Top 200 Vocabulary for JC 2 H2 Computing Syllabus | A-Level 9569 Python, Databases, Cybersecurity and AI

JC2 H2 Computing vocabulary for SEAB syllabus 9569 (2027): here are 200 carefully defined terms covering Python, algorithms, data structures, SQL, networks, privacy, machine learning and the practical examination. Find your topic in the contents, learn the distinction, then use the small worked examples below to connect language with actual code and decisions.

The revised 2027 H2 Computing syllabus has six modules: Programming Fundamentals; Data Structures and Algorithms; Data and Information; Computer Networks; Social, Ethical and Security Impact of Computing; and Emerging Technologies. Paper 1 is a written examination; Paper 2 is a lab-based examination in which candidates use Python, HTML, CSS and tools including SQLite, Flask, JupyterLab and scikit-learn.

Quick information:

200 unique, numbered glossary entries in ten topic groups, accessible from the links below.
2027 H2 Computing code 9569, with revised content first examined in 2027.
Paper 1: 3 hours, 100 marks, 60%; six to eight compulsory written questions across six modules.
Paper 2: 3 hours, 100 marks, 40%; four compulsory laptop-based questions on modules 1, 2, 3, 4 and 6.
Includes sample Python/SQL, exam distinctions, a four-week study plan, FAQs and official reference materials.

Editorial scope: an independently curated educational glossary, not a list of 200 prescribed SEAB terms or an official product of MOE and SEAB. Check your examination year because the 2027 syllabus is revised.

Find the topic you need

Top 200 JC2 H2 Computing vocabulary terms

1. Programming and Python foundations

program — Set of executable instructions that directs a computer to carry out a task; distinguish intended behaviour from actual output.
source code — Human-readable instructions written in a programming language before interpretation or translation.
algorithm — Finite, unambiguous sequence of steps for solving a well-defined problem, independent of the particular language used.
variable — Named reference to a value in a program, whose binding may change during execution.
assignment — Operation giving a variable a value, such as total = 0; it is not a mathematical assertion of equality.
data type — Category governing representation and allowed operations, such as integer, float, Boolean or string.
integer — Whole-number type used for counts and indices; unlike a floating-point approximation it has no fractional part.
floating-point number — Computer representation of approximate real values, which may introduce rounding errors into comparisons.
Boolean — Logical value True or False used in conditions and program decisions.
string — Sequence of characters representing text; a digit inside a string is not automatically a number.
type conversion — Explicit change from one data type to another, for example converting validated numeric input to an integer.
expression — Combination of operands and operators evaluated to produce a result.
operator precedence — Rules determining which operations are evaluated first, unless parentheses specify another order.
selection — Decision structure using conditions to choose between alternative execution paths.
iteration — Repeated execution of statements, such as a for loop or while loop.
condition — Logical expression evaluated to decide whether code should run or a loop should continue.
function — Named, reusable block of code that receives arguments and may return a value.
parameter — Variable declared in a function definition to receive a value when the function is called.
argument — Value supplied to a function call, corresponding to one of its parameters.
return value — Result passed back to a calling program by a function, as distinct from output merely printed on screen.

2. Decomposition, correctness and efficiency

computational thinking — Approach combining decomposition, abstraction, representation and algorithm design to solve problems.
decomposition — Breaking a large problem into smaller, more manageable tasks with clear responsibilities.
abstraction — Focusing on essential information while ignoring detail irrelevant to the current model.
pattern recognition — Identifying similarities across examples that may support a general algorithm or reusable solution.
pseudocode — Language-independent description of an algorithm using structured, readable steps.
flowchart — Diagram representing decisions and the flow of control among operations.
precondition — Requirement expected to be true before an operation starts, such as a nonempty input list.
postcondition — Property guaranteed by a correctly executed operation under its stated preconditions.
invariant — Property remaining true at a specified stage of repeated or structured computation.
termination — Property that an algorithm eventually stops rather than looping indefinitely.
correctness — Property that an algorithm satisfies its specification for all valid inputs.
robustness — Ability to handle malformed input or unexpected circumstances without unsafe failure.
time complexity — How an algorithm’s time requirements grow with input size under a chosen measure.
space complexity — How memory use grows with problem size, including temporary storage where relevant.
Big O notation — Asymptotic upper bound on growth rate, used to compare scalability rather than exact seconds.
linear search — Examination of sequence elements until a target is found or the collection ends.
binary search — Search of sorted data that repeatedly discards about half the remaining candidates.
bubble sort — Sorting technique repeatedly exchanging adjacent out-of-order elements.
insertion sort — Sorting technique inserting each new item into its correct place in the processed prefix.
merge sort — Divide-and-conquer sorting that recursively sorts partitions then combines them in order.

