Visualization of Personalized Pruning of Decision Tree
- Yue Kang
- Hua Feng
Display the result of decision tree from Scikit-learn of python. Python Scikit-learn provides the functionality of classification using decision tree model. Although a high test accuracy can be reached by a classifier with sufficient feature and appropriate tuning, if people can’t see the reasoning behind the classifier, they might regard it as untrustworthy. Based on the idea of LIME, we define our task to visualize the classfication process of a decision tree. This task aims to aid users make their judgement on whether to trust the classifier or not by making the decision making process transparent. It would also aid users to understand the necessity of clean the data while training the decision tree.
This project will provide the user with function to upload their decision tree and visualize it. Then it will make classification with the decision tree and testing data. We will try to make the classfication process dynamic to help users understand the decision making process. It should also show how samples are partitioned by decision nodes on each level by gradually developing a sunburst chart.
For implementation, we plan to visualize with D3. The classifers and data would be exported to D3 as json or csv file from python.
Project Progress Presentation
- Executable project : http://web.stanford.edu/~yuekang/cs448b/index.html