Simple Undersampler for Python 2.7
- A simple program for undersampling asymmetric data sets saved in CSV format
The purpose of this program is to make quick undersampled data sets for training supervised classification models meant for classifying asymmetric data (data with lots of occurrences of one label but fewer of others). This program reads CSV data sets and writes an undersampled set to a new CSV file. The new set will contain a number of records with each label proportional to the label with the lowest number of records. Records with labels occurring at higher proportions will be selected at random from the original set and written to the output file.