3. Data structures and recursive methods

array — Indexed collection used for storing multiple items, with implementation-dependent storage behaviour.
list — Ordered Python collection supporting indexed access and mutation, useful for sequences of values.
tuple — Ordered Python collection whose entries cannot be reassigned after creation.
dictionary — Mapping between unique keys and associated values, useful for fast lookup by identifier.
set — Unordered collection of distinct elements, useful for membership tests and removing duplicates.
stack — Last-in-first-out structure supporting push and pop operations, useful for nested tasks.
queue — First-in-first-out structure supporting enqueue and dequeue operations, useful for processing tasks in arrival order.
linked list — Structure of nodes containing data and references to subsequent nodes, supporting flexible insertion.
node — Unit of a linked structure holding data and one or more connections.
pointer — Reference identifying another memory location or object, depending on language and structure.
tree — Hierarchical structure of nodes with parent-child relations and no cycles.
binary tree — Tree in which each node has no more than two children.
binary search tree — Ordered binary tree enabling searches by relative key values when its ordering conditions hold.
root node — Top node from which other tree nodes descend.
leaf node — Node in a tree with no children.
graph — Collection of vertices connected by edges, representing networks or relationships.
vertex — Individual element or point within a graph.
edge — Connection between two vertices, possibly directed or weighted.
recursion — Technique in which a function solves smaller instances by invoking itself.
base case — Simplest case that terminates recursive descent and prevents unbounded recursion.

4. Web applications and the network

client — Software or device requesting a service from another system.
server — Computer or application responding to client requests and supplying data or operations.
request — Message asking a service to perform an action or return a resource.
response — Information and status returned after a request has been processed.
HTTP — Application protocol governing web requests and responses.
HTTPS — HTTP carried over an authenticated encrypted transport connection to protect communication.
URL — Address identifying the location or route of a resource on a network.
domain name — Human-readable identifier resolved to network addresses by a naming system.
DNS — Distributed system translating domain names into IP addresses or other network records.
IP address — Logical address used to identify network interfaces and route packets.
TCP — Transport protocol providing ordered, reliable data delivery between endpoints.
packet — Small unit of transmitted data with addressing and control information.
router — Network device forwarding packets between distinct networks.
switch — Network device forwarding traffic among devices within a local network.
latency — Delay between initiating a network operation and receiving its result.
bandwidth — Maximum rate at which a link can carry information, distinct from actual throughput.
HTML — Markup language expressing the meaning and structure of web content.
CSS — Style rules controlling presentation, layout and appearance of web pages.
Flask — Python web framework used to handle routes, requests and responses in a small application.
route — Web application mapping between a URL path and the code processing its request.

5. Databases and information handling

database — Organised collection of data managed for retrieval, consistency and updating.
DBMS — Database management system providing structured storage, queries and control of data.
relational model — Approach representing data as tables connected through keys.
table — Collection of related records organised into named columns.
record — One row representing a particular entity or occurrence.
field — Named data attribute stored in a table column.
primary key — Attribute or combination uniquely identifying each row of a table.
foreign key — Field whose value refers to a key in another table, expressing a relationship.
referential integrity — Rule ensuring relationships between table records remain valid.
normalisation — Restructuring relational data to reduce anomalies and needless repetition.
data redundancy — Unnecessary duplication of stored information, which can create inconsistent updates.
SQL — Language used to query and manipulate relational databases.
SELECT — SQL clause specifying which columns or expressions a query returns.
WHERE — SQL clause filtering rows according to a condition.
ORDER BY — SQL clause specifying sorting of selected rows.
JOIN — Operation combining related rows from different tables based on matching conditions.
GROUP BY — SQL clause grouping records so aggregate functions can summarise them.
aggregate function — Calculation such as COUNT, SUM or AVG applied to multiple records.
SQLite — Lightweight relational database engine used in the syllabus’s programming environment.
parameterised query — SQL statement separating query syntax from user-supplied values, helping prevent injection.

