felix-last / evaluate-kmeans-smote

Evaluate the performance of the oversampling method KMeans-SMOTE

Evaluate K-Means SMOTE

This project is used to evaluate the performance of the oversampling method k-means SMOTE.

Dependencies

1. pip3 install -r requirements.txt

Usage

1. Set local folder paths using config.yml (see config.sample.yml for an example).
2. Open imbalanced_benchmark.py to check and adapt experiment_config, classifiers and oversampling_methods.
3. Execute imbalanced_benchmark.py.
4. Once the script has run, open the results folder created with the current timestamp. It contains a pickled experiment (or CSV files for older versions of imbalanced-tools) which can be read by imbalanced-tools.
5. Execute plot_imbalanced_benchmark.py with the timestamp as a parameter. This will generate a PDF file with plots in the experiment's folder.