Decision tree implementation from scratch
-
project folder structure :
- DecisionTree - contains the implemntation of decision tree
- Test - contain the classification model build based on top of iris dataset (comparision with sklearn version of decision tree) - no parameter tunning is performed
-
Python version : v3.6
-
dependency : numpy v1.13.1
- Our Model Accuracy : 0.7368421052631579
- SK-Learn Model Accuracy : 0.7631578947368421
- Analyse the reson for the performance deviation with sklearn(76 % accuracy) to 73 % accuracy.
- use other performance metric - right now its a raw accuracy number used for comaprision
- test on more dataset fro UCI machine learning repository
- implement tree purning technique to reduce overfitting
- adapt tree for regression by creating differnt mechanism for creating terminal node
- try cross entropy for evaluting the split