CV0003: ace the quiz, not just read the notes
Your complete guide to Nanyang Technological University's introduction to data science and artificial intelligence course. See where the marks are, work real practice questions, and study with an AI tutor that knows CV0003.
Sia generates CV0003 practice questions, walks through the data pipeline and data exploration step by step, and quizzes you on the material the quiz that weights most heavily.
Find what is wrong
A student is building a classifier for the CV0003 mini project and reports 96% accuracy on the held-out test set. The grader is unconvinced. Here is the relevant code:
scaler = StandardScaler()
X_scaled = scaler.fit_transform(X) # all rows
X_train, X_test, y_train, y_test = train_test_split(
X_scaled, y, test_size=0.2, random_state=0)
model = DecisionTreeClassifier().fit(X_train, y_train)
print(model.score(X_test, y_test))
What is wrong?The split happens after the scaling. That ordering is the bug: fit_transform computed the mean and standard deviation over every row, including the rows that later became the test set.
The fix is to split first and fit the scaler only on the training portion: scaler.fit(X_train), then apply the fitted scaler to both: scaler.transform(X_train) and scaler.transform(X_test).
In practice a Pipeline that chains the scaler and the model makes the mistake harder to commit, because the transformation is refitted inside each training fold rather than once over everything.
The trap: Reading a suspiciously high accuracy as success rather than as a symptom. Data leakage does not raise an error, it raises your score, which is exactly why it survives to the presentation. When a mini project reports a number well above what the problem plausibly allows, check the order of operations before celebrating: any step that learns parameters from data, including scaling, imputation and feature selection, has to be fitted on the training set alone. classic slip!
One quiz decides 40% of your grade. Compulsory: the outline states that sitting this quiz session is a requirement for passing the course. It is both the largest single component and a hurdle. This whole page is built around that.
Overview
What CV0003 is, and where it sits
CV0003 is the introduction to data science and artificial intelligence in the NTU College of Engineering, and it is unusually broad for a first-year course: it runs from problem formulation and data wrangling through to reinforcement learning in thirteen weeks. Python is the language throughout.
The first half is the data pipeline end to end. You start with what a data science problem even is, then move through data types and wrangling with Pandas, exploratory analysis and basic statistics, visualisation, and then inference: prediction with regression and time series, classification with decision trees and support vectors, and identification through clustering and anomaly detection, all using Scikit-Learn. Digital storytelling closes that arc, because the course treats communicating a result as part of producing it.
The second half turns to artificial intelligence: state space representation and search, including breadth-first, depth-first, iterative deepening and uniform-cost, then reinforcement learning with Markov processes and Q-learning. The course finishes on ethics in data science and AI and the state of the art in big data, neural networks and deep learning.
Delivery is split: topics arrive as online video sessions, and the face-to-face example classes are used for discussion and hands-on work, including guidance on the mini project. That structure means the timetabled hours are not where most of the learning happens.
Always treat your own course outline and the exam timetable as authoritative.
Difficulty & time commitment
Is CV0003 hard, and how much time does it take?
CV0003 is manageable if you keep a weekly rhythm and treat the back half as the main event. The pattern is consistent: it starts gently and steepens, and the heaviest assessment is the part that separates grades.
The difficulty curve and the assessment weighting point the same way: the back half is harder and worth more. Front-loading effort there is the highest-return decision in the course.
Is this course for you
Who tends to do well, and who tends to struggle
You will likely do well if
- You keep pace with the online video sessions. Delivery is front-loaded online, and the TEL component tracks whether you are actually watching them.
- You are comfortable in Python and willing to work in Pandas and Scikit-Learn from the early weeks.
- You can carry a team project alongside continuous assessment, since the two run in parallel all semester.
- You enjoy breadth. The course covers the full pipeline plus AI foundations rather than going deep in one place.
You may struggle if
- You treat the quiz date casually. It is 40% of the mark and sitting it is a stated requirement for passing.
- You expect to learn Python during the course. The prerequisite assumes it, and Week 2 starts wrangling immediately.
- You leave the abstract material to the end. Markov processes and Q-learning arrive in Weeks 11 and 12, when the project is also due attention.
- You are a passive team member. The modification factor from peer assessment and panel judging means the team mark is not evenly distributed.
- Pick a mini project topic where the data is genuinely messy. Clean data hides the wrangling skill the course spends its first weeks building, and graders notice.
- Validate before you present. Check every model result for leakage, and be able to say what would falsify your conclusion.
- Implement one search algorithm by hand rather than reading about all four. Understanding why uniform-cost differs from breadth-first is worth more than being able to name them.
