The Enhancement of Evolving Spiking Neural Network with Firefly Algorithm

Abstract

This study presents the integration between Evolving Spiking Neural Network (ESNN) and Firefly Algorithm (FA) for parameter optimization of ESNN model. Since ESNN lacks the ability to automatically select the optimum parameters, Firefly Algorithm (FA), as one of nature inspired metaheuristic algorithms is used as a new parameter optimizer for ESNN. The proposed method, ESNN-FA is used to determine the optimum value of ESNN parameter which is modulation factor (Mod), similarity factor (Sim) and threshold factor (C). Five standard datasets from UCI machine learning are used to measure the effectiveness of the proposed work. The classification results obtained shown an increase in accuracy than standard ESNN for all dataset except for iris dataset. Classification accuracy for iris dataset is 84% which less than standard ESNN with 89.33%. The rest of datasets achieved higher classification accuracy than standard ESNN which for breast cancer with 92.12% than 66.18%, diabetes with 68.25% than 38.46%, heart with 78.15% than 66.3% and wine with 78.66% than 44.45%.