Loading…

Worked solutions

IB Computer Science HL · Paper 2

Computer Science, Higher Level

Session: Nov 2023
Marks: 260
Duration: 1 hour 20 min
Calculator: Not allowed
Questions: 18

Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9 Q10 Q11 Q12 Q13 Q14 Q15 Q16 Q17 Q18

Keep your copy of the paper open alongside. We don't reproduce the questions: each part below is labelled to match the Nov 2023 Global paper, which your school or the IB can provide.

Practice mode

Hide our working until you've had a go.

Option A: Databases · Option

Questions 1–4 · 65 marks

Question 1

15 marks

Restaurant ordering database structure, date formats, composite keys, and SQL queries

Databases

(a)

4 marks

1. Data integrity and normalization: A relational database enforces structural constraints and referential integrity across related tables, ensuring that customer and meal records remain consistent and free of redundant anomalies.
2. Concurrent multi-user access: A database management system supports record locking and transaction management so multiple restaurant staff members can enter orders and update stock simultaneously without corrupting shared files (unlike a flat spreadsheet loaded entirely into memory).

Mentor tip

Provide two distinct reasons, each paired with a clear expansion on how the database improves operations compared to a spreadsheet.

(b)

2 marks

Storing dates in the ISO 8601 format (`yyyy/mm/dd`) ensures that chronological sorting matches alphabetical/lexicographical sorting (ordering by year first, then month, then day).
It also facilitates seamless data integration and compatibility across international database systems without ambiguity between day-first and month-first conventions.

Mentor tip

Highlight how year-month-day order allows standard string or numeric sorting to naturally place dates in true chronological sequence.

(c)

3 marks

Combining `OrderID` and `ItemID` forms a natural composite key because each distinct meal item appears at most once per customer order (with `Quantity` tracking multiple portions).
Using existing foreign keys eliminates the need to generate and store an arbitrary surrogate `OrderlineID` column, saving disk space and avoiding a meaningless attribute that would never be searched directly.

Mentor tip

Explain both why the pair is unique per order line and how omitting a separate surrogate key saves storage.

(d)

2 marks

When joining `ORDERLINE` and `ITEM` in a query, referencing `UnitPrice` without a table prefix creates an ambiguous column name conflict.
The query parser will fail to execute and return a syntax/ambiguity error unless the attribute is explicitly qualified using dot notation (such as `ORDERLINE.UnitPrice` or `ITEM.UnitPrice`).

Mentor tip

Mention column name ambiguity during multi-table queries and the requirement to qualify field names with their table names.

(e)

4 marks

1. Specify the required output columns: `CUSTOMER.FirstName`, `CUSTOMER.Surname`, and `ORDER.OrderDate`.
2. Identify the source tables: `CUSTOMER` and `ORDER`.
3. Join the tables on their common key: `CUSTOMER.CustomerID = ORDER.CustomerID`.
4. Filter the records using a `WHERE` clause matching both name fields:
`SELECT CUSTOMER.FirstName, CUSTOMER.Surname, ORDER.OrderDate FROM CUSTOMER INNER JOIN ORDER ON CUSTOMER.CustomerID = ORDER.CustomerID WHERE CUSTOMER.FirstName = 'Akemi' AND CUSTOMER.Surname = 'Watene';`

Mentor tip

You can express the query steps either in SQL syntax or clear structured English, ensuring the join condition and both filter criteria are explicit.

Question 2

15 marks

Derived attributes, database transactions, validation vs verification, concurrency, and recovery

Databases

(a)

3 marks

`LinePrice` is a derived (calculated) attribute that can always be computed on the fly as `Quantity * UnitPrice`.
Storing a calculated value violates Third Normal Form (3NF) by introducing transitive dependency, wastes storage space, and risks update anomalies if `Quantity` or `UnitPrice` is modified without recalculating `LinePrice`.

LinePrice = Quantity \times UnitPrice

Mentor tip

Connect derived fields to 3NF compliance, storage efficiency, and prevention of calculation discrepancies.

(b)

1 mark

A database transaction is a single logical unit of work comprising one or more database operations that must be executed atomically (either completed in full or rolled back entirely) to transition the database from one consistent state to another.

Mentor tip

Emphasize that a transaction is an all-or-nothing logical unit of work.

(c)(i)

1 mark

Apply a range check (for example, ensuring `hours_worked` is between `0` and `16` hours per shift) or a numeric type check.

Mentor tip

Validation is an automated rule check performed by software on input (range check, type check, presence check).

(c)(ii)

2 marks

Use double data entry (requiring the shift manager to type the `payrate` twice and comparing both entries for an exact match) or visual proofreading against an authoritative paper contract / lookup schedule.

Mentor tip

Do not confuse validation with verification: verification confirms that data matches the original source (e.g., double entry or visual comparison).

(d)

4 marks

The DBMS applies concurrency control such as record-level locking (or two-phase locking) when the shift manager initiates an edit on the employee's shift row.
While the write lock is active, the employee's session is either blocked from reading/confirming the uncommitted record (preventing a dirty read) or shown a notification that the schedule is currently being updated.
Alternatively, Multi-Version Concurrency Control (MVCC) can serve the employee a consistent read-only snapshot of the prior committed hours while the manager's transaction is in progress, only allowing confirmation once the write lock is released.

Mentor tip

Explain how record locking or MVCC isolates the manager's in-progress edit from the employee's read/confirmation request.

(e)

4 marks

Instant recovery (recovery in place / shadow paging with virtual snapshots): The system maintains periodic consistent snapshots of the database state on a backup virtual machine or mirror.
If a crash occurs, user workloads and new write operations are immediately redirected to the clean snapshot VM so operations continue seamlessly.
Meanwhile, the background recovery protocol inspects interrupted transactions on the primary instance, rolling back incomplete changes and syncing updated blocks before switching traffic back.

Mentor tip

Describe either snapshot-based instant recovery (failover to a backup image while restoring in the background) or checkpoint/rollback crash recovery.

Question 3

15 marks

European pet passport ERD, secondary keys, 3NF normalization, and data protection

Databases

(a)

2 marks

1. `OWNER` to `PET` is a one-to-many ( $1 : M$ ) relationship (one pet owner can own multiple pets).
2. `PET` to `PASSPORT` is a one-to-one ( $1 : 1$ ) relationship (each pet holds exactly one individual passport).
Diagram notation: `OWNER (1) ----< (M) PET (1) ---- (1) PASSPORT`

Mentor tip

Clearly label both cardinalities: 1-to-many between Owner and Pet, and 1-to-1 between Pet and Passport.

(b)

1 mark

A secondary key is a non-primary attribute (which may or may not be unique) that is indexed to speed up searching, sorting, and querying records in a table.

Mentor tip

State that a secondary key is an indexed field used for faster lookup/retrieval rather than primary entity identification.

(c)

8 marks

Decompose the unnormalized passport and vaccination data into four additional 3NF relations alongside `VET`: `OWNER`, `PET`, `PASSPORT`, `VACCINE`, and `PETVACCINATION` (or combine 1:1 `PET`/`PASSPORT` with separate `OWNER` or `VACCINE` tables):
1. `PET(TransponderCode, PetName, Species, Breed, Sex, DOB)` — Primary Key: `TransponderCode`.
2. `PASSPORT(PassportID, TransponderCode*, VetID*, DateOfIssue)` (or using `TransponderCode` as PK) — Foreign Keys: `TransponderCode` referencing `PET`, `VetID` referencing `VET`.
3. `VACCINE(VaccineID, VaccineName, Manufacturer, BatchNumber)` — Primary Key: `VaccineID` (separating vaccine batch details from individual pet injections).
4. `PETVACCINATION(VaccinationID, TransponderCode*, VaccineID*, VetID*, VaccineDate, ValidFrom, ValidUntil)` — Primary Key: `VaccinationID` (or composite key `TransponderCode, VaccineID, VaccineDate`), with Foreign Keys linking the pet, the vaccine batch, and the administering vet.

Mentor tip

Ensure vaccine product/manufacturer details are separated into their own `VACCINE` table and linked via a junction/event table `PETVACCINATION` so repeating vaccination stickers are in 3NF.

(d)

4 marks

1. Purpose limitation and consent: Personal contact details of the pet owner must only be collected and processed for legitimate pet passport and veterinary traceability purposes, never shared with third parties without consent.
2. Cross-border transfer safeguards: Because the passport is used across European Union countries, owner data must only be transferred to jurisdictions that enforce equivalent data protection standards (such as GDPR).
3. Accuracy and right of access: Records must be kept accurate and up to date, and pet owners have the legal right to inspect and correct their stored personal data.
4. Storage limitation and security: Data must be protected against unauthorized access and not retained longer than necessary.

Mentor tip

Reference core data protection principles (purpose limitation, consent, accuracy, retention limits, and secure cross-border sharing).

Question 4

20 marks

Emergency Management Information Systems (EMIS), spatial databases, and spatial data mining

Databases

(a)(i)

2 marks

Spatial data represents the geometric location, shape, and topological properties of physical features on Earth.
It is stored using geographic coordinate systems such as latitude and longitude pairs.

Mentor tip

Mention both geographic coordinates (latitude/longitude) and geometric/topological representation of locations.

(a)(ii)

2 marks

1. Points (e.g., individual emergency call coordinates).
2. Lines / Polylines (e.g., road networks or highway segments) or Polygons (e.g., city districts or patrol zones).

Mentor tip

Any two standard spatial primitives—Points, Lines, or Polygons—earn full credit.

(b)

1 mark

One-to-one ( $1 : 1$ ) relationship (one map coordinate point links to one specific table record) or one-to-many ( $1 : M$ ) relationship (one location point links to multiple reported incident rows).

Mentor tip

State a standard relational cardinality such as one-to-one or one-to-many.

(c)

6 marks

1. Density clustering / Heat map generation (3 marks): Spatial clustering groups nearby incident coordinates and renders a colour-coded density gradient (e.g., red for high concentration, yellow for low) using classification methods like quantiles or natural breaks. This highlights high-risk crime hotspots more clearly than overlapping dots, guiding patrol allocation.
2. Attribute or temporal filtering / Association analysis (3 marks): Filtering incidents by crime category (e.g., burglary vs assault) or time window (e.g., night shifts or annual public events) isolates specific patterns and removes confounding noise like general population density, helping dispatchers pre-position emergency vehicles.

Mentor tip

Explain two distinct spatial analysis techniques (3 marks each), detailing both how the map view changes and how emergency planners use that insight.

(d)

5 marks

1. Risk of manipulation and public deception: Viewers often perceive maps as objective truth without realizing that altering colour scales, dot sizes, or classification thresholds can exaggerate or downplay crime severity to serve a political agenda.
2. Beneficial persuasion and clarity: Visual tools can legitimately simplify complex multidimensional datasets for decision-makers, and highlighting all incidents (including minor ones) may justify increased funding for public safety and emergency response.
3. Balanced conclusion: Using visual analysis tools is ethical provided maps include transparent legends, objective metadata, and clear disclosure of filtering criteria so audiences can critically evaluate competing perspectives.

Mentor tip

Include negative ethical risks (bias/deception), positive uses (funding/clarity), and a reasoned conclusion emphasizing transparency in legends and metadata.

(e)

4 marks

Deviation detection compares incoming spatial and temporal sensor/call data against historical baselines to identify statistically significant anomalies outside normal patterns.
For example, an unusual spike in emergency calls or temperature/smoke readings clustered in a specific zone during dry seasons flags an emerging fire hazard.
By applying a statistical confidence threshold (such as 95%) to filter out random background noise, the system alerts authorities to deploy preventative fire units early.

Mentor tip

Connect historical baseline comparison, spatial/temporal clustering of outliers, and noise filtering via confidence thresholds.

