Python for data science — Final Project

PYTHON PROGRAMMING FOR DATA SCIENCE

FINAL PROJECT
Capstone data science project

Format: Group project + individual defense

Page 1
 Python for data science — Final Project

1. Overview
The final project is the capstone assessment of the entire course. Each group completes a full,
end-to-end data science project: starting from raw data, then cleaning, analyzing, building
predictive models, evaluating them, and drawing insights. It is your opportunity to demonstrate
the skills you have built over the 15 sessions.

2. Objectives
By completing this project, each student must demonstrate the ability to:
1. Carry out the full data science lifecycle: collect → clean → analyze → model → evaluate
→ interpret.
2. Flexibly apply Python, Pandas, visualization, and scikit-learn to a real, un-guided problem.
3. Build, compare, and justify the choice of machine learning models.
4. Present and defend their individual work through oral Q&A and live code modification.

3. Project at a glance
Item Detail
Format Group of 3–4 students (same groups) + individual defense
Released At the final session
Deliverables Notebook (.ipynb) + Report + Slides
Submission Google Drive, no later than 1 DAY BEFORE THE DEFENSE
Defense Individual — every member defends

4. Choosing your topics

Your role: Your group acts as a team of data analysts / data scientists given a real dataset and
a business question to answer. The mission is to turn raw data into a well-grounded decision or
recommendation.
Each group picks its own dataset and problem, but it must be approved by the instructor before
you start. The problem must be one of the following types (or a combination):
• Regression — predict a continuous value (e.g., house price, revenue, score).
• Classification — predict a label (e.g., customer churn yes/no, spam/not spam, pass/fail).

5. Where to find datasets

Use one of the reputable sources below to find your dataset. Prefer downloadable tabular files
(CSV / Excel). Always record the exact source link in your report and README.
Source Link Best for
Kaggle Datasets [Link]/datasets Huge variety across every domain;
free account needed to download

Page 2
 Python for data science — Final Project

UCI ML Repository [Link] Classic, well-documented

classification & regression datasets
Google Dataset Search [Link] Search engine that indexes datasets
across the whole web
[Link] (US) [Link] US government open data: health,
transport, economy
World Bank Open Data [Link] Global development, economic &
demographic indicators
Our World in Data [Link] Curated global data: health, climate,
energy, society
Hugging Face Datasets [Link]/datasets ML-ready datasets, including text &
tabular
[Link] [Link] Community-shared datasets, easy to
browse and preview
GSO Vietnam [Link] Vietnam official statistics — useful for
local-context topics

Tips for picking a good dataset:

• Pick something you find interesting — you will spend weeks with it.
• Open the file first and confirm it has a clear target variable to predict.
• “Messier” real data is fine — cleaning it is part of the grade.
• Avoid datasets so clean they need no work, or so famous that solutions are everywhere.

6. Requirements
The project must go through all the stages below. Each stage should be clearly shown in the
notebook (markdown explanation + code) and summarized in the report.

6.1. Data understanding

• Describe the dataset: source, size, meaning of each column, data types.
• State the business question clearly and identify the target variable to predict.

6.2. Data cleaning & preparation

• Handle missing values — state the strategy (drop / fill / impute) and the reason.
• Handle duplicates and outliers.
• Fix data types; process dates / strings as needed.
• Feature engineering: create / encode / scale features for the model.

6.3. Exploratory data analysis (EDA) + statistics

• At least 5 meaningful charts — each with a title, axis labels, and a one-line insight.
• Descriptive statistics for key variables (mean, median, std, distribution).
• Correlation analysis between features and with the target variable.

6.4. Modeling

Page 3
 Python for data science — Final Project

CORE REQUIREMENT
Required: train and COMPARE at least 4 models.
• Use a proper train/test split (train_test_split) and explain how you avoid data leakage.
• Justify why you chose each model — not just “run it for the sake of it”.
• At least one model must be tuned (features / parameters) to improve results.

Suggested models by problem type:

Regression
from sklearn.linear_model import LinearRegression
from [Link] import DecisionTreeRegressor
from [Link] import RandomForestRegressor

Classification
from sklearn.linear_model import LogisticRegression
from [Link] import DecisionTreeClassifier
from [Link] import RandomForestClassifier

6.5. Evaluation
• Regression: MSE / RMSE, MAE, R² — interpret what each number means in practice.
• Classification: confusion matrix, accuracy, precision, recall, F1 — state which metric
matters most for your problem and why.
• Compare models in a results table; pick the best model with justification.
• Comment on overfitting / underfitting if present.

