Multi-granulation Strategy via Feature Subset Extraction by Using a Genetic Algorithm and a Rough Sets-Based Measure of Dependence

Rough Set Theory (RST) is an effective technique for data analysis, which aims at approximate any concept (domain subset) by a pair of exact sets called the lower and upper approximations. In this research, we develop a machine learning method based on contexts (sets of attributes). It performs a multi-granulation based on a genetic algorithm in a way that it searches for the subsets of features having best values for the measure at hand (the degree of dependence from RST), but at the same time more distinct from each other. Then, an ensemble algorithm of the models obtained in each granule is applied. The proposed Genetic Algorithm and Rough Sets-based Multi-Granulation method exhibits satisfactory results compared to outstanding state-of-the-art algorithms.