Option B: Modelling and simulation · Option

Questions 5–8 · 65 marks

Question 5

16 marks

Cattery5S optimal stopping problem, permutation tables, and spreadsheet modelling

Modelling and simulation

(a)

1 mark

`0.5` (or `50%` / `1/2`).

P (best) = \frac{1}{2} = 0.5

Mentor tip

With two cats and no prior information, either choice has an equal 50% chance of being the higher-rated cat.

(b)

1 mark

Complete rows 4, 5, and 6 with the remaining three permutations of `{1, 2, 3}`:
Row 4: `2 | 3 | 1`
Row 5: `3 | 1 | 2`
Row 6: `3 | 2 | 1`

Mentor tip

Ensure all 6 permutations of (1, 2, 3) are present without duplicating any row.

(c)

2 marks

Use an `IF` function comparing the second cat in column B (`B1`) against the first cat in column A (`A_1`, written in spreadsheet syntax without underscore):
`=IF(B1 > A_1, B1, C1)` (where `A_1` denotes cell column A, row 1)
If the second cat rating exceeds the first cat rating, return `B1`; otherwise, return the third cat `C1`.

Mentor tip

Check that the condition compares `B1` to the first cat in column A row 1, returning `B1` if true and `C1` if false.

(d)(i)

2 marks

Use `COUNTIF` with an absolute row reference for the outcome range `D $1 : D$ 6` and a criteria value (such as `1`, `2`, `3` or a cell reference `E1` containing the target rating):
`=COUNTIF( $D$ 1: $D$ 6, E1)` (or `=COUNTIF(D $1 : D$ 6, 3)` for rating 3).

Mentor tip

Include dollar signs (` $D$ 1: $D$ 6`) to lock the range when copying the counting formula across the three cat ratings.

(d)(ii)

3 marks

Evaluate the chosen cat in column D across all 6 permutations:
• Row 1 (`1, 2, 3`): $2 > 1 ⟹$ chooses `2`.
• Row 2 (`1, 3, 2`): $3 > 1 ⟹$ chooses `3`.
• Row 3 (`2, 1, 3`): $1 < 2 ⟹$ moves to C, chooses `3`.
• Row 4 (`2, 3, 1`): $3 > 2 ⟹$ chooses `3`.
• Row 5 (`3, 1, 2`): $1 < 3 ⟹$ moves to C, chooses `2`.
• Row 6 (`3, 2, 1`): $2 < 3 ⟹$ moves to C, chooses `1`.
Frequencies out of 6: Rating 1 is chosen 1 time ( $1/6 \approx 16.7%$ ); Rating 2 is chosen 2 times ( $2/6 \approx 33.3%$ ); Rating 3 (best cat) is chosen 3 times ( $3/6 = 50%$ ).

P (1) = \frac{1}{6} \approx 16.7%, P (2) = \frac{2}{6} \approx 33.3%, P (3) = \frac{3}{6} = 50%

Mentor tip

State the probability or percentage for each of the three ratings (16.7%, 33.3%, and 50%).

(e)

1 mark

Nest a second `IF` statement to check the third cat (`C1`) against the first cat in column A row 1 (`A_1`) if the second cat (`B1`) is lower, defaulting to the fourth cat (`D1`):
`=IF(B1 > A_1, B1, IF(C1 > A_1, C1, D1))` (where `A_1` denotes cell column A, row 1)

Mentor tip

Both comparisons (`B1` and `C1`) benchmark against the first cat in column A, row 1.

(f)

6 marks

1. Advantages (up to 3 marks): A spreadsheet automates repetitive conditional evaluations without manual arithmetic errors, visually displays the exact chosen cat for every permutation row, supports quick charting of probability distributions, and allows easy tweaking of decision formulas.
2. Disadvantages (up to 3 marks): Spreadsheets do not natively generate ordered permutations, requiring manual entry of rows ( $N!$ grows factorially from 6 to 24 to 40,320). Deeply nested `IF` formulas become unwieldy and error-prone as the number of cats increases.
3. Conclusion: A spreadsheet model is effective and transparent for small problem sizes (3 or 4 cats), but becomes impractical for larger numbers where a programmatic algorithm is required.

Mentor tip

Balance advantages (visual clarity, automatic recalculation) with disadvantages (manual permutation entry, nested IF complexity as N! grows) and finish with a clear conclusion.

Question 6

16 marks

Algorithmic simulation of 8-cat and 20-cat selection strategies and hardware scaling

Modelling and simulation

(a)

2 marks

Initialize an accumulator `TOTAL = 1` and multiply it by each integer from `1` to `N` in a loop (or use recursion with base case `N <= 1` returning `1`):
```text function FACTORIAL(N) TOTAL = 1 loop X from 1 to N TOTAL = TOTAL * X end loop return TOTAL end function ```

N! = x = 1 \prod N x

Mentor tip

Either an iterative loop multiplying from 1 to N or a recursive function with a base case of N = 1 is valid.

(b)

2 marks

Initialize `HIGHEST` to `ARR[0]` (or `-1`) and compare across the first three elements (indices `0` to `2`):
```text function FINDHIGH(ARR) HIGHEST = ARR[0] loop I from 1 to 2 if ARR[I] > HIGHEST then HIGHEST = ARR[I] end if end loop return HIGHEST end function ```

Mentor tip

Remember that 0-indexed arrays store the first three cats at indices 0, 1, and 2.

(c)

