torgeros / ecse-557-project-testcase

McGill ECSE 557 Project Test Case

ECSE 557 Project Test Case

This belongs to a McGill University project. The related dataset modification tool can be found in elements72/ECSE-557-project

classifier.py usage

Warning: enum_str does enumerate just fine, but in the case of binned/grouped values, they are still not ordered...

Datasets

Datasets can be found in /datasets.

filename	description
full	original dataset as exported from kaggle
base	blank ID for all entries. all following are based on base. Run enum on ID to set to 0
base_debiased	as many zero-targets as one-targets. training bias is too big otherwise.
b_d_inc_*	base_debiased, but with income grouped to filename based bin size. Create using our tool.

Classifier python files

Executable python scripts can be found in the root directory, the are named according to the used dataset (see above table).

The print_b_d_json_accuracies.py script is used to evaluate the json files logged by the b_d_inc_* classifier and create confusion matrices used in the report.

Files that should not be ran directly (because they are internal libraries of sort) are

• classifier.py, providing the actual classifier code
• conf_mat_plt.py, a utility for creating a confusion matrix with matplotlib

The helpers directory contains a script to create the base_debiased data set from base.