6. Data representation, storage and systems

bit — Single binary digit with value zero or one.
byte — Group of eight bits, often used to measure storage capacity.
binary representation — Encoding values using two symbols, zero and one.
denary — Base-ten numeral system used in everyday arithmetic.
hexadecimal — Base-sixteen notation using 0–9 and A–F for compact representation of binary values.
two’s complement — Method for encoding signed integers using a fixed number of bits.
character encoding — Rule mapping written characters into stored numeric codes.
Unicode — Standard assigning code points across writing systems and symbols.
UTF-8 — Variable-width encoding of Unicode text commonly used on the web.
pixel — Individual element of a digital image’s rectangular grid.
resolution — Count or density of sampled image elements, affecting level of detail.
colour depth — Bits used to encode the colour of a pixel in an image.
sampling — Measuring a continuous signal at discrete points to represent it digitally.
sample rate — Number of signal measurements taken per second during digitisation.
quantisation — Mapping sampled values to one of a finite set of representation levels.
lossless compression — Reducing file size without losing information needed to reconstruct the original exactly.
lossy compression — Reducing file size by discarding selected information that cannot be fully recovered.
metadata — Information describing data, such as format, origin or time collected.
data validation — Checking that input meets stated rules before storing or processing it.
data verification — Checking that copied or entered data corresponds to its intended source.

7. Testing, debugging and software development

software requirement — Statement of behaviour or quality a solution must provide to its users.
functional requirement — Specified action or output that a system must support.
non-functional requirement — Constraint on qualities such as security, usability or response time.
user interface — Part of a program through which people provide input and interpret results.
user experience — Overall usability and satisfaction arising from interaction with a system.
modular design — Separating code into components with defined responsibilities and interfaces.
encapsulation — Hiding internal implementation behind a documented interface.
code readability — Extent to which another programmer can understand a solution accurately.
documentation — Written description of design decisions, interfaces and usage.
test case — Specified input, procedure and expected outcome used to check behaviour.
unit test — Test of a small isolated piece of code, such as a function.
integration test — Test of multiple connected components interacting.
boundary-value test — Test near the limits of valid input, where errors often occur.
edge case — Uncommon but valid or foreseeable situation that may expose a defect.
exception — Exceptional event affecting normal execution that may be caught and handled.
debugging — Systematic process of locating and correcting faults in software.
syntax error — Violation of the language’s grammar preventing valid interpretation.
logic error — Program executes but produces the wrong result because the algorithm is flawed.
regression test — Test rerun after a change to ensure previously working behaviour remains correct.
version control — System tracking changes to source code and supporting recovery or comparison.

8. Privacy, ethics and cybersecurity

personal data — Information relating to an identifiable individual, protected in Singapore under applicable laws.
PDPA — Singapore’s Personal Data Protection Act governing collection, use and disclosure of personal data.
data minimisation — Limiting information collected to what is genuinely needed for a stated purpose.
informed consent — Meaningful permission granted with awareness of the relevant collection and use.
confidentiality — Protection of information against unauthorised access or disclosure.
integrity — Property that information remains accurate and protected against improper modification.
availability — Property that authorised users can obtain system services or data when needed.
authentication — Process of confirming a user’s or system’s claimed identity.
authorisation — Decision defining which actions an authenticated party is allowed to perform.
encryption — Transformation of readable data into encoded form that requires a suitable key to recover.
digital signature — Cryptographic method supporting authenticity and detection of unauthorised changes to signed data.
phishing — Deceptive communication attempting to obtain information or trigger harmful actions.
pretexting — Social engineering using a fabricated situation to manipulate a target.
shoulder surfing — Obtaining secrets by observing another person’s screen or input.
malware — Malicious software designed to damage, compromise or abuse systems.
ransomware — Malware typically blocking access to data and demanding payment.
denial-of-service attack — Disruption of a service by overwhelming or exhausting resources.
SQL injection — Attack arising when untrusted text changes the meaning of a database query.
cross-site request forgery — Attack inducing a browser to submit an unintended authorised request.
privacy by design — Including privacy safeguards within system design rather than adding them only after problems arise.