6 marks

Iterate through all `FACTORIAL(8)` permutations, find the max of the first 3 cats using `FINDHIGH(ARR)`, scan from index `3` to `7` for the first cat exceeding `HIGHEST` (or default to the last cat `ARR[7]`), count how often rating `8` (the best cat) is selected, and advance `ARR` via `NEXTPERMUTATION(ARR)`:
```text TOTAL_BEST = 0 NUM_PERMS = FACTORIAL(8) loop ROW from 0 to NUM_PERMS - 1 HIGHEST = FINDHIGH(ARR) FOUND = false POS = 3 loop while FOUND = false AND POS < 8 if ARR[POS] > HIGHEST then CHOICE = ARR[POS] FOUND = true end if POS = POS + 1 end loop if FOUND = false then CHOICE = ARR[7] end if if CHOICE = 8 then TOTAL_BEST = TOTAL_BEST + 1 end if ARR = NEXTPERMUTATION(ARR) end loop PROBABILITY = (TOTAL_BEST / NUM_PERMS) * 100 output PROBABILITY ```

Mentor tip

Ensure you handle the fallback case (`if FOUND = false then CHOICE = ARR[7]`) and check `if CHOICE = 8` before advancing to `NEXTPERMUTATION(ARR)`.

(d)

2 marks

Modify the threshold number of initial cats rejected (for example, test rejecting the first 2 cats or the first 4 cats instead of 3).
Compare the resulting probabilities of selecting the best cat across each threshold to verify which rejection count maximizes success.

Mentor tip

State both what parameter to vary (the number of initially rejected cats) and how to evaluate the outcome (comparing the resulting success probabilities).

(e)

4 marks

Evaluating $20! \approx 2.4 \times 1 0^{18}$ permutations is computationally intractable on a standard single-CPU laptop due to combinatorial explosion.
It requires high-performance parallel processing architecture—such as multi-core supercomputing clusters, GPUs adapted for massive parallel thread execution, or dedicated ASICs/math co-processors—alongside substantial high-speed RAM to manage state across parallel workers.

20! \approx 2.43 \times 1 0^{18}

Mentor tip

Link the $2.4 \times 1 0^{18}$ scale to parallel processing architectures (multi-CPU clusters, adapted GPU compute, high clock speed, and large RAM).

Question 7

13 marks

2D map visualizations vs 3D building walkthroughs for emergency response

Modelling and simulation

(a)

1 mark

A 2D visualization is a flat, two-dimensional graphical representation (using horizontal and vertical axes) of data, objects, or processes.

Mentor tip

Define it clearly as a flat (two-axis / non-3D) visual representation of data or objects.

(b)(i)

4 marks

1. Alignment with geographic coordinates: City maps naturally use 2D latitude and longitude coordinates, allowing GPS locations of emergency calls to be plotted accurately without needing altitude.
2. Visual clarity and rapid interpretation: A 2D map avoids 3D perspective distortion and occlusion, enabling dispatchers to read street layouts and incident densities at a glance.

Mentor tip

Give two distinct reasons with expansions: 2D GPS coordinate mapping and visual simplicity/readability.

(b)(ii)

2 marks

Because a 2D map only stores flat coordinate vectors/pixels without complex 3D polygon meshes, z-buffering, or volumetric lighting textures, it requires minimal RAM and video memory.

Mentor tip

Contrast the low memory footprint of flat 2D graphics with the heavy memory demands of 3D rendering.

(c)

6 marks

1. Rendering speed vs urgency (2 marks): Because people are trapped in an active fire, there is no time for slow photorealistic ray tracing; fast wireframe or scanline rendering must be chosen to generate the model immediately.
2. File size and wireless transmission (2 marks): High-detail 3D animations produce huge files that may buffer or fail over mobile networks; a lightweight low-polygon model ensures instant transmission to handheld tablets.
3. Functional detail vs visual realism (2 marks): Rescue personnel only need structural navigation cues (corridors, stairwells, exits, and hazards) rather than realistic surface textures or shadows.

Mentor tip

Outline three distinct considerations (time/rendering method, file size/bandwidth, and required level of functional detail) for 2 marks each.

Question 8

20 marks

Genetic algorithms for keyboard layout optimization and natural language processing (NLP)

Modelling and simulation

(a)(i)

2 marks

Increasing the initial population size increases genetic diversity (variation of keyboard chromosomes), which reduces the risk of converging on a poor local optimum (or increases the computation time required per generation).

Mentor tip

Mention increased chromosome diversity and its effect on either solution quality or computational time per generation.

(a)(ii)

3 marks

The search space of all possible keyboard permutations ( $26!$ ) is astronomically large, so purely random initial layouts will mostly have terrible typing efficiency.
Seeding the initial population with known ergonomic layouts (such as Dvorak or Colemak) starts the search much closer to an optimal region, significantly reducing the number of generations needed to converge.

Mentor tip

Connect the vast size of the permutation space to the benefit of seeding with existing high-performing layouts (like Dvorak) to speed up convergence.

(b)

3 marks

The fitness function evaluates each individual keyboard layout (for example, by calculating total finger travel distance or typing time across a corpus of common English text) and assigns a numerical score.
Layouts with higher fitness scores have a higher probability of being selected as parents to pass their key arrangements to the next generation.

Mentor tip

Explain both what the fitness score measures (closeness to the optimal typing target) and how it drives parent selection for reproduction.

(c)

2 marks

Taking the left half of keys from Parent 1 and the right half from Parent 2 via single-point crossover would result in duplicate letters (e.g., two 'E' keys) and missing letters, producing an invalid keyboard layout.

Mentor tip

Point out that standard single-point crossover violates the permutation constraint by duplicating some letters and omitting others.

(d)

4 marks

1. Massive increases in processing power and GPU hardware acceleration.
2. Availability of Big Data (vast corpora of recorded human speech and linguistic datasets for training and testing).
3. Advancements in deep learning and neural network architectures.
4. Deeper computational modelling of human phonology, syntax, and semantics.

Mentor tip

List four concise factors: computing power, large speech datasets (Big Data), advanced ML/AI algorithms, and linguistic modeling.

(e)

4 marks

Human language is inherently ambiguous, relying on shared real-world context, tone, emotion, or sarcasm to convey meaning beyond literal word definitions.
Without explicit contextual cues across phonological, semantic, and pragmatic levels, a machine-learning model may misinterpret homonyms or figurative expressions.
Training supervised models to handle context requires massive, richly annotated datasets covering diverse conversational scenarios, and unsupervised models often struggle to infer pragmatic intent.

Mentor tip

Discuss ambiguity, pragmatic factors like sarcasm/emotion, and the difficulty of capturing real-world context in training datasets.

(f)

2 marks

Machine translation relies on large parallel corpora of translated texts; far more bilingual training data exists between English and French than directly between Bengali and Polish (often forcing translation through English as an intermediate pivot language, compounding errors).
Additionally, English and French share the Latin alphabet and closer grammatical roots, whereas Bengali and Polish use completely different scripts and grammatical structures.

Mentor tip

Mention either the scarcity of parallel training datasets (requiring an English intermediate step) or the difference in alphabets and linguistic structures.

Option C: Web science · Option

Questions 9–13 · 65 marks

Question 9

18 marks

Friends for Dogs webpage components, HTTP characteristics, PHP server-side processing, and templates

Web science

(a)

2 marks

1. Text input box / HTML form (`<input type="text">`) and submit button.
2. Images of dogs, hyperlinks, page header/banner, or body text paragraphs.

Mentor tip

Name any two visible or structural webpage elements such as forms/buttons, images, headers, or scripts.

(b)

2 marks

1. Stateless protocol: Each HTTP request is processed independently without retaining session memory of previous requests.
2. Request-response client-server model (using methods like `GET` and `POST`, status codes, and plain-text transmission over TCP port 80 by default).

Mentor tip

State any two core HTTP features: statelessness, client-server request/response methods (GET/POST), unencrypted text messaging, or status codes.

(c)

4 marks

1. The script retrieves the submitted `dogname` value from the HTTP `POST` request array (`$_POST["dogname"]`).
2. It opens a connection to the `dogDB` MySQL database on `localhost` using the server credentials.
3. It constructs and executes a `SELECT` query on the `Dogs` table to fetch `dname`, `image`, and `price` matching `$dogname`.
4. If one or more matching rows exist, it loops through the result set and outputs each dog's image, name, and price; otherwise it outputs `"not found"`, and finally closes the database connection.

Mentor tip

Trace the PHP script chronologically: read POST input -> connect to DB -> run SELECT query -> loop/echo results or 'not found' -> close connection.

(d)

2 marks

Server-side execution protects sensitive database credentials (` $u ser nam e ‘ an d ‘$ pass`) by keeping them hidden on the server rather than exposing them in client-viewable source code.
It also allows direct querying of the centralized `dogDB` database residing on the server without transferring the entire database to the client browser.

Mentor tip

Emphasize hiding database login credentials from the user or interacting directly with the server-hosted database.

(e)

3 marks

Lossy compression significantly reduces image file sizes compared to lossless formats.
Smaller file sizes reduce bandwidth consumption and server storage costs, allowing dog photos to download and render much faster for users browsing or sharing them.
Although minor pixel detail is permanently discarded, the visual degradation is imperceptible for casual web viewing.

Mentor tip

Connect smaller file size to faster download speeds/lower bandwidth and tolerable visual quality for web photos.

(f)

5 marks

1. Advantages (up to 3 marks): Templates provide a consistent, familiar navigation layout across all pages, speed up development and lower costs by reusing pre-built structural modules (headers, footers), and often feature pre-tested responsive code.
2. Disadvantages (up to 3 marks): Templates restrict creative flexibility and customization, may include unnecessary code bloat from unused features, and can make the Friends for Dogs website look generic and indistinguishable from other template-based sites.

Mentor tip

Weigh both advantages (consistency, rapid/cheaper development, modular reuse) and disadvantages (limited customization, generic appearance, code bloat).

Question 10

14 marks

Website interactivity, copyright compliance, search ranking metrics, and white hat SEO

Web science

(a)

2 marks

1. Add online booking forms or live chat support so visitors can reserve tours and ask questions in real time.
2. Incorporate interactive maps, customer rating/review sections, or virtual 360-degree tour previews.

Mentor tip

Identify any two interactive features relevant to a travel website (booking forms, live chat, customer reviews, interactive maps).

(b)

2 marks

Developers can create original photographs, text, and code in-house, or obtain explicit licensing/permission from copyright holders before using external media (such as adhering to Creative Commons attribution and non-commercial terms).

Mentor tip

Mention either creating original media or obtaining proper licenses/permissions (including Creative Commons compliance) for third-party assets.

(c)

1 mark

The number and authority of incoming backlinks (as used in PageRank/HITS), page load speed, user dwell time, or click-through rate.

Mentor tip

State one specific search ranking metric such as backlink quality/quantity, page load time, or visitor dwell time.

(d)

2 marks

1. Including accurate, relevant keywords in `<meta>` tags, headings, and high-quality body content.
2. Earning natural backlinks from reputable travel directories and optimizing site structure for mobile responsiveness.

Mentor tip

List two ethical (white hat) SEO methods: relevant keywords/meta tags, quality original content, reputable backlinks, or mobile optimization.

(e)

4 marks

1. Long-term sustainability and penalty avoidance: White hat SEO complies with search engine guidelines, preventing the travel site from being down-ranked or banned (delisted) during algorithm updates.
2. Enhanced user experience and credibility: Focusing on clear navigation, fast loading, and genuine travel content builds customer trust, increases visitor retention, and drives organic bookings.

Mentor tip

Explain two reasons (2 marks each): avoiding search engine penalties/de-indexing and building sustainable user trust/engagement.

(f)

3 marks

If search engine developers assume that webpage importance is strictly proportional to the number of incoming backlinks, the algorithm will prioritize established, heavily linked domains.
Consequently, newly published or niche websites containing highly accurate, relevant information may be buried in search results because they have not yet accumulated backlinks, reducing the diversity and freshness of surfaced content.

Mentor tip

Link a specific design assumption (e.g., equating backlinks with quality, or static indexing intervals) to its effect on search result accuracy and diversity.

Question 11

13 marks

Client-server architecture, identification vs authentication, cloud storage, and web decentralization

Web science

(a)

2 marks

A server is a centralized host that manages, stores, and provides shared resources or services across the network.
A client is an end-user workstation or device that initiates requests to the server to access those shared services or files.

Mentor tip

Clearly contrast the server's role (providing/hosting resources) with the client's role (requesting services).

(b)

2 marks

Identification is the act of claiming an identity (for example, entering a username or employee ID).
Authentication is the process of verifying that claim using secret credentials or evidence (for example, checking a password, biometric scan, or security token).

Mentor tip

Remember: Identification states *who you claim to be* (username); authentication *proves it* (password/biometrics).

(c)

3 marks

Scalability and cost efficiency: Moving to cloud storage eliminates the capital expense of purchasing, powering, and maintaining on-premises file servers.
The training company can dynamically scale storage capacity up or down on a pay-as-you-go basis while enabling employees to access and share training materials remotely from any internet-connected device.

Mentor tip

Develop one core reason thoroughly—such as hardware/maintenance cost reduction, elastic scalability, ubiquitous remote access, or automated disaster recovery.

(d)

6 marks

1. Arguments that cloud storage supports decentralization: Cloud infrastructure distributes data across geographically dispersed data centres globally and allows small organizations and individuals to publish services and share files from anywhere without owning local server hardware.
2. Arguments that cloud storage increases centralization (ownership concentration): Despite geographic distribution, cloud infrastructure is dominated by a small oligopoly of mega-providers (such as AWS, Google Cloud, Azure). Businesses migrating away from independent local servers to these few platforms consolidate control, create single points of failure/cyber-attack targets, and lock users into proprietary ecosystems with high network effects.
3. Conclusion: While cloud storage decentralizes physical access and workplace collaboration, it has largely acted as a catalyst for *re-centralizing* web infrastructure ownership until peer-to-peer or blockchain storage alternatives achieve mainstream adoption.

Mentor tip

Distinguish between *geographic/access distribution* (access from anywhere) and *ownership concentration* (control by a few tech giants) to build a strong 6-mark evaluation.

Question 12

10 marks

Bow-tie web graph structure, strongly connected core (SCC), tubes vs tendrils, and power laws

Web science

(a)

2 marks

A web graph is a directed graph representing the entire World Wide Web, where vertices represent webpages and directed edges represent hyperlinks between them.
A sub-graph represents a smaller, localized subset of nodes and links extracted from the overall web graph (often focused on a specific domain or topic).

Mentor tip

Contrast the full World Wide Web directed graph with a smaller subset of pages/links.

(b)

2 marks

1. Mutual reachability: Every webpage within the Strongly Connected Core (SCC) can be reached from every other webpage in the SCC via a directed path of hyperlinks.
2. High link density and resilience: The SCC forms the central, densely interlinked backbone of the web graph with multiple redundant pathways between nodes.

Mentor tip

Highlight that any node in the SCC can reach any other node in the SCC via directed links.

(c)

2 marks

A tube is a directed pathway of webpages that connects the `IN` component directly to the `OUT` component while bypassing the Strongly Connected Core (SCC) entirely.
A tendril is a dead-end branch that either leads out from the `IN` component without ever reaching the SCC, or leads into the `OUT` component from nowhere without being reachable from the SCC.

Mentor tip

A tube bridges IN to OUT bypassing the SCC; a tendril branches off IN (or into OUT) without connecting the two.

(d)

2 marks

Under a power-law (scale-free / rich-get-richer) distribution, a tiny fraction of dominant hub websites attract the vast majority of new incoming links and high PageRank visibility.
As user traffic and links concentrate around these mega-platforms, smaller independent sites either lose reciprocal links or become obsolete, shrinking the relative proportion of pages that remain mutually interlinked in the main SCC.

Mentor tip

Connect power-law preferential attachment ('rich get richer' via high PageRank) to link concentration on major hubs and the shrinking of the broader SCC.

(e)

2 marks

Sites like `wordpress.org`, `youtube.com`, and `wikipedia.org` are massive, authoritative content hubs or publishing platforms that provide embeddable widgets, reference articles, and template footer links.
Because millions of external blogs and websites naturally cite Wikipedia articles, embed YouTube videos, or use WordPress themes, they accumulate an enormous volume of incoming backlinks.

Mentor tip

Explain that these sites act as authoritative reference hubs or content/CMS platforms that millions of external websites embed or cite.

Question 13

10 marks

Ambient intelligence in smart homes for dementia care and online health data behaviour

Web science

(a)

4 marks

1. Sensor input: Non-intrusive IoT sensors embedded around the smart home continuously monitor the patient's vital signs (body temperature, heart rate), movement/fall detection, and room conditions (temperature, door state, stove usage).
2. Real-time processing: The ambient intelligence system compares live sensor streams against the patient's baseline behavioural profile to identify anomalies (such as wandering at night or forgetting medication).
3. Automated output and alerts: The environment automatically adjusts lighting/heating, plays audio/visual reminders for medication, and dispatches immediate alerts to caregivers or emergency services if a fall or hazardous deviation is detected.

Mentor tip

Structure your explanation using the Input (sensors) -> Processing (anomaly detection against baseline) -> Output (automated adjustments/caregiver alerts) framework.

(b)

6 marks

1. Positive behavioural shifts (empowerment and prevention): Patients actively research symptoms, treatments, and preventive lifestyle habits online, improving health literacy. Online support communities and patient forums foster collective engagement, and patients enter medical consultations better informed.
2. Negative behavioural shifts (self-diagnosis and anxiety): The overwhelming volume of unverified or commercialized medical claims on the web can trigger information overload, cyberchondria (health anxiety), and dangerous self-diagnosis or self-medication with unproven remedies.
3. Systemic impact and conclusion: While online health data has transformed patients from passive recipients into proactive participants, disparities in digital literacy and the prevalence of misinformation mean professional medical guidance remains essential.

Mentor tip

Discuss both constructive changes (health literacy, support forums, preventive care) and risks (self-diagnosis, misinformation, anxiety, doctor-patient friction).

Option D: Object-oriented programming · Option

Questions 14–18 · 65 marks

Question 14

11 marks

Monster Battle class definitions, static constants, encapsulation, and inheritance constructors

Object-oriented programming

(a)

1 mark

The `static` keyword means the variable belongs to the `Monster` class itself rather than to any individual object instance, so a single shared copy exists in memory and can be accessed via `Monster.ICE_MONSTER_TYPE` without instantiating an object.

Mentor tip

Do not confuse `static` with `final`: `static` means class-level/shared across all instances, whereas `final` means constant.

(b)(i)

2 marks

Declaring instance variables `private` enforces encapsulation (data hiding), preventing external classes from directly modifying or corrupting a monster's state (such as setting `health` to an invalid value) outside controlled methods.

Mentor tip

Link `private` visibility to encapsulation/data hiding and protection against accidental external modification.

(b)(ii)

2 marks

1. Accessor methods (getters).
2. Mutator methods (setters).

Mentor tip

State both accessor (getter) and mutator (setter) methods.

(c)

2 marks

The `super(...)` call invokes the constructor of the parent superclass (`Monster`), passing `name`, `health`, and `Monster.WATER_MONSTER_TYPE` to initialize the inherited attributes.

Mentor tip

State both that `super` refers to the superclass (`Monster`) and that it calls the superclass constructor to initialize inherited fields.

(d)

2 marks

1. Transparency and peer review: Making source code publicly accessible so developers worldwide can inspect, audit, and improve security and reliability.
2. Affordability, flexibility, and perpetuity: Providing free-to-use software that users can modify, localize, and maintain without vendor lock-in.

Mentor tip

Identify any two aims of open source: transparency, affordability, flexibility/customizability, interoperability, or community security auditing.

(e)

2 marks

Instantiate a `FireMonster` object (or `Monster` with `FIRE_MONSTER_TYPE`), passing `"Flamey"` and `40` to the constructor:
```java FireMonster flamey = new FireMonster("Flamey", 40); ```

