Forecasting stock prices using a hybrid Artificial Bee Colony based neural network

Abstract

Financial Stock prediction presents a challenging task that attracts great interest from researchers and investors because of potential substantial rewards. However, the field still requires a more precise process. This paper presents an integrated system formed by data preprocessing techniques and a hybrid algorithm combining