6.6. Conclusions & insights

• At least 3 insights drawn from the data and the model.
• Recommendations based on the analysis.
• Limitations of the project and directions for future work.

7. Deliverables
Upload everything to your group's Google Drive folder, clearly named:
# Deliverable Requirement
1 Jupyter Notebook (.ipynb) Clean, commented, runs top-to-bottom; markdown cells
explain each step.
2 Report (PDF/DOCX) Max 10 pages. Required structure is given in Section 8.
3 Presentation slides Max 12 slides, used for the defense.

TIP
“Runs top-to-bottom” is a graded criterion: the instructor will open your notebook and run it
from the start. If it breaks midway due to a missing file/library or out-of-order cells, your group
loses points. Check it with Restart & Run All before submitting.

Page 4
 Python for data science — Final Project

8. Report structure
Your report (max 30 pages) must contain the following sections in order. Keep it concise —
quality of reasoning matters more than length.
Section What to include
1. Cover Page Project title, group name, all members with student IDs, course,
instructor, date.
2. Executive Summary ~½ page in plain language: the problem, your approach, and the
single most important result.
3. Introduction & Business Context/motivation, the specific question you answer, and your
Question objectives.
4. Dataset Description Source and link, size (rows × columns), meaning of key columns, and
the target variable.
5. Data Cleaning & Issues found (missing, duplicates, outliers) and how you handled
Preparation them, with justification; feature engineering.
6. Exploratory Data Analysis Key charts and the insight each reveals; descriptive statistics;
notable correlations.
7. Modeling Models tried and why; train/test split; tuning steps.
8. Evaluation & Comparison Metrics table comparing models; the chosen best model with
reasons; overfitting/underfitting notes.
9. Key Findings & Insights At least 3 insights tied back to the business question.
10. Conclusions, What you conclude, what you recommend, and the limits of your
Recommendations & work + future directions.
Limitations
11. References Data source, libraries, and any external material used.
12. Appendix AI Transparency Log summary and a link to the notebook / Drive
folder.

9. Individual defense (most important)

Because the course has no separate individual final exam, the defense is the decisive
checkpoint for individual ability. Every member must defend.
Item Detail
Duration About 5–10 minutes / student
Format Present your part + Q&A + live code modification

9.1. Q&A
The instructor may ask any member about any part of the project (including parts done by other
teammates). Example questions:
• “Explain what this block of code does, line by line.”
• “Why did the group choose this missing-value strategy and not another?”
• “Why is model A better than model B for this problem?”
• “Which part did you use AI for? Where did the AI get it wrong, and how did you fix it?”

9.2. Live code modification

Page 5
 Python for data science — Final Project

LIVE MODIFICATION
Required for every student.
The instructor gives a small request and the student must edit / extend the group's own code on
the spot, without AI. For example:
• Add a new chart or change the chart type.
• Filter the data by a different condition and re-print the result.
• Add an evaluation metric or change a model parameter and explain the change.

IMPORTANT
If a member cannot explain the code they submitted, or cannot perform a basic modification, that
member's individual score will be heavily reduced — regardless of how strong the group's work is.

10. Grading
Each student's final score = 50% group score + 50% individual score. A detailed rubric is
provided alongside this brief.

Group portion (50%) — shared by the whole group

Criterion Weight
Data cleaning & preparation 10%
EDA & descriptive statistics 10%
Modeling & evaluation 15%
Visualization quality 8%
Report & notebook structure (runs top-to-bottom) 7%

Individual portion (50%) — per student

Criterion Weight
Q&A + live code modification 30%
Individual contribution to the project 10%

TIP
The high individual weight (50%) is intentional: it ensures the grade reflects each person's real
ability, not how much they leaned on AI or on teammates.

11. AI usage policy

AI is a tool — the course does not ban it, but demands full transparency.
1. You may use AI (ChatGPT, Copilot, Claude, Gemini…) while working on the group
project.
2. You must log every AI use in the AI Transparency Log: the prompt, what the AI returned,
how you used/edited it, and what you learned.
3. You may not use AI during the individual defense.

Page 6
 Python for data science — Final Project

4. You must be able to explain every line of code you submit. Submitting code you do not
understand = no credit for that part.

12. Submission & Academic integrity

SUBMISSION
Submission: upload all deliverables to your group's Google Drive folder no later than 1 DAY
BEFORE THE DEFENSE.

• Late submission: penalized per class policy; not submitting before the deadline = not
eligible to defend.
• Plagiarism — copying another group's project/notebook, or work from a previous cohort:
zero for the entire project.

Page 7