Mentor tip

Ensure the variable type, `new FireMonster(...)` constructor call, and parameter order `("Flamey", 40)` match the class definition.

Question 15

19 marks

Monster attack logic, class vs instance distinction, UML relationships, and Arena array management

Object-oriented programming

(a)

4 marks

Call `rollDice()` to compute the damage, subtract it from `defender.health`, output the remaining health if `> 0`, or clamp `defender.health` to `0` and output the elimination message:
```java public void attack(Monster defender) { int damage = rollDice(); defender.health = defender.health - damage; if (defender.health > 0) { output(defender.name + " sustained " + damage + " damage and has " + defender.health + " health points."); } else { defender.health = 0; output(defender.name + " must leave the game"); } } ```

Mentor tip

Store the result of `rollDice()` in a local `damage` variable so you can both subtract it from `defender.health` and display it in the output message.

(b)

4 marks

1. Blueprint vs concrete entity: The `FireMonster` class is the static code template/blueprint defining attributes and behaviours, whereas an instance is a concrete object created at runtime with its own specific attribute values (such as `"Flamey"` with `40` health).
2. Multiplicity and memory allocation: There is only one `FireMonster` class definition in the program, whereas multiple distinct `FireMonster` instances can be allocated in heap memory simultaneously.

Mentor tip

Contrast template/blueprint vs runtime object with specific state, and single class definition vs multiple memory-allocated instances.

(c)

2 marks

1. `FireMonster` to `Monster`: Draw an inheritance arrow (closed, unfilled triangle pointing from `FireMonster` to `Monster`) representing an **is-a** relationship.
2. `Arena` to `Monster`: Draw an aggregation connector (open diamond on the `Arena` side connected to `Monster`) representing a **has-a** relationship:
`[Arena] <>---- [Monster] <|---- [FireMonster]`

Mentor tip

Use a hollow triangle pointing to `Monster` for inheritance (`FireMonster` is a `Monster`) and a diamond on `Arena` for aggregation (`Arena` has `Monster`s).

(d)

5 marks

Check if `monsterCount < monsters.length` (the arena is not full), search indices `0` to `monsterCount - 1` to verify `M` is not already present, insert `M` at `monsters[monsterCount]` and increment `monsterCount`, or output the specific reason for rejection:
```java public void addMonster(Monster M) { if (monsterCount >= monsters.length) { output("Arena is full"); } else { boolean exists = false; for (int i = 0; i < monsterCount; i++) { if (monsters[i] == M) { exists = true; } } if (!exists) { monsters[monsterCount] = M; monsterCount++; } else { output(M.name + " is already in the Arena"); } } } ```

Mentor tip

Include two distinct `output(...)` messages: one when the array is full (`monsterCount == monsters.length`) and another when `M` is already in the array.

(e)

4 marks

1. Increased complexity and code size for simple tasks: Designing class hierarchies, constructors, and encapsulation adds structural overhead that is unnecessary for small, straightforward scripts.
2. Runtime memory and execution overhead: Instantiating, referencing, and garbage-collecting numerous objects at runtime can consume more memory and execute slightly slower than procedural code.

Mentor tip

Give two distinct disadvantages (2 marks each): design/code complexity for small programs and runtime memory/speed overhead from object instantiation.

Question 16

15 marks

Polymorphism, array element removal with shifting, and inheritance hierarchy design

Object-oriented programming

(a)

2 marks

Polymorphism is the ability of different classes in an inheritance hierarchy to respond to the same method signature in their own specialized way (via method overriding in subclasses, or method overloading with different parameter lists).

Mentor tip

Mention either subclass method overriding (same signature, specialized behaviour) or method overloading (same name, different parameters).

(b)

6 marks

Locate the index of `Monster M` in `monsters`, shift all subsequent elements one position to the left to close the gap, set the vacated last slot `monsters[monsterCount - 1]` to `null`, and decrement `monsterCount`:
```java public void removeMonster(Monster M) { for (int i = 0; i < monsterCount; i++) { if (monsters[i] == M) { for (int j = i; j < monsterCount - 1; j++) { monsters[j] = monsters[j + 1]; } monsters[monsterCount - 1] = null; monsterCount--; break; } } } ```

Mentor tip

Remember all four steps: find index `i`, shift elements from `i` up to `monsterCount - 2` left by one (`monsters[j] = monsters[j+1]`), clear `monsters[monsterCount - 1] = null`, and decrement `monsterCount`.

(c)(i)

2 marks

Inheritance is the mechanism whereby a child subclass derives (extends) the non-private attributes and methods of a parent superclass, allowing shared state and behaviour to be defined once in the parent.

Mentor tip

State that a subclass derives/extends public and protected attributes and methods from a parent superclass.

(c)(ii)

5 marks

1. Superclass implementation of `ability1`: Because every monster shares the universal `'Heal'` ability, `ability1(Monster defender)` is fully implemented once in the base `Monster` class and inherited by all subclasses without duplication.
2. Type-level overriding of `ability2`: Each monster-type subclass (such as `WaterMonster` and `FireMonster`) extends `Monster` and overrides `ability2(Monster defender)` to implement type-specific behaviour (e.g., `'Put out fire'` for all water monsters).
3. Card-level overriding of `ability3`: Each individual monster card class (such as `Squishy`) extends its respective type class (`WaterMonster`), inheriting both `ability1` and `ability2` automatically while overriding `ability3(Monster defender)` to define its unique ability (`'Clean'`).

Mentor tip

Map each ability to its exact level in the 3-tier hierarchy: `ability1` implemented in `Monster`, `ability2` overridden in `WaterMonster`/`FireMonster`, and `ability3` overridden in `Squishy`.

Question 17

10 marks

Library collections, 2D array indexing for battle effectiveness, and best-attacker selection

Object-oriented programming

(a)

2 marks

1. Pre-tested reliability: Standard library collections are thoroughly debugged and optimized, reducing the risk of pointer or bounds errors.
2. Faster development time: Programmers save time by calling built-in methods (like dynamic resizing and removal) instead of coding and maintaining custom data structures from scratch.

Mentor tip

Cite development speed/convenience and pre-tested reliability.

(b)

2 marks

Since `monsterType` constants are `1` (Ice), `2` (Water), and `3` (Fire), subtract `1` from both `attacker.monsterType` (row index) and `defender.monsterType` (column index) to match 0-based array indexing:
```java return effectiveness[attacker.monsterType - 1][defender.monsterType - 1]; ```

Mentor tip

Don't forget to subtract 1 from both `monsterType` values because Java arrays are 0-indexed (`0, 1, 2`) while the type constants are `1, 2, 3`.

(c)

