The first argument of tree_predictor.cpp is the test (prediction) data in the format same as training data, and the output is the accuray of the predictor on the test data. For example, ./tree_predictor heart Usage of the Makefile.ta: 1. make -f Makefile.ta run_tree FNAME=heart.train EPS=3 : run the decision tree on data heart.train with epsilon = 3. 2. make -f Makefile.ta predict_tree PNAME=heart.test : make prediction by the decision tree model on data heart.test . You can play around with different FNAME, PNAME and EPS.