9. AI, learning and responsible prediction

artificial intelligence — Field building systems able to carry out tasks associated with intelligent behaviour.
machine learning — Approach in which a system derives predictive patterns from examples rather than only fixed rules.
training data — Examples used to fit a machine learning model.
test data — Examples held aside to estimate performance on data not used during training.
feature — Measured input property used by a model to make a prediction or grouping.
label — Known target category or value in supervised learning.
supervised learning — Learning from examples with known target outputs.
unsupervised learning — Finding patterns or groupings in data without supplied target labels.
classification — Task of assigning an input to one of several categories.
regression model — Predictive model estimating numerical outputs, distinct from the debugging term regression test.
k-nearest neighbours — Method predicting a new case from nearby labelled examples under a chosen distance.
k-means clustering — Unsupervised method partitioning points into k groups around repeatedly updated centroids.
centroid — Representative centre of a cluster used in k-means updates.
model training — Process of adjusting internal parameters or rules using data.
hyperparameter — Setting chosen outside the training process, such as the value of k.
overfitting — Model learning training-specific patterns that do not generalise well.
underfitting — Model too simple to capture important relationships in its training data.
confusion matrix — Table comparing predicted classes against actual classes for classification evaluation.
precision score — Share of cases predicted positive that truly belong to the positive class.
recall score — Share of actual positive cases correctly identified by the classifier.

10. Lab execution, evaluation and real-world use

JupyterLab — Interactive development environment supporting Python notebooks and exploratory experiments.
scikit-learn — Python library offering machine learning methods and model-evaluation tools.
data preprocessing — Cleaning and preparing raw data into a form suitable for analysis.
missing value — Data field lacking an observed value, requiring a deliberate handling strategy.
data leakage — Information from evaluation cases improperly entering training, giving misleadingly optimistic scores.
class imbalance — Unequal frequencies of target categories that may distort naive accuracy measures.
model evaluation — Assessment of performance using appropriate metrics and independent evidence.
accuracy metric — Proportion of total predictions that are correct; may hide weak minority-class detection.
F1 score — Harmonic mean of precision and recall, balancing two classification errors.
false positive — Incorrect prediction that an actual negative case belongs to the positive class.
false negative — Incorrect prediction that an actual positive case belongs to the negative class.
explainability — Ability to describe the basis or main factors shaping a model’s outputs.
algorithmic bias — Systematic unfair or distorted results arising from data, design or use.
human oversight — Meaningful ability of people to review, question and correct automated decisions.
reproducibility — Ability to obtain the same computational result under documented data and conditions.
deployment — Making a program or model available for actual use outside development.
input sanitisation — Handling untrusted input safely according to its intended context, alongside validation.
access control — Rules limiting which users or processes can read or change system resources.
incident response — Planned process for detecting, containing and recovering from security problems.
audit trail — Traceable record of actions or data changes used for review and accountability.

How computing vocabulary helps you solve a problem

Computer programming turns a human problem into structured steps. That begins by identifying inputs, outputs and constraints. It continues through decomposition into smaller responsibilities; selection of a data structure; design of a correct algorithm; implementation; and testing. A student who can quote the definition of a queue but chooses a stack for a first-come-first-served service has not yet used the vocabulary productively.

The revised syllabus also joins programming to public responsibility. A form collecting names and email addresses is simultaneously a web application, a database task and a personal-data concern. The technical response involves validating inputs, avoiding unsafe string-built SQL, restricting access to records and retaining only data genuinely required for the function. The user’s real-life problem, not a list of technologies, should determine how the system is designed.

Three short applied explanations

1. Algorithm choice: why a sorted list changes searching

Suppose a school stores 10,000 student reference numbers in sorted order. A linear search might compare many entries one after another, taking time that grows proportionally to the number of entries in the worst case. Binary search checks a middle value and removes half the remaining possibilities each time, so the number of comparisons grows logarithmically. But binary search requires the ordering assumption; if the list is unsorted, it can give a wrong result. The key issue is not which method sounds advanced but whether its preconditions hold.

2. Why database parameters are different from string concatenation

Imagine a search box asking for an item code. The safe program gives the database a fixed query template with a placeholder and passes the code separately as data. That prevents the input from becoming part of the SQL instructions. It is much safer than directly concatenating the user-provided text into a SQL command. Parameterised queries are a core design habit, but should be combined with validation, access controls and careful error handling.