6 marks

Initialize trackers for the best monster (`chosen = monsters[0]`), highest effectiveness (`mostEffective = -9`, below the minimum possible `-8`), and highest health (`healthiest = 0`), then iterate through the active `monsterCount` elements in `monsters`:
```java public Monster recommendAttackingMonster(Monster defender) { Monster chosen = monsters[0]; int mostEffective = -9; int healthiest = 0; for (int i = 0; i < monsterCount; i++) { int eff = Monster.getAttackingWeight(monsters[i], defender); if (eff > mostEffective || (eff == mostEffective && monsters[i].health > healthiest)) { chosen = monsters[i]; mostEffective = eff; healthiest = monsters[i].health; } } return chosen; } ```

Mentor tip

Use `<` / `>` strictly (not `>=`) when comparing ties so that if two monsters have identical effectiveness and health, the first one encountered is retained.

Question 18

10 marks

Linked list vs stack for battle turn logs and implementing rewindTurns

Object-oriented programming

(a)

4 marks

1. Order and speed of access (2 marks): A stack operates in Last-In-First-Out (LIFO) order, granting immediate $O (1)$ access to the most recent turns at the top, whereas a singly linked list starting at the head (`01`) must be traversed sequentially from the beginning to reach recent turns at the tail.
2. Destructive vs non-destructive reading (2 marks): Reading previous turns from a stack requires popping (removing) elements—which would then have to be pushed back to leave `gameLog` unchanged—whereas a linked list can be traversed via node pointers without removing or modifying any nodes.

Mentor tip

Contrast LIFO top-access vs head-to-tail traversal (2 marks) and destructive `pop()` vs non-destructive pointer traversal (2 marks).

(b)

6 marks

1. Count total turns: Traverse the `gameLog` linked list from the head to count the total number of recorded `Turn` nodes (`size`), since a singly linked list cannot be accessed directly from the rear.
2. Validate request: Check if `numberOfTurns > size`; if so, output an error message indicating that the requested number of turns exceeds the battle history.
3. Locate starting turn: Calculate the starting node index (`size - numberOfTurns`) and traverse from the head to reach the earliest turn that needs to be undone.
4. Restore health selectively: Continue traversing through the remaining `numberOfTurns` nodes to the end of `gameLog`. At each node, check if the defending monster (`turn.to`) equals `rewinder`; if so, add `turn.damage` back to `rewinder.health` (ignoring turns where `rewinder` was the attacker `turn.from`).

Mentor tip

Since the prompt asks to *explain* how to implement `rewindTurns`, clearly state both each algorithmic step and *why* it is needed (e.g., counting size first because linked lists are front-accessed, and checking `to == rewinder` so only sustained damage is healed).

Our analysis

Where the marks are

Each part sorted by the topic it mostly tests. Databases carries the most marks on this paper.

Databases65 marks · 25%
Q1, Q2, Q3, Q4
Modelling and simulation65 marks · 25%
Q5, Q6, Q7, Q8
Web science65 marks · 25%
Q9, Q10, Q11, Q12, Q13
Object-oriented programming65 marks · 25%
Q14, Q15, Q16, Q17, Q18

Mentor insights

Traps to watch for

01
Databases: Emergency Management Information Systems (EMIS), spatial database · Q4
Mention both geographic coordinates (latitude/longitude) and geometric/topological representation of locations.
02
Modelling and simulation: Genetic algorithms for keyboard layout optimization and natural l · Q8
Mention increased chromosome diversity and its effect on either solution quality or computational time per generation.
03
Web science: Friends for Dogs webpage components, HTTP characteristics, PHP se · Q9
Name any two visible or structural webpage elements such as forms/buttons, images, headers, or scripts.
04
Object-oriented programming: Monster attack logic, class vs instance distinction, UML relation · Q15
Store the result of `rollDice()` in a local `damage` variable so you can both subtract it from `defender.health` and display it in the output message.

Solutions and tips are GeorgeTheTutor's own; the questions are not reproduced here. Not affiliated with or endorsed by the International Baccalaureate Organization.

Want someone to go through your paper with you?

Our maths mentors took these exams recently. One-to-one, they'll find where your marks are going and fix it before May.

Find your maths mentor More past papers

Worked solutions

IB Computer Science HL · Paper 2

Computer Science, Higher Level

Session: Nov 2023
Marks: 260
Duration: 1 hour 20 min
Calculator: Not allowed
Questions: 18

Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9 Q10 Q11 Q12 Q13 Q14 Q15 Q16 Q17 Q18

Keep your copy of the paper open alongside. We don't reproduce the questions: each part below is labelled to match the Nov 2023 Global paper, which your school or the IB can provide.

Practice mode

Hide our working until you've had a go.

Option A: Databases · Option

Questions 1–4 · 65 marks

Question 1

15 marks

Restaurant ordering database structure, date formats, composite keys, and SQL queries

Databases

(a)

4 marks

1. Data integrity and normalization: A relational database enforces structural constraints and referential integrity across related tables, ensuring that customer and meal records remain consistent and free of redundant anomalies.
2. Concurrent multi-user access: A database management system supports record locking and transaction management so multiple restaurant staff members can enter orders and update stock simultaneously without corrupting shared files (unlike a flat spreadsheet loaded entirely into memory).

Mentor tip

Provide two distinct reasons, each paired with a clear expansion on how the database improves operations compared to a spreadsheet.

(b)

2 marks

Storing dates in the ISO 8601 format (`yyyy/mm/dd`) ensures that chronological sorting matches alphabetical/lexicographical sorting (ordering by year first, then month, then day).
It also facilitates seamless data integration and compatibility across international database systems without ambiguity between day-first and month-first conventions.

Mentor tip

Highlight how year-month-day order allows standard string or numeric sorting to naturally place dates in true chronological sequence.

(c)

3 marks

Combining `OrderID` and `ItemID` forms a natural composite key because each distinct meal item appears at most once per customer order (with `Quantity` tracking multiple portions).
Using existing foreign keys eliminates the need to generate and store an arbitrary surrogate `OrderlineID` column, saving disk space and avoiding a meaningless attribute that would never be searched directly.

Mentor tip

Explain both why the pair is unique per order line and how omitting a separate surrogate key saves storage.

(d)

2 marks

When joining `ORDERLINE` and `ITEM` in a query, referencing `UnitPrice` without a table prefix creates an ambiguous column name conflict.
The query parser will fail to execute and return a syntax/ambiguity error unless the attribute is explicitly qualified using dot notation (such as `ORDERLINE.UnitPrice` or `ITEM.UnitPrice`).

Mentor tip

Mention column name ambiguity during multi-table queries and the requirement to qualify field names with their table names.

(e)

4 marks

1. Specify the required output columns: `CUSTOMER.FirstName`, `CUSTOMER.Surname`, and `ORDER.OrderDate`.
2. Identify the source tables: `CUSTOMER` and `ORDER`.
3. Join the tables on their common key: `CUSTOMER.CustomerID = ORDER.CustomerID`.
4. Filter the records using a `WHERE` clause matching both name fields:
`SELECT CUSTOMER.FirstName, CUSTOMER.Surname, ORDER.OrderDate FROM CUSTOMER INNER JOIN ORDER ON CUSTOMER.CustomerID = ORDER.CustomerID WHERE CUSTOMER.FirstName = 'Akemi' AND CUSTOMER.Surname = 'Watene';`

Mentor tip

You can express the query steps either in SQL syntax or clear structured English, ensuring the join condition and both filter criteria are explicit.

Question 2

15 marks

Derived attributes, database transactions, validation vs verification, concurrency, and recovery

Databases

(a)

3 marks

`LinePrice` is a derived (calculated) attribute that can always be computed on the fly as `Quantity * UnitPrice`.
Storing a calculated value violates Third Normal Form (3NF) by introducing transitive dependency, wastes storage space, and risks update anomalies if `Quantity` or `UnitPrice` is modified without recalculating `LinePrice`.

LinePrice = Quantity \times UnitPrice

Mentor tip

Connect derived fields to 3NF compliance, storage efficiency, and prevention of calculation discrepancies.

(b)

1 mark

Mentor tip

Emphasize that a transaction is an all-or-nothing logical unit of work.

(c)(i)

1 mark

Apply a range check (for example, ensuring `hours_worked` is between `0` and `16` hours per shift) or a numeric type check.

Mentor tip

Validation is an automated rule check performed by software on input (range check, type check, presence check).

(c)(ii)

2 marks

Mentor tip

Do not confuse validation with verification: verification confirms that data matches the original source (e.g., double entry or visual comparison).

(d)

4 marks

The DBMS applies concurrency control such as record-level locking (or two-phase locking) when the shift manager initiates an edit on the employee's shift row.
While the write lock is active, the employee's session is either blocked from reading/confirming the uncommitted record (preventing a dirty read) or shown a notification that the schedule is currently being updated.
Alternatively, Multi-Version Concurrency Control (MVCC) can serve the employee a consistent read-only snapshot of the prior committed hours while the manager's transaction is in progress, only allowing confirmation once the write lock is released.

Mentor tip

Explain how record locking or MVCC isolates the manager's in-progress edit from the employee's read/confirmation request.

(e)

4 marks

Instant recovery (recovery in place / shadow paging with virtual snapshots): The system maintains periodic consistent snapshots of the database state on a backup virtual machine or mirror.
If a crash occurs, user workloads and new write operations are immediately redirected to the clean snapshot VM so operations continue seamlessly.
Meanwhile, the background recovery protocol inspects interrupted transactions on the primary instance, rolling back incomplete changes and syncing updated blocks before switching traffic back.

Mentor tip

Describe either snapshot-based instant recovery (failover to a backup image while restoring in the background) or checkpoint/rollback crash recovery.

Question 3

15 marks

European pet passport ERD, secondary keys, 3NF normalization, and data protection

Databases

(a)

2 marks

1. `OWNER` to `PET` is a one-to-many ( $1 : M$ ) relationship (one pet owner can own multiple pets).
2. `PET` to `PASSPORT` is a one-to-one ( $1 : 1$ ) relationship (each pet holds exactly one individual passport).
Diagram notation: `OWNER (1) ----< (M) PET (1) ---- (1) PASSPORT`

Mentor tip

Clearly label both cardinalities: 1-to-many between Owner and Pet, and 1-to-1 between Pet and Passport.

(b)

1 mark

A secondary key is a non-primary attribute (which may or may not be unique) that is indexed to speed up searching, sorting, and querying records in a table.

Mentor tip

State that a secondary key is an indexed field used for faster lookup/retrieval rather than primary entity identification.

(c)

8 marks

Decompose the unnormalized passport and vaccination data into four additional 3NF relations alongside `VET`: `OWNER`, `PET`, `PASSPORT`, `VACCINE`, and `PETVACCINATION` (or combine 1:1 `PET`/`PASSPORT` with separate `OWNER` or `VACCINE` tables):
1. `PET(TransponderCode, PetName, Species, Breed, Sex, DOB)` — Primary Key: `TransponderCode`.
2. `PASSPORT(PassportID, TransponderCode*, VetID*, DateOfIssue)` (or using `TransponderCode` as PK) — Foreign Keys: `TransponderCode` referencing `PET`, `VetID` referencing `VET`.
3. `VACCINE(VaccineID, VaccineName, Manufacturer, BatchNumber)` — Primary Key: `VaccineID` (separating vaccine batch details from individual pet injections).
4. `PETVACCINATION(VaccinationID, TransponderCode*, VaccineID*, VetID*, VaccineDate, ValidFrom, ValidUntil)` — Primary Key: `VaccinationID` (or composite key `TransponderCode, VaccineID, VaccineDate`), with Foreign Keys linking the pet, the vaccine batch, and the administering vet.

Mentor tip

Ensure vaccine product/manufacturer details are separated into their own `VACCINE` table and linked via a junction/event table `PETVACCINATION` so repeating vaccination stickers are in 3NF.

(d)

4 marks

1. Purpose limitation and consent: Personal contact details of the pet owner must only be collected and processed for legitimate pet passport and veterinary traceability purposes, never shared with third parties without consent.
2. Cross-border transfer safeguards: Because the passport is used across European Union countries, owner data must only be transferred to jurisdictions that enforce equivalent data protection standards (such as GDPR).
3. Accuracy and right of access: Records must be kept accurate and up to date, and pet owners have the legal right to inspect and correct their stored personal data.
4. Storage limitation and security: Data must be protected against unauthorized access and not retained longer than necessary.

Mentor tip

Reference core data protection principles (purpose limitation, consent, accuracy, retention limits, and secure cross-border sharing).

Question 4

20 marks

Emergency Management Information Systems (EMIS), spatial databases, and spatial data mining

Databases

(a)(i)

2 marks

Spatial data represents the geometric location, shape, and topological properties of physical features on Earth.
It is stored using geographic coordinate systems such as latitude and longitude pairs.

Mentor tip

Mention both geographic coordinates (latitude/longitude) and geometric/topological representation of locations.

(a)(ii)

2 marks

1. Points (e.g., individual emergency call coordinates).
2. Lines / Polylines (e.g., road networks or highway segments) or Polygons (e.g., city districts or patrol zones).

Mentor tip

Any two standard spatial primitives—Points, Lines, or Polygons—earn full credit.

(b)

1 mark

Mentor tip

State a standard relational cardinality such as one-to-one or one-to-many.

(c)

6 marks

1. Density clustering / Heat map generation (3 marks): Spatial clustering groups nearby incident coordinates and renders a colour-coded density gradient (e.g., red for high concentration, yellow for low) using classification methods like quantiles or natural breaks. This highlights high-risk crime hotspots more clearly than overlapping dots, guiding patrol allocation.
2. Attribute or temporal filtering / Association analysis (3 marks): Filtering incidents by crime category (e.g., burglary vs assault) or time window (e.g., night shifts or annual public events) isolates specific patterns and removes confounding noise like general population density, helping dispatchers pre-position emergency vehicles.

Mentor tip

Explain two distinct spatial analysis techniques (3 marks each), detailing both how the map view changes and how emergency planners use that insight.

(d)

5 marks

1. Risk of manipulation and public deception: Viewers often perceive maps as objective truth without realizing that altering colour scales, dot sizes, or classification thresholds can exaggerate or downplay crime severity to serve a political agenda.
2. Beneficial persuasion and clarity: Visual tools can legitimately simplify complex multidimensional datasets for decision-makers, and highlighting all incidents (including minor ones) may justify increased funding for public safety and emergency response.
3. Balanced conclusion: Using visual analysis tools is ethical provided maps include transparent legends, objective metadata, and clear disclosure of filtering criteria so audiences can critically evaluate competing perspectives.

Mentor tip

Include negative ethical risks (bias/deception), positive uses (funding/clarity), and a reasoned conclusion emphasizing transparency in legends and metadata.

(e)

4 marks

Deviation detection compares incoming spatial and temporal sensor/call data against historical baselines to identify statistically significant anomalies outside normal patterns.
For example, an unusual spike in emergency calls or temperature/smoke readings clustered in a specific zone during dry seasons flags an emerging fire hazard.
By applying a statistical confidence threshold (such as 95%) to filter out random background noise, the system alerts authorities to deploy preventative fire units early.

Mentor tip

Connect historical baseline comparison, spatial/temporal clustering of outliers, and noise filtering via confidence thresholds.

Option B: Modelling and simulation · Option

Questions 5–8 · 65 marks

Question 5

16 marks

Cattery5S optimal stopping problem, permutation tables, and spreadsheet modelling

Modelling and simulation

(a)

1 mark

`0.5` (or `50%` / `1/2`).

P (best) = \frac{1}{2} = 0.5

Mentor tip

With two cats and no prior information, either choice has an equal 50% chance of being the higher-rated cat.

(b)

1 mark

Complete rows 4, 5, and 6 with the remaining three permutations of `{1, 2, 3}`:
Row 4: `2 | 3 | 1`
Row 5: `3 | 1 | 2`
Row 6: `3 | 2 | 1`

Mentor tip

Ensure all 6 permutations of (1, 2, 3) are present without duplicating any row.

(c)

2 marks

Use an `IF` function comparing the second cat in column B (`B1`) against the first cat in column A (`A_1`, written in spreadsheet syntax without underscore):
`=IF(B1 > A_1, B1, C1)` (where `A_1` denotes cell column A, row 1)
If the second cat rating exceeds the first cat rating, return `B1`; otherwise, return the third cat `C1`.

Mentor tip

Check that the condition compares `B1` to the first cat in column A row 1, returning `B1` if true and `C1` if false.

(d)(i)

2 marks

Use `COUNTIF` with an absolute row reference for the outcome range `D $1 : D$ 6` and a criteria value (such as `1`, `2`, `3` or a cell reference `E1` containing the target rating):
`=COUNTIF( $D$ 1: $D$ 6, E1)` (or `=COUNTIF(D $1 : D$ 6, 3)` for rating 3).

Mentor tip

Include dollar signs (` $D$ 1: $D$ 6`) to lock the range when copying the counting formula across the three cat ratings.

(d)(ii)

3 marks

Evaluate the chosen cat in column D across all 6 permutations:
• Row 1 (`1, 2, 3`): $2 > 1 ⟹$ chooses `2`.
• Row 2 (`1, 3, 2`): $3 > 1 ⟹$ chooses `3`.
• Row 3 (`2, 1, 3`): $1 < 2 ⟹$ moves to C, chooses `3`.
• Row 4 (`2, 3, 1`): $3 > 2 ⟹$ chooses `3`.
• Row 5 (`3, 1, 2`): $1 < 3 ⟹$ moves to C, chooses `2`.
• Row 6 (`3, 2, 1`): $2 < 3 ⟹$ moves to C, chooses `1`.
Frequencies out of 6: Rating 1 is chosen 1 time ( $1/6 \approx 16.7%$ ); Rating 2 is chosen 2 times ( $2/6 \approx 33.3%$ ); Rating 3 (best cat) is chosen 3 times ( $3/6 = 50%$ ).

P (1) = \frac{1}{6} \approx 16.7%, P (2) = \frac{2}{6} \approx 33.3%, P (3) = \frac{3}{6} = 50%

Mentor tip

State the probability or percentage for each of the three ratings (16.7%, 33.3%, and 50%).

(e)

1 mark

Nest a second `IF` statement to check the third cat (`C1`) against the first cat in column A row 1 (`A_1`) if the second cat (`B1`) is lower, defaulting to the fourth cat (`D1`):
`=IF(B1 > A_1, B1, IF(C1 > A_1, C1, D1))` (where `A_1` denotes cell column A, row 1)

Mentor tip

Both comparisons (`B1` and `C1`) benchmark against the first cat in column A, row 1.

(f)

6 marks

1. Advantages (up to 3 marks): A spreadsheet automates repetitive conditional evaluations without manual arithmetic errors, visually displays the exact chosen cat for every permutation row, supports quick charting of probability distributions, and allows easy tweaking of decision formulas.
2. Disadvantages (up to 3 marks): Spreadsheets do not natively generate ordered permutations, requiring manual entry of rows ( $N!$ grows factorially from 6 to 24 to 40,320). Deeply nested `IF` formulas become unwieldy and error-prone as the number of cats increases.
3. Conclusion: A spreadsheet model is effective and transparent for small problem sizes (3 or 4 cats), but becomes impractical for larger numbers where a programmatic algorithm is required.

Mentor tip

Balance advantages (visual clarity, automatic recalculation) with disadvantages (manual permutation entry, nested IF complexity as N! grows) and finish with a clear conclusion.

Question 6

16 marks

Algorithmic simulation of 8-cat and 20-cat selection strategies and hardware scaling

Modelling and simulation

(a)

2 marks

Initialize an accumulator `TOTAL = 1` and multiply it by each integer from `1` to `N` in a loop (or use recursion with base case `N <= 1` returning `1`):
```text function FACTORIAL(N) TOTAL = 1 loop X from 1 to N TOTAL = TOTAL * X end loop return TOTAL end function ```

