I. Association rules and decision trees
Given by Sven Laur
Brief summary: Advantages and drawbacks of machine learning. When is it appropriate to use machine and when knowledge based modelling is more appropriate. overview of standard experiment design. Potential applications and limits of machine learning. Broad classification of machine learning methods. Roadmap for the course. What is association rule and which rules are useful? Notions of support, confidence and applicability. Heuristic ways to define rules. Decision trees. ID3 algorithm for learning a good decision tree from examples. C4.5 algorithm for finding decision trees. CART algorithm for predicting continuous variables. Over-fitting and tree pruning.
Slides: PDF
Video: UTTV(2013)
Literature
- Lecture slides by Tom Mitchell
- Thomas Mitchell: Machine learning (1997) pages 52 - 80
Complementary exercises
- Thomas Mitchell: Machine learning (1997) pages 77 - 78
- Iris dataset
- Tutorial for using C4.5
Free implementations