A new predictive approach to variables selection through Genetic Algorithm and Fuzzy Adaptive Resonance Theory Using medical diagnosis as a case

:Variables selection is challenging task due mainly to huge search space. This study addresses the increasingly encountered chal- lenge of variables selection. It addresses the application of machine learning techniques to the problem of variables selection. We detailed the various models of the variables selection and examined the basic steps that are used to select the cost-effective predictors. We also walked through the initial settings and all variables selection stages, including architecture configuration, strat- egy generation, learning, model induction, and scoring. Results from this study show that the cost and generalization were seen to improve significantly in terms of computing time and recognition accuracy when the proposed system is applied for medical diagnosis. Good comparisons with an experimental study demonstrate the multidisciplinary applications of our approach.