N! = x = 1 \prod N x

Mentor tip

Either an iterative loop multiplying from 1 to N or a recursive function with a base case of N = 1 is valid.

(b)

2 marks

Initialize `HIGHEST` to `ARR[0]` (or `-1`) and compare across the first three elements (indices `0` to `2`):
```text function FINDHIGH(ARR) HIGHEST = ARR[0] loop I from 1 to 2 if ARR[I] > HIGHEST then HIGHEST = ARR[I] end if end loop return HIGHEST end function ```

Mentor tip

Remember that 0-indexed arrays store the first three cats at indices 0, 1, and 2.

(c)

6 marks

Iterate through all `FACTORIAL(8)` permutations, find the max of the first 3 cats using `FINDHIGH(ARR)`, scan from index `3` to `7` for the first cat exceeding `HIGHEST` (or default to the last cat `ARR[7]`), count how often rating `8` (the best cat) is selected, and advance `ARR` via `NEXTPERMUTATION(ARR)`:
```text TOTAL_BEST = 0 NUM_PERMS = FACTORIAL(8) loop ROW from 0 to NUM_PERMS - 1 HIGHEST = FINDHIGH(ARR) FOUND = false POS = 3 loop while FOUND = false AND POS < 8 if ARR[POS] > HIGHEST then CHOICE = ARR[POS] FOUND = true end if POS = POS + 1 end loop if FOUND = false then CHOICE = ARR[7] end if if CHOICE = 8 then TOTAL_BEST = TOTAL_BEST + 1 end if ARR = NEXTPERMUTATION(ARR) end loop PROBABILITY = (TOTAL_BEST / NUM_PERMS) * 100 output PROBABILITY ```

Mentor tip

Ensure you handle the fallback case (`if FOUND = false then CHOICE = ARR[7]`) and check `if CHOICE = 8` before advancing to `NEXTPERMUTATION(ARR)`.

(d)

2 marks

Modify the threshold number of initial cats rejected (for example, test rejecting the first 2 cats or the first 4 cats instead of 3).
Compare the resulting probabilities of selecting the best cat across each threshold to verify which rejection count maximizes success.

Mentor tip

State both what parameter to vary (the number of initially rejected cats) and how to evaluate the outcome (comparing the resulting success probabilities).

(e)

4 marks

Evaluating $20! \approx 2.4 \times 1 0^{18}$ permutations is computationally intractable on a standard single-CPU laptop due to combinatorial explosion.
It requires high-performance parallel processing architecture—such as multi-core supercomputing clusters, GPUs adapted for massive parallel thread execution, or dedicated ASICs/math co-processors—alongside substantial high-speed RAM to manage state across parallel workers.

20! \approx 2.43 \times 1 0^{18}

Mentor tip

Link the $2.4 \times 1 0^{18}$ scale to parallel processing architectures (multi-CPU clusters, adapted GPU compute, high clock speed, and large RAM).

Question 7

13 marks

2D map visualizations vs 3D building walkthroughs for emergency response

Modelling and simulation

(a)

1 mark

A 2D visualization is a flat, two-dimensional graphical representation (using horizontal and vertical axes) of data, objects, or processes.

Mentor tip

Define it clearly as a flat (two-axis / non-3D) visual representation of data or objects.

(b)(i)

4 marks

1. Alignment with geographic coordinates: City maps naturally use 2D latitude and longitude coordinates, allowing GPS locations of emergency calls to be plotted accurately without needing altitude.
2. Visual clarity and rapid interpretation: A 2D map avoids 3D perspective distortion and occlusion, enabling dispatchers to read street layouts and incident densities at a glance.

Mentor tip

Give two distinct reasons with expansions: 2D GPS coordinate mapping and visual simplicity/readability.

(b)(ii)

2 marks

Because a 2D map only stores flat coordinate vectors/pixels without complex 3D polygon meshes, z-buffering, or volumetric lighting textures, it requires minimal RAM and video memory.

Mentor tip

Contrast the low memory footprint of flat 2D graphics with the heavy memory demands of 3D rendering.

(c)

6 marks

1. Rendering speed vs urgency (2 marks): Because people are trapped in an active fire, there is no time for slow photorealistic ray tracing; fast wireframe or scanline rendering must be chosen to generate the model immediately.
2. File size and wireless transmission (2 marks): High-detail 3D animations produce huge files that may buffer or fail over mobile networks; a lightweight low-polygon model ensures instant transmission to handheld tablets.
3. Functional detail vs visual realism (2 marks): Rescue personnel only need structural navigation cues (corridors, stairwells, exits, and hazards) rather than realistic surface textures or shadows.

Mentor tip

Outline three distinct considerations (time/rendering method, file size/bandwidth, and required level of functional detail) for 2 marks each.

Question 8

20 marks

Genetic algorithms for keyboard layout optimization and natural language processing (NLP)

Modelling and simulation

(a)(i)

2 marks

Mentor tip

Mention increased chromosome diversity and its effect on either solution quality or computational time per generation.

(a)(ii)

3 marks

The search space of all possible keyboard permutations ( $26!$ ) is astronomically large, so purely random initial layouts will mostly have terrible typing efficiency.
Seeding the initial population with known ergonomic layouts (such as Dvorak or Colemak) starts the search much closer to an optimal region, significantly reducing the number of generations needed to converge.

Mentor tip

Connect the vast size of the permutation space to the benefit of seeding with existing high-performing layouts (like Dvorak) to speed up convergence.

(b)

3 marks

The fitness function evaluates each individual keyboard layout (for example, by calculating total finger travel distance or typing time across a corpus of common English text) and assigns a numerical score.
Layouts with higher fitness scores have a higher probability of being selected as parents to pass their key arrangements to the next generation.

Mentor tip

Explain both what the fitness score measures (closeness to the optimal typing target) and how it drives parent selection for reproduction.

(c)

2 marks

Mentor tip

Point out that standard single-point crossover violates the permutation constraint by duplicating some letters and omitting others.

(d)

4 marks

1. Massive increases in processing power and GPU hardware acceleration.
2. Availability of Big Data (vast corpora of recorded human speech and linguistic datasets for training and testing).
3. Advancements in deep learning and neural network architectures.
4. Deeper computational modelling of human phonology, syntax, and semantics.

Mentor tip

List four concise factors: computing power, large speech datasets (Big Data), advanced ML/AI algorithms, and linguistic modeling.

