Intelligent hybrid system for pattern recognition and classification

In this study, we critically analyse and compare performances of several global optimization (GO) approaches with our hybrid GLP_τS method, which uses meta-heuristic rules and a local search in the final stage of finding a global solution. We also critically investigate a Stochastic Genetic Algorithm (StGA) method to demonstrate that there are some loopholes in its algorithm and assumptions. Subsequently, we employ the GLP_τS method for neural network (NN) supervised learning, when using our intelligent system for solving real-world pattern recognition and classification problem. In the preprocessing data phase, our system also uses Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) for dimensionality reduction and minimization of the chosen number of features for the classification problem. Finally, the reported results are compared with Backpropagation (BP) to demonstrate the competitive properties and the efficiency of our system.