Rule Induction Partitioning Estimator

To find an easy way to describe a complex model with a high accuracy is an important objective for machine learning. Many research fields such as medicine, marketing, or finance need algorithms able to give a reason for each prediction made. Until now, a common solution to achieve this goal has been to use induction rule to describe cells of a partition of the features space X. A rule is an If-Then statement which is understood by everyone and easily interpreted by experts (medical doctors, asset managers, etc.). We focus on rules with a If condition defined as a hyperrectangle of X. Sets of such rules have always been seen as decision trees, which means that there is a one-to-one correspondence between a rule and a generated partition cell.