Multiobjective Differential Evolutionary Neural Network for Multi Class Pattern Classification

In this paper, a Differential Evolution (DE) algorithm for solving multiobjective optimization problems to solve the problem of tuning Artificial Neural Network (ANN) parameters is presented. The multiobjective evolutionary used in this study is a Differential Evolution algorithm while ANN used is Three-Term Backpropagation network (TBP). The proposed algorithm, named (MODETBP) utilizes the advantages of multi objective differential evolution to design the network architecture in order to find an appropriate number of hidden nodes in the hidden layer along with the network error rate. For performance evaluation, indicators, such as accuracy, sensitivity, specificity and 10-fold cross validation are used to evaluate the outcome of the proposed method. The results show that our proposed method is viable in multi class pattern classification problems when compared with TBP Network Based on Elitist Multiobjective Genetic Algorithm (MOGATBP) and some other methods found in literature. In addition, the empirical analysis of the numerical results shows the efficiency of the proposed algorithm. 1 Introduction Over the past few decades, there was a significant increase in using soft computing approaches. Artificial Neural Network (ANN) has become the substrate of soft computing methods, successfully used for solving different problems. Due to the importance of using ANNs in many applications, there are some different methods in the previous studies that focused on solving the problems of ANNs optimization, the training and structure of the network [1, 2]. Recently, there has been a remarkable increase in the use of Evolutionary algorithms (EAs) for solving optimization problems. The design of ANNs is considered one of the most important problems that need to be solved using this kind of algorithms. The earlier approaches tackled the single objective optimization problems in some of the previous works, PSO [3], GA[2] and DE[4] were considered for optimizing ANNs. These optimization techniques optimize only one factor, such as, hidden nodes or connection weights or optimizing training error rate. Though in ANNs optimization, there is more than one parameter that need to be optimized. Therefore, multiobjective optimization problems are preferred because of their ability to optimize more than one objective simultaneously. Evolutionary Algorithms (EAs) are good candidates for Multi objective optimization problems (MOOPs). This is because of their abilities to search for multiple Pareto optimal solutions and they perform better in global search space. Multiobjective evolutionary algorithms (MOEAs) research area has become one of the hottest areas in the field of evolutionary computation [5]. They are …