Rule and Decision Tree Induction
Andrew R. Webb, Keith D. Copsey · 2011
Classification or decision trees lie at the intersection of the areas of statistical pattern recognition and machine learning. On one hand they are an example of a nonparametric approach that models the classification/regression function as a weighted sum of basis functions. On the other hand, they can be used to generate interpretable rules, which can be very important in many applications. This chapter begins with a description of the basic decision tree model and methods for its construction. It then introduces rule-based approaches and presents two basic paradigms: (1) rule extraction from decision trees (indirect methods); and (2) rule induction using a sequential covering approach (direct methods). The chapter also considers the development of the recursive partitioning approach employed for decision tree construction to the modelling of continuous functions through the multivariate adaptive regressions splines (MARS) procedure. Controlled Vocabulary Terms multivariate adaptive regression splines; sequential analysis; statistical measures; tree digrams