Feature Selection by Evolutionary Computing: application on Diagnosis by Pattern Recognition Approach

摘要

The choice of relevant parameters for pattern recognition is a key point of the diagnosis procedure. This paper presents a method for selecting the parameters based on the genetic algorithm (GA) which optimizes the choice of parameters by minimizing a selective function. This function is defined by a criterion which takes into account the data dispersion. In this paper, it is shown that GA is more efficient than well-known parameters selection method. Because of its robustness, it can be applied to any problem. Moreover, a strategy to determine the dimension of the representation space (choice of the number of parameters to be selected) is developed. This method is validated on an application: the detection of the various operating modes (healthy and faulty) on an induction machine.