Concept learning and tree models

4 Outubro 2016, 13:00 • João Marques Silva

The fourth lecture reviewed the construction of decision trees and covered three main topics. First, we introduced notation and definitions to be used throughout the semester. These include definitions for binary classification, formalization of different ML scenarios, and definitions of coverage plots and curves and associated ROC plots and curves. Second, we continued the study of tree models, by relating learned decision trees with coverage curves. Third, we studied rule models, namely the inference of lists of rules and sets of rules. In addition, we studied two examples of descriptive ML: subgroup discovery and association rules. We also studied frequent itemset mining, to be used for learning association rules.