An Artificial Electric Field Algorithm and Artificial Neural Network-Based Hybrid Model for Software Reliability Prediction

Artificial neural networks (ANNs) are popular nonlinear approximation techniques for solving complex multimodal functions. Performance of such methods is vastly depending on the learning method. In contrast to gradient-based ANN learning, nature-inspired optimization algorithms-based ANN training are found competent. Artificial electric field algorithm (AEFA) is a new optimization technique, needs fewer controlling parameters, and possesses robust learning ability. Its application to data mining problems is not yet explored. This article used AEFA to discover the best feasible weight and bias set as well as the number of hidden neurons of ANN, thus crafting an optimal ANN structure on fly. The hybrid model thus formed, i.e., AEFA + ANN is evaluated on modelling and predicting software reliability datasets. Experiments are conducted on real software failure datasets considering normalized root mean squared error statistics. Outcomes of result analysis, comparative studies, and statistical tests suggest that AEFA-ANN based model is suitable for forecasting.