Suppose there are X1...Xn numerical variables predicting a target variable Y (0 or 1)
Objective: to obtain the best possible thresholds and combinations of X1...Xn that can predict Y
Example: (X1>60 and X3<20) predicts Y=1 with 90% accuracy
Approach: Train a decision tree and select the conditions leading upto leaf nodes with high "purity"
Question 1: Is there a name for this process for generating threshold based rules by combination of variables?
Question 2: Are there better alternatives to decision tree for this process?