3. Why accuracy alone can disguise a weak ML classifier

Consider a data set containing 990 negative and ten positive examples. A model predicting every case as negative reports 99% accuracy yet detects none of the positive cases. Its recall for the positive class is zero. A confusion matrix makes the failure visible, and precision and recall help assess whether the model is useful for the actual purpose. When the task involves false alarms or missed cases with unequal consequences, no single score should be treated as a universal answer.

Understanding the 2027 examination

Paper	What it tests	Time and weight
1. Written	All six syllabus modules, including cybersecurity, ethics and emerging technologies	3 hours; 100 marks; 60%
2. Lab-based	Compulsory computer solutions using modules 1, 2, 3, 4 and 6	3 hours; 100 marks; 40%

Important 2027 change: older 9569 materials described four sections, but the syllabus revised for 2027 explicitly organises learning into six modules. Paper 2 involves writing and saving soft copies of working code within examination time, using the approved laptop environment. The centre provides a programming reference guide. Knowing commands without practising a full run, test, correction and saved submission can leave students exposed.

A realistic lab-exam workflow

Read requirements: identify what is given, what must be produced and which edge cases a grader or user could encounter.
Sketch the data route: note which variables, lists or tables carry values between inputs, processing and outputs.
Build a minimal working version: implement an input, one key operation and an observable output before adding complexity.
Test ordinary and boundary cases: include empty data, repeated entries, unexpected text and the smallest valid values where appropriate.
Review database actions: use parameterised queries and check that keys and data types remain consistent.
Save exactly as instructed: use the exam’s required file names, locations and submission procedure; do not assume a notebook being open means a file is saved.

Common confusions

Term pair	Important difference
Assignment versus equality	Assignment changes a binding; equality comparison evaluates whether two values match.
Stack versus queue	A stack reverses processing order; a queue preserves first-in-first-out order.
Validation versus verification	Validation checks rules; verification checks agreement with source information.
Authentication versus authorisation	Identity checking does not itself grant permission to every resource.
Training accuracy versus test performance	Excellent training scores do not establish that a model generalises.
SQL data versus SQL instructions	Untrusted values should never be made part of SQL syntax through careless string construction.
Big O versus actual execution time	Asymptotic growth describes scalability, not an exact runtime across every input and computer.

A four-week JC2 revision cycle

Week 1: Python and algorithms. Study words 1–60; write short functions for searching, sorting and recursion, and trace both normal and edge cases.
Week 2: web, data and systems. Study 61–120; design a small Flask route, store records with SQLite, and explain the path from browser input through server to database.
Week 3: development, security and AI. Study 121–180; build tests and a confusion matrix, and explain how permissions, data quality and model choice change outcomes.
Week 4: complete lab tasks. Study 181–200; practise a timed end-to-end task, saving files properly and revising based on an explicit error log.

This four-week cycle is for organising study, not for learning two years of Computing from scratch. Students needing stronger programming foundations should first stabilise functions, loops and data structures before increasing lab-exam speed.

Frequently asked questions

Is the 2027 H2 Computing 9569 syllabus the same as the 2026 syllabus?

No. The 2027 syllabus has revised modules, including the addition of an explicit Emerging Technologies module and machine learning outcomes. Use the 2027 syllabus rather than copying an older course outline.

What software is expected for the lab paper?

The official 2027 syllabus lists Python and its relevant modules including Flask, JupyterLab and scikit-learn, plus DB Browser for SQLite and Notepad++. The examination centre provides the required environment; consult the latest SEAB listing for the exact permitted versions.

Does Paper 2 test the cybersecurity module directly?

The scheme of assessment lists Modules 1, 2, 3, 4 and 6 for Paper 2, while Paper 1 tests all six modules. Good security habits are still important whenever a student writes applications.

How do I practise AI without using a large commercial model?

Use a small, well-understood labelled data set and implement an appropriate model such as k-nearest neighbours. Evaluate it on separate examples and explain what mistakes the model makes. The aim is comprehension and justified design.

Will memorising syntax guarantee Paper 2 marks?

No. Students must also translate the problem into correct algorithms, debug unexpected cases and submit functioning files. Practice should include full programs, not isolated code snippets only.

How can a parent support a Computing student?

Ask the student to demonstrate their program with one normal example and one edge case and explain what they changed when it failed. This focuses attention on clear reasoning rather than extra hours of typing.

Official syllabus and eduKate companion reading

Editorially checked against the SEAB 2027 syllabus on 7 November 2026. Educational material; not an official MOE or SEAB resource.