- Write the data story for a reader who has not seen your notebook. The course assesses communication explicitly, and most teams under-invest there.
Syllabus
The 13 topics, week by week
The exam-weight marker on each topic shows where the marks concentrate. The amber topics carry the highest exam weight.
W1 · Data-analytic thinking
Topic 1What data science is, the core problems it addresses, and how to turn a real situation into a formulated problem. The framing week that the mini project later depends on.
W2 · The data pipeline
Topic 2Types of data in practical scenarios, then extraction, wrangling, cleaning and preparation using Pandas. Unglamorous and where most real project time goes.
W3 · Data exploration
Topic 3Basic statistics and exploratory data analysis, applied to case studies with both structured and unstructured data.
W4 · Data presentation and visualisation
Topic 3Visualisation tools in Python and the principles behind choosing a representation that supports the inference rather than decorating it.
W5 · Data-driven predictions
Topic 4Prediction using regression and time series techniques, implemented with Scikit-Learn.
W6 · Data-driven classification
Topic 4Classification with decision trees and support vector methods, and the question of how you know a classifier is actually working.
W7 · Data-driven identification
Topic 5Clustering and anomaly detection: finding structure without labels, and the judgement that requires.
W8 · Digital storytelling
Topic 6Dashboards, notebooks and presentation with Plotly. The course assesses communication of a result as part of the result, so this is not a soft week.
W9 · Artificial intelligence and state space
Topic 7What AI is, a short history and current state, then the principle of representing a problem as a state space to be searched.
W10 · Uninformed search
Topic 7Breadth-first, depth-first, iterative deepening and uniform-cost search, with the trade-offs between completeness, optimality and cost.
W11 · Reinforcement learning I
Topic 8Reinforcement learning in the context of AI, and the fundamentals of Markov processes.
W12 · Reinforcement learning II
Topic 8Q-learning and case studies. Conceptually the densest fortnight of the course for most students.
W13 · Ethics and the state of the art
Topic 9, Topic 10Ethical considerations and responsible practice in data science and AI, then progress in big data, neural networks and deep learning.
How it's assessed
Assessment structure
| Component | Weight | Format & timing |
|---|---|---|
| TEL participation and MCQs | 10% | Individual, online. Scores, correct answers and explanations are returned immediately on submission. Across the semester. Runs alongside the online video delivery, so it tracks whether you are keeping up with the material week by week. |
| Quiz | 40% | Individual. The date is announced in the first teaching week. Announced in Week 1. Compulsory: the outline states that sitting this quiz session is a requirement for passing the course. It is both the largest single component and a hurdle. |
| Coding assignment in class | 20% | Individual. Partly online MCQ exercises with immediate feedback, partly classwork submissions evaluated individually. In class, across the semester. There are no make-up opportunities for in-class activities. |
| Mini project with presentation | 30% | Team, developed with instructor guidance through the example classes, with regular progress check-ins. Across the semester, presented at the end. A modification factor derived from panel judging and peer assessment is applied to this 30%, so individual contribution changes the individual mark. |
- Sitting the compulsory quiz session, which carries 40%, is stated as a requirement for passing the course. There are no make-up opportunities for in-class activities.
- There is no final examination. The mark is assembled from online participation, one substantial quiz, an in-class coding assignment and a team mini project with presentation.
- Calculator policy: Not stated in the published course outline. The course is conducted in Python, using Pandas, Scikit-Learn and Plotly.
This is a coursework course. Coursework carries 100% of the grade and the quiz is the single heaviest piece at 40%, so steady work across the semester decides your result more than any one sitting. Compulsory: the outline states that sitting this quiz session is a requirement for passing the course. It is both the largest single component and a hurdle.
How to actually pass it
A weekly rhythm, two checklists, and the traps to avoid
The course rewards consistency over cramming, and practice over re-reading. Here is the loop that works, then what to have nailed before each exam.
The weekly loop
Before the mid-semester checklist
- Wrangling and cleaning in Pandas, including handling missing and malformed data
- Exploratory analysis and the basic statistics that support it
- Visualisation choices that carry the argument rather than decorate it
- Regression and classification with Scikit-Learn, plus how you would know the model is working
Before the final heaviest topics
- Clustering and anomaly detection, and when unlabelled structure is meaningful
- State space representation of a problem
- Breadth-first, depth-first, iterative deepening and uniform-cost search, compared on completeness and cost
- Markov processes and Q-learning at the level the course develops them
- Ethical considerations in data science and AI, applied to a concrete case rather than recited
The mistakes that cost marks
Fitting a transformation before splitting the data. Scaling, imputation and feature selection all learn parameters from data. Fitting them on the full dataset leaks test information into training and inflates your reported score without raising any error.
