A Novel Group Teaching Optimization Algorithm Based Artificial Neural Network for Classification

Abstract

AbstractTraditional techniques of training Artificial Neural Networks (ANNs) i.e. the Back Propagation algorithm (BPA), have high computational time and number of iterations and hence have been improved over the years with the induction of meta-heuristic algorithms that introduce randomness into the training process but even they have been seen to be prone to falling into local minima cost solutions at high dimensional search space and/or have low convergence rate. To cater for the inefficiencies of training such an ANN, a novel neural network classifier is presented in this paper using the simulation of the group teaching mechanism to update weights and biases of the neural network. The proposed network, the group teaching optimization algorithm based neural network (GTOA-NN) consists of an input layer, a single hidden layer of 10 neurons, and an output layer. Two University of California Irvine (UCI) database sample datasets have been used as benchmark for this study, namely ‘Iris’ and ‘Blood Transfusion Service Center’, for which the training accuracy is 97% and 77.9559% whereas the testing accuracy is 98% and 83.5341% respectively. Comparative analysis with PSO-NN and GWO-NN unveil that the proposed GTOA-NN outperforms by 0.4% and 4% in training accuracy and 2.6% and 6.8% in testing accuracy respectively. KeywordsMeta-heuristicNeural networkIntelligent control systemGroup teaching optimization algorithmClassification