A new algorithm relief hybrid (HRelief) for biological motifs selection

Feature selection plays a crucial role in the automatic learning field, since the non relevant and /or redundant ones can influence the strength of discrimination of a learning algorithm. In fact, select a minimum set of informative and relevant features can increase the performance of algorithms and the precision of prediction, minimize the time of data treatment, facilitates their visualization as well as their analysis. In this paper, we present a series of adaptations of algorithms for the motifs selection of Relief filtering algorithm. In the first two adaptation ways (HRelief1 and HRelief2) we transformed Relief in hybrid algorithms by using a classifier to evaluate the subset of the features generated. The third way of adaptation (HRelief3) helps in treating the problem of redundancy of features. Based on the experimentations done so far, these improvements resulted in an interesting outcome that encourages us to go into the depth of this orientation field.