Reporting accuracy alone. On imbalanced data a high accuracy can be produced by a model that never predicts the minority class. Say what the number means before presenting it.
Treating the online sessions as optional. Delivery is deliberately split, with content online and application in class. Skipping the videos means arriving at the assessed hands-on session without the material.
Splitting the project by section rather than by understanding. With a peer-assessed modification factor, a member who only knows their own slice is exposed in both the presentation and the assessment of contribution.
Teaching team
Who teaches CV0003
The bios below are factual. We do not rate lecturers; any star ratings are submitted by students who have taken CV0003.
Lee-Chua Lee Hong
Listed as the faculty member proposing and revising CV0003 in the course outline, and the author of its assessment structure and planned schedule.
Teaching team as listed in the course materials reviewed. AskSia does not rate lecturers; star ratings are submitted by students who have taken CV0003.
Formula & concept sheet
The vocabulary and formulas you must own
- Data wrangling
- Extracting, cleaning and reshaping raw data into a form that can be analysed. Usually the largest share of real project time.
- Exploratory data analysis (EDA)
- Examining a dataset's structure and distribution before modelling, to find what questions it can actually answer.
- Regression
- Predicting a continuous outcome from input features; in this course the first supervised technique applied.
- Classification
- Predicting a category, using methods including decision trees and support vector approaches.
- Clustering
- Grouping observations without labels, so that structure is discovered rather than supervised.
- Anomaly detection
- Identifying observations that do not fit the pattern the rest of the data establishes.
- Train and test split
- Holding back part of the data so performance is measured on observations the model has not seen.
- Data leakage
- Information from the held-out data influencing training, which inflates measured performance without any error being raised.
- State space
- Representing a problem as states and transitions, so that solving it becomes searching a graph.
- Uninformed search
- Search that uses no problem-specific guidance: breadth-first, depth-first, iterative deepening and uniform-cost.
- Markov process
- A process where the next state depends only on the current state, the foundation for the reinforcement learning material.
- Q-learning
- Learning the value of taking an action in a state from experience, without a model of the environment.
- Digital storytelling
- Communicating an analysis so that the audience can follow the inference, treated here as part of the work rather than a wrapper on it.
Common acronyms: {'term': 'DS&AI', 'def': 'Data science and artificial intelligence'} · {'term': 'EDA', 'def': 'Exploratory data analysis'} · {'term': 'TEL', 'def': 'Technology-enhanced learning, the online delivery component'} · {'term': 'MF', 'def': 'Modification factor, applied to the team project weighting'} · {'term': 'CA', 'def': 'Continuous assessment'} · {'term': 'AU', 'def': 'Academic Units, the NTU credit measure'} · {'term': 'ILO', 'def': 'Intended Learning Outcome'}.
Where it fits
Prerequisites, related courses & why it matters
Prerequisite: CV1014 Introduction to Computational Thinking. The course assumes you can already write basic Python before the Pandas work begins in Week 2.
Your CV0003 study toolkit
Study the course with Sia, not just read about it
Each tool already knows CV0003: your syllabus, your texts, and where the marks are. Grouped by how you study, from first contact to exam week.
FAQ
Frequently asked questions
How is CV0003 assessed?
Quiz 40%, mini project with presentation 30%, in-class coding assignment 20%, TEL participation and MCQs 10%. There is no final examination.
Is the quiz really compulsory?
Yes. The outline states that sitting the 40% quiz session is a requirement for passing the course, and the date is announced in the first teaching week. Treat it as fixed from day one.
What programming background do I need?
The prerequisite is CV1014 Introduction to Computational Thinking. You should be able to write basic Python before Week 2, when data wrangling with Pandas begins.
Is there a textbook?
No single set text. The outline names three references: the Python Data Science Handbook, An Introduction to Statistical Learning, and Artificial Intelligence: A Modern Approach, with further material shared through the online sessions and example classes.
How does the team project mark work?
The 30% team component has a modification factor applied, derived from panel judging and peer assessment, so individual contribution affects the individual mark rather than everyone receiving the team score.
Which part do students find hardest?
The reinforcement learning fortnight in Weeks 11 and 12. Markov processes and Q-learning are the most abstract material in the course and arrive late, when project work is also demanding attention.
Study CV0003 with Sia
Work through the data pipeline, data exploration, data-driven predictions and the rest of the course with a tutor that knows it and quizzes you on the topics the assessments weight most heavily.
Start studying with Sia
