Single Multiplicative Neuron Model Artificial Neuron Network Trained by Bat Algorithm for Time Series Forecasting

In recent years, artificial neural networks have been commonly used for time series forecasting. One popular type of artificial neural networks is feed forward artificial neural networks. While feed forward artificial neural networks give successful forecasting results, they have an architecture selection problem. In order to eliminate this problem, Yadav et al. (2007) proposed single multiplicative neuron model artificial neural network (SMNM-ANN). There are various learning algorithms for SMNM-ANN in the literature such as particle swarm optimization and differential evolution algorithm. In this study, differently from these learning algorithms, bat algorithm is used for the training of SMNM-ANN for forecasting of time series. The SMNM-ANN trained by bat algorithm is applied to two well-known real world time series data sets and its superior forecasting performance is proved by comparing with the results of other techniques suggested in the literature.