STAT3612 Homework 3: Binary Classification
Date: November 24, 2020
Sumbit (in the ipynb format) via Moodle before 11:59pm December 6, 2020.
Use the load_breast_cancer() from sklearn.datasets to get a copy of the breast cancer
(diagnostic) data with 569 instances and 30 numeric predictive attributes. The binary
responses include 212 Malignant and 357 Benign cases. Use train_test_split to divide
the data into 80% training data and 20% testing data, then perform the following machine
learning tasks.
Step 1. (20%) Fit a decision tree classifier with max_depth =3. Visualize the fitted tree by
export_graphviz. Report the training and testing accuracy.
Step 2. (20%) Fit the random forests and gradient boosting machines. Report the training
and testing accuracy for both models.
Step 3. (20%) Fit support vector classifiers with linear and RBF kernels. Report the training
and testing accuracy for both models.
Step 4. (20%) Fit a neural network with two hidden layers each having 40 nodes. Report
the training and testing accuracy.
Step 5. (20%) From the above model fits, pick the one with the best testing accuracy. Run
post-hoc analysis for model interpretation in terms of a) feature importance and b) partial
dependence plots of the 5 leading important features.
1