(e)

4 marks

Human language is inherently ambiguous, relying on shared real-world context, tone, emotion, or sarcasm to convey meaning beyond literal word definitions.
Without explicit contextual cues across phonological, semantic, and pragmatic levels, a machine-learning model may misinterpret homonyms or figurative expressions.
Training supervised models to handle context requires massive, richly annotated datasets covering diverse conversational scenarios, and unsupervised models often struggle to infer pragmatic intent.

Mentor tip

Discuss ambiguity, pragmatic factors like sarcasm/emotion, and the difficulty of capturing real-world context in training datasets.

(f)

2 marks

Machine translation relies on large parallel corpora of translated texts; far more bilingual training data exists between English and French than directly between Bengali and Polish (often forcing translation through English as an intermediate pivot language, compounding errors).
Additionally, English and French share the Latin alphabet and closer grammatical roots, whereas Bengali and Polish use completely different scripts and grammatical structures.

Mentor tip

Mention either the scarcity of parallel training datasets (requiring an English intermediate step) or the difference in alphabets and linguistic structures.

Option C: Web science · Option

Questions 9–13 · 65 marks

Question 9

18 marks

Friends for Dogs webpage components, HTTP characteristics, PHP server-side processing, and templates

Web science

(a)

2 marks

1. Text input box / HTML form (`<input type="text">`) and submit button.
2. Images of dogs, hyperlinks, page header/banner, or body text paragraphs.

Mentor tip

Name any two visible or structural webpage elements such as forms/buttons, images, headers, or scripts.

(b)

2 marks

1. Stateless protocol: Each HTTP request is processed independently without retaining session memory of previous requests.
2. Request-response client-server model (using methods like `GET` and `POST`, status codes, and plain-text transmission over TCP port 80 by default).

Mentor tip

State any two core HTTP features: statelessness, client-server request/response methods (GET/POST), unencrypted text messaging, or status codes.

(c)

4 marks

1. The script retrieves the submitted `dogname` value from the HTTP `POST` request array (`$_POST["dogname"]`).
2. It opens a connection to the `dogDB` MySQL database on `localhost` using the server credentials.
3. It constructs and executes a `SELECT` query on the `Dogs` table to fetch `dname`, `image`, and `price` matching `$dogname`.
4. If one or more matching rows exist, it loops through the result set and outputs each dog's image, name, and price; otherwise it outputs `"not found"`, and finally closes the database connection.

Mentor tip

Trace the PHP script chronologically: read POST input -> connect to DB -> run SELECT query -> loop/echo results or 'not found' -> close connection.

(d)

2 marks

Server-side execution protects sensitive database credentials (` $u ser nam e ‘ an d ‘$ pass`) by keeping them hidden on the server rather than exposing them in client-viewable source code.
It also allows direct querying of the centralized `dogDB` database residing on the server without transferring the entire database to the client browser.

Mentor tip

Emphasize hiding database login credentials from the user or interacting directly with the server-hosted database.

(e)

3 marks

Lossy compression significantly reduces image file sizes compared to lossless formats.
Smaller file sizes reduce bandwidth consumption and server storage costs, allowing dog photos to download and render much faster for users browsing or sharing them.
Although minor pixel detail is permanently discarded, the visual degradation is imperceptible for casual web viewing.

Mentor tip

Connect smaller file size to faster download speeds/lower bandwidth and tolerable visual quality for web photos.

(f)

5 marks

1. Advantages (up to 3 marks): Templates provide a consistent, familiar navigation layout across all pages, speed up development and lower costs by reusing pre-built structural modules (headers, footers), and often feature pre-tested responsive code.
2. Disadvantages (up to 3 marks): Templates restrict creative flexibility and customization, may include unnecessary code bloat from unused features, and can make the Friends for Dogs website look generic and indistinguishable from other template-based sites.

Mentor tip

Weigh both advantages (consistency, rapid/cheaper development, modular reuse) and disadvantages (limited customization, generic appearance, code bloat).

Question 10

14 marks

Website interactivity, copyright compliance, search ranking metrics, and white hat SEO

Web science

(a)

2 marks

1. Add online booking forms or live chat support so visitors can reserve tours and ask questions in real time.
2. Incorporate interactive maps, customer rating/review sections, or virtual 360-degree tour previews.

Mentor tip

Identify any two interactive features relevant to a travel website (booking forms, live chat, customer reviews, interactive maps).

(b)

2 marks

Mentor tip

Mention either creating original media or obtaining proper licenses/permissions (including Creative Commons compliance) for third-party assets.

(c)

1 mark

The number and authority of incoming backlinks (as used in PageRank/HITS), page load speed, user dwell time, or click-through rate.

Mentor tip

State one specific search ranking metric such as backlink quality/quantity, page load time, or visitor dwell time.

(d)

2 marks

1. Including accurate, relevant keywords in `<meta>` tags, headings, and high-quality body content.
2. Earning natural backlinks from reputable travel directories and optimizing site structure for mobile responsiveness.

Mentor tip

List two ethical (white hat) SEO methods: relevant keywords/meta tags, quality original content, reputable backlinks, or mobile optimization.

(e)

4 marks

1. Long-term sustainability and penalty avoidance: White hat SEO complies with search engine guidelines, preventing the travel site from being down-ranked or banned (delisted) during algorithm updates.
2. Enhanced user experience and credibility: Focusing on clear navigation, fast loading, and genuine travel content builds customer trust, increases visitor retention, and drives organic bookings.

Mentor tip

Explain two reasons (2 marks each): avoiding search engine penalties/de-indexing and building sustainable user trust/engagement.

(f)

3 marks

If search engine developers assume that webpage importance is strictly proportional to the number of incoming backlinks, the algorithm will prioritize established, heavily linked domains.
Consequently, newly published or niche websites containing highly accurate, relevant information may be buried in search results because they have not yet accumulated backlinks, reducing the diversity and freshness of surfaced content.

Mentor tip

Link a specific design assumption (e.g., equating backlinks with quality, or static indexing intervals) to its effect on search result accuracy and diversity.

Question 11

13 marks

Client-server architecture, identification vs authentication, cloud storage, and web decentralization

Web science

(a)

2 marks

A server is a centralized host that manages, stores, and provides shared resources or services across the network.
A client is an end-user workstation or device that initiates requests to the server to access those shared services or files.

Mentor tip

Clearly contrast the server's role (providing/hosting resources) with the client's role (requesting services).

(b)

2 marks

Identification is the act of claiming an identity (for example, entering a username or employee ID).
Authentication is the process of verifying that claim using secret credentials or evidence (for example, checking a password, biometric scan, or security token).

Mentor tip

Remember: Identification states *who you claim to be* (username); authentication *proves it* (password/biometrics).

(c)

3 marks

Scalability and cost efficiency: Moving to cloud storage eliminates the capital expense of purchasing, powering, and maintaining on-premises file servers.
The training company can dynamically scale storage capacity up or down on a pay-as-you-go basis while enabling employees to access and share training materials remotely from any internet-connected device.

Mentor tip

Develop one core reason thoroughly—such as hardware/maintenance cost reduction, elastic scalability, ubiquitous remote access, or automated disaster recovery.

(d)

6 marks

1. Arguments that cloud storage supports decentralization: Cloud infrastructure distributes data across geographically dispersed data centres globally and allows small organizations and individuals to publish services and share files from anywhere without owning local server hardware.
2. Arguments that cloud storage increases centralization (ownership concentration): Despite geographic distribution, cloud infrastructure is dominated by a small oligopoly of mega-providers (such as AWS, Google Cloud, Azure). Businesses migrating away from independent local servers to these few platforms consolidate control, create single points of failure/cyber-attack targets, and lock users into proprietary ecosystems with high network effects.
3. Conclusion: While cloud storage decentralizes physical access and workplace collaboration, it has largely acted as a catalyst for *re-centralizing* web infrastructure ownership until peer-to-peer or blockchain storage alternatives achieve mainstream adoption.

Mentor tip

Distinguish between *geographic/access distribution* (access from anywhere) and *ownership concentration* (control by a few tech giants) to build a strong 6-mark evaluation.

Question 12

10 marks

Bow-tie web graph structure, strongly connected core (SCC), tubes vs tendrils, and power laws

Web science

(a)

2 marks

A web graph is a directed graph representing the entire World Wide Web, where vertices represent webpages and directed edges represent hyperlinks between them.
A sub-graph represents a smaller, localized subset of nodes and links extracted from the overall web graph (often focused on a specific domain or topic).

Mentor tip

Contrast the full World Wide Web directed graph with a smaller subset of pages/links.

(b)

2 marks

1. Mutual reachability: Every webpage within the Strongly Connected Core (SCC) can be reached from every other webpage in the SCC via a directed path of hyperlinks.
2. High link density and resilience: The SCC forms the central, densely interlinked backbone of the web graph with multiple redundant pathways between nodes.

Mentor tip

Highlight that any node in the SCC can reach any other node in the SCC via directed links.

(c)

2 marks

A tube is a directed pathway of webpages that connects the `IN` component directly to the `OUT` component while bypassing the Strongly Connected Core (SCC) entirely.
A tendril is a dead-end branch that either leads out from the `IN` component without ever reaching the SCC, or leads into the `OUT` component from nowhere without being reachable from the SCC.

Mentor tip

A tube bridges IN to OUT bypassing the SCC; a tendril branches off IN (or into OUT) without connecting the two.

(d)

2 marks

Under a power-law (scale-free / rich-get-richer) distribution, a tiny fraction of dominant hub websites attract the vast majority of new incoming links and high PageRank visibility.
As user traffic and links concentrate around these mega-platforms, smaller independent sites either lose reciprocal links or become obsolete, shrinking the relative proportion of pages that remain mutually interlinked in the main SCC.

Mentor tip

Connect power-law preferential attachment ('rich get richer' via high PageRank) to link concentration on major hubs and the shrinking of the broader SCC.

(e)

2 marks

Sites like `wordpress.org`, `youtube.com`, and `wikipedia.org` are massive, authoritative content hubs or publishing platforms that provide embeddable widgets, reference articles, and template footer links.
Because millions of external blogs and websites naturally cite Wikipedia articles, embed YouTube videos, or use WordPress themes, they accumulate an enormous volume of incoming backlinks.

Mentor tip

Explain that these sites act as authoritative reference hubs or content/CMS platforms that millions of external websites embed or cite.

Question 13

10 marks

Ambient intelligence in smart homes for dementia care and online health data behaviour

Web science

(a)

4 marks

1. Sensor input: Non-intrusive IoT sensors embedded around the smart home continuously monitor the patient's vital signs (body temperature, heart rate), movement/fall detection, and room conditions (temperature, door state, stove usage).
2. Real-time processing: The ambient intelligence system compares live sensor streams against the patient's baseline behavioural profile to identify anomalies (such as wandering at night or forgetting medication).
3. Automated output and alerts: The environment automatically adjusts lighting/heating, plays audio/visual reminders for medication, and dispatches immediate alerts to caregivers or emergency services if a fall or hazardous deviation is detected.

Mentor tip

Structure your explanation using the Input (sensors) -> Processing (anomaly detection against baseline) -> Output (automated adjustments/caregiver alerts) framework.

(b)

6 marks

1. Positive behavioural shifts (empowerment and prevention): Patients actively research symptoms, treatments, and preventive lifestyle habits online, improving health literacy. Online support communities and patient forums foster collective engagement, and patients enter medical consultations better informed.
2. Negative behavioural shifts (self-diagnosis and anxiety): The overwhelming volume of unverified or commercialized medical claims on the web can trigger information overload, cyberchondria (health anxiety), and dangerous self-diagnosis or self-medication with unproven remedies.
3. Systemic impact and conclusion: While online health data has transformed patients from passive recipients into proactive participants, disparities in digital literacy and the prevalence of misinformation mean professional medical guidance remains essential.

