reduce number of calls to pandas.DataFrame.values

Results in a 3 order of magnitude speedup for prediction on shallow trees trained on the adult dataset, based on our experiments

Author: HaydenMcT

gosdt/model/tree_classifier.py

@@ -152,8 +152,9 @@ def predict(self, X):
 
         predictions = []
         (n, m) = X.shape
+        samples = X.values
         for i in range(n):
-            prediction, _ = self.classify(X.values[i,:])
+            prediction, _ = self.classify(samples[i,:])
             predictions.append(prediction)
         return array(predictions)