Author: Pavlo Boidachenko
Description:
Small Python script for performing holdout(dividing
dataset on test and train parts) for LIBSVM dataset format.
Script has guarantee that class ratio will be the same in train
and test parts.
Warning: there is no check for correctness your dataset file.
Warning: script assumes that your classes is a first column
Usage:
For classification:
python3 h4c.py dataset percent_on_test train_part test_part
For regression:
python3 h4r.py dataset percent_on_test train_part test_part
dataset - input dataset in libsvm format
percent_on_test - how much percent from dataset you divide for testing.
train_part - output file will contain training part
test_part - output file will contain testing part
Algorithm:
Classification Script devides dataset by classes than takes random lines from
each class and puts them into training and testing files. Ratio
beetween test and train parts is defined by user. Basically classification script
tries to save class ratio in test dataset.
Regression Script randomizes dataset and puts coresponding parts into train_part
and test_part.