ISEA2011 Machine Learning and Data Science (5 op)

By the end of the course, students will be able to:

1. Frame a data problem clearly and select suitable methods such as regression, classification, clustering, and dimensionality reduction based on the goal and constraints.

2. Acquire, clean, transform, and document datasets, and create clear visualizations that communicate findings and data quality issues.

3. Build baseline and improved models into software and structure the work into reusable, well‑tested components and pipelines.

4. Evaluate models with appropriate measures, perform error analysis, and justify choices for balanced and imbalanced classes.

5. Validate models rigorously using training, validation, and test splits; use cross‑validation; interpret learning and validation curves; and avoid data leakage.

6. Tune model hyperparameters systematically and reason about the bias–variance trade‑off and the role of regularization.

7. Engineer effective features for data, and integrate them into reproducible pipelines.

8. Explain model behavior using model‑specific and model‑agnostic techniques, and state limitations clearly.

9. Assess fairness and privacy risks in data and models, propose basic mitigations, and produce transparent documentation for data and models.

10. Package a trained model and expose it as a small web service, validate inputs, and write unit tests for data transformations and inference code.

11. Track experiments and version code, data, and models so that the teammates can reproduce results reliably.

12. Collaborate effectively in a small team using issues, reviews, and clear repository structure; communicate results with concise writing, visuals, and short demonstrations. Reflect on ethical and societal impacts of data‑driven features and define responsible deployment criteria for your project.