Mentor tip

Discuss both constructive changes (health literacy, support forums, preventive care) and risks (self-diagnosis, misinformation, anxiety, doctor-patient friction).

Option D: Object-oriented programming · Option

Questions 14–18 · 65 marks

Question 14

11 marks

Monster Battle class definitions, static constants, encapsulation, and inheritance constructors

Object-oriented programming

(a)

1 mark

Mentor tip

Do not confuse `static` with `final`: `static` means class-level/shared across all instances, whereas `final` means constant.

(b)(i)

2 marks

Mentor tip

Link `private` visibility to encapsulation/data hiding and protection against accidental external modification.

(b)(ii)

2 marks

1. Accessor methods (getters).
2. Mutator methods (setters).

Mentor tip

State both accessor (getter) and mutator (setter) methods.

(c)

2 marks

The `super(...)` call invokes the constructor of the parent superclass (`Monster`), passing `name`, `health`, and `Monster.WATER_MONSTER_TYPE` to initialize the inherited attributes.

Mentor tip

State both that `super` refers to the superclass (`Monster`) and that it calls the superclass constructor to initialize inherited fields.

(d)

2 marks

1. Transparency and peer review: Making source code publicly accessible so developers worldwide can inspect, audit, and improve security and reliability.
2. Affordability, flexibility, and perpetuity: Providing free-to-use software that users can modify, localize, and maintain without vendor lock-in.

Mentor tip

Identify any two aims of open source: transparency, affordability, flexibility/customizability, interoperability, or community security auditing.

(e)

2 marks

Instantiate a `FireMonster` object (or `Monster` with `FIRE_MONSTER_TYPE`), passing `"Flamey"` and `40` to the constructor:
```java FireMonster flamey = new FireMonster("Flamey", 40); ```

Mentor tip

Ensure the variable type, `new FireMonster(...)` constructor call, and parameter order `("Flamey", 40)` match the class definition.

Question 15

19 marks

Monster attack logic, class vs instance distinction, UML relationships, and Arena array management

Object-oriented programming

(a)

4 marks

Call `rollDice()` to compute the damage, subtract it from `defender.health`, output the remaining health if `> 0`, or clamp `defender.health` to `0` and output the elimination message:
```java public void attack(Monster defender) { int damage = rollDice(); defender.health = defender.health - damage; if (defender.health > 0) { output(defender.name + " sustained " + damage + " damage and has " + defender.health + " health points."); } else { defender.health = 0; output(defender.name + " must leave the game"); } } ```

Mentor tip

Store the result of `rollDice()` in a local `damage` variable so you can both subtract it from `defender.health` and display it in the output message.

(b)

4 marks

1. Blueprint vs concrete entity: The `FireMonster` class is the static code template/blueprint defining attributes and behaviours, whereas an instance is a concrete object created at runtime with its own specific attribute values (such as `"Flamey"` with `40` health).
2. Multiplicity and memory allocation: There is only one `FireMonster` class definition in the program, whereas multiple distinct `FireMonster` instances can be allocated in heap memory simultaneously.

Mentor tip

Contrast template/blueprint vs runtime object with specific state, and single class definition vs multiple memory-allocated instances.

(c)

2 marks

1. `FireMonster` to `Monster`: Draw an inheritance arrow (closed, unfilled triangle pointing from `FireMonster` to `Monster`) representing an **is-a** relationship.
2. `Arena` to `Monster`: Draw an aggregation connector (open diamond on the `Arena` side connected to `Monster`) representing a **has-a** relationship:
`[Arena] <>---- [Monster] <|---- [FireMonster]`

Mentor tip

Use a hollow triangle pointing to `Monster` for inheritance (`FireMonster` is a `Monster`) and a diamond on `Arena` for aggregation (`Arena` has `Monster`s).

(d)

5 marks

Check if `monsterCount < monsters.length` (the arena is not full), search indices `0` to `monsterCount - 1` to verify `M` is not already present, insert `M` at `monsters[monsterCount]` and increment `monsterCount`, or output the specific reason for rejection:
```java public void addMonster(Monster M) { if (monsterCount >= monsters.length) { output("Arena is full"); } else { boolean exists = false; for (int i = 0; i < monsterCount; i++) { if (monsters[i] == M) { exists = true; } } if (!exists) { monsters[monsterCount] = M; monsterCount++; } else { output(M.name + " is already in the Arena"); } } } ```

Mentor tip

Include two distinct `output(...)` messages: one when the array is full (`monsterCount == monsters.length`) and another when `M` is already in the array.

(e)

4 marks

1. Increased complexity and code size for simple tasks: Designing class hierarchies, constructors, and encapsulation adds structural overhead that is unnecessary for small, straightforward scripts.
2. Runtime memory and execution overhead: Instantiating, referencing, and garbage-collecting numerous objects at runtime can consume more memory and execute slightly slower than procedural code.

Mentor tip

Give two distinct disadvantages (2 marks each): design/code complexity for small programs and runtime memory/speed overhead from object instantiation.

Question 16

15 marks

Polymorphism, array element removal with shifting, and inheritance hierarchy design

Object-oriented programming

(a)

2 marks

Mentor tip

Mention either subclass method overriding (same signature, specialized behaviour) or method overloading (same name, different parameters).

(b)

6 marks

Locate the index of `Monster M` in `monsters`, shift all subsequent elements one position to the left to close the gap, set the vacated last slot `monsters[monsterCount - 1]` to `null`, and decrement `monsterCount`:
```java public void removeMonster(Monster M) { for (int i = 0; i < monsterCount; i++) { if (monsters[i] == M) { for (int j = i; j < monsterCount - 1; j++) { monsters[j] = monsters[j + 1]; } monsters[monsterCount - 1] = null; monsterCount--; break; } } } ```

Mentor tip

(c)(i)

2 marks

Mentor tip

State that a subclass derives/extends public and protected attributes and methods from a parent superclass.

(c)(ii)

5 marks

1. Superclass implementation of `ability1`: Because every monster shares the universal `'Heal'` ability, `ability1(Monster defender)` is fully implemented once in the base `Monster` class and inherited by all subclasses without duplication.
2. Type-level overriding of `ability2`: Each monster-type subclass (such as `WaterMonster` and `FireMonster`) extends `Monster` and overrides `ability2(Monster defender)` to implement type-specific behaviour (e.g., `'Put out fire'` for all water monsters).
3. Card-level overriding of `ability3`: Each individual monster card class (such as `Squishy`) extends its respective type class (`WaterMonster`), inheriting both `ability1` and `ability2` automatically while overriding `ability3(Monster defender)` to define its unique ability (`'Clean'`).

Mentor tip

Map each ability to its exact level in the 3-tier hierarchy: `ability1` implemented in `Monster`, `ability2` overridden in `WaterMonster`/`FireMonster`, and `ability3` overridden in `Squishy`.

Question 17

10 marks

Library collections, 2D array indexing for battle effectiveness, and best-attacker selection

Object-oriented programming

(a)

2 marks

1. Pre-tested reliability: Standard library collections are thoroughly debugged and optimized, reducing the risk of pointer or bounds errors.
2. Faster development time: Programmers save time by calling built-in methods (like dynamic resizing and removal) instead of coding and maintaining custom data structures from scratch.

Mentor tip

Cite development speed/convenience and pre-tested reliability.

(b)

2 marks

Since `monsterType` constants are `1` (Ice), `2` (Water), and `3` (Fire), subtract `1` from both `attacker.monsterType` (row index) and `defender.monsterType` (column index) to match 0-based array indexing:
```java return effectiveness[attacker.monsterType - 1][defender.monsterType - 1]; ```

Mentor tip

Don't forget to subtract 1 from both `monsterType` values because Java arrays are 0-indexed (`0, 1, 2`) while the type constants are `1, 2, 3`.

(c)

6 marks

Initialize trackers for the best monster (`chosen = monsters[0]`), highest effectiveness (`mostEffective = -9`, below the minimum possible `-8`), and highest health (`healthiest = 0`), then iterate through the active `monsterCount` elements in `monsters`:
```java public Monster recommendAttackingMonster(Monster defender) { Monster chosen = monsters[0]; int mostEffective = -9; int healthiest = 0; for (int i = 0; i < monsterCount; i++) { int eff = Monster.getAttackingWeight(monsters[i], defender); if (eff > mostEffective || (eff == mostEffective && monsters[i].health > healthiest)) { chosen = monsters[i]; mostEffective = eff; healthiest = monsters[i].health; } } return chosen; } ```

Mentor tip

Use `<` / `>` strictly (not `>=`) when comparing ties so that if two monsters have identical effectiveness and health, the first one encountered is retained.

Question 18

10 marks

Linked list vs stack for battle turn logs and implementing rewindTurns

Object-oriented programming

(a)

4 marks

1. Order and speed of access (2 marks): A stack operates in Last-In-First-Out (LIFO) order, granting immediate $O (1)$ access to the most recent turns at the top, whereas a singly linked list starting at the head (`01`) must be traversed sequentially from the beginning to reach recent turns at the tail.
2. Destructive vs non-destructive reading (2 marks): Reading previous turns from a stack requires popping (removing) elements—which would then have to be pushed back to leave `gameLog` unchanged—whereas a linked list can be traversed via node pointers without removing or modifying any nodes.

Mentor tip

Contrast LIFO top-access vs head-to-tail traversal (2 marks) and destructive `pop()` vs non-destructive pointer traversal (2 marks).

(b)

6 marks

1. Count total turns: Traverse the `gameLog` linked list from the head to count the total number of recorded `Turn` nodes (`size`), since a singly linked list cannot be accessed directly from the rear.
2. Validate request: Check if `numberOfTurns > size`; if so, output an error message indicating that the requested number of turns exceeds the battle history.
3. Locate starting turn: Calculate the starting node index (`size - numberOfTurns`) and traverse from the head to reach the earliest turn that needs to be undone.
4. Restore health selectively: Continue traversing through the remaining `numberOfTurns` nodes to the end of `gameLog`. At each node, check if the defending monster (`turn.to`) equals `rewinder`; if so, add `turn.damage` back to `rewinder.health` (ignoring turns where `rewinder` was the attacker `turn.from`).

Mentor tip

Our analysis

Where the marks are

Each part sorted by the topic it mostly tests. Databases carries the most marks on this paper.

Databases65 marks · 25%
Q1, Q2, Q3, Q4
Modelling and simulation65 marks · 25%
Q5, Q6, Q7, Q8
Web science65 marks · 25%
Q9, Q10, Q11, Q12, Q13
Object-oriented programming65 marks · 25%
Q14, Q15, Q16, Q17, Q18

Mentor insights

Traps to watch for

01
Databases: Emergency Management Information Systems (EMIS), spatial database · Q4
Mention both geographic coordinates (latitude/longitude) and geometric/topological representation of locations.
02
Modelling and simulation: Genetic algorithms for keyboard layout optimization and natural l · Q8
Mention increased chromosome diversity and its effect on either solution quality or computational time per generation.
03
Web science: Friends for Dogs webpage components, HTTP characteristics, PHP se · Q9
Name any two visible or structural webpage elements such as forms/buttons, images, headers, or scripts.
04
Object-oriented programming: Monster attack logic, class vs instance distinction, UML relation · Q15
Store the result of `rollDice()` in a local `damage` variable so you can both subtract it from `defender.health` and display it in the output message.

Solutions and tips are GeorgeTheTutor's own; the questions are not reproduced here. Not affiliated with or endorsed by the International Baccalaureate Organization.

Want someone to go through your paper with you?

Our maths mentors took these exams recently. One-to-one, they'll find where your marks are going and fix it before May.

Find your maths mentor More past papers