zzeng13/Fourier-Feature-Selection

Fourier-Feature-Selection

Fourier Feature Generation and Selection (FFGS) method.

There are three major parts:

1. Disjunctive Normal Form (DNF) formulas generation (rules generation)
2. Parity feature generation and selection
3. Testing with multiple classifiers

Phase I: DNF formulas generation

This phase takes a raw data file in .arff format and generate a set of rules in .csv format.

Phase II: Parity feature generation and selection

This phase take DNF rules (.csv) and dataset with raw features (.csv) to generate and select best parity features.

The code is written in python with Google Colab notebook. You can also run it locally by downloading the .ipynb file.

Phase III: Testing with classifiers

In this phase, we compare the performance of raw features and the parity features using several classifiers. The code is written in Java, using the WEKA JAVA API.

Run ‘wekaClassificationTest.java’