Noise PGA

A simple genetic algorithm for feature selection with DEAP library

"Zhu, M., & Chipman, H. (2006). Darwinian Evolution in Parallel Universes: A Parallel Genetic Algorithm for Variable Selection. Technometrics, 48(4), 491-502" http://www.jstor.org/stable/25471241

"Chun-Xia Zhang et al.(2016) Randomizing outputs to increase variable selection accuracy. Neurocomputing, Volume 218, 19 December 2016, Pages 91-102, ISSN 0925-2312" http://www.sciencedirect.com/science/article/pii/S0925231216309638

Installation

Install the following requirements:

• deap
• scikit-learn

Example of usage

>>> import numpy as np
>>> from noisePGA import NoiseGAEnsemble
>>> from sklearn.datasets import make_regression
>>> from sklearn.linear_model import LinearRegression
>>> X, y, coef = make_regression(n_samples=150, n_features=100, n_informative=15, n_targets=1,
                                 noise=0.0, coef=True)
>>> lr = LinearRegression()
>>> nga = NoiseGAEnsemble(lr, ngen=15, ens_size=50, cv=5, s=1.2)
>>> nga.fit(X, y)
>>> print('TRUE: ', np.where(coef > 0)[0])
>>> print('PRED: ', np.where(nga.get_rmean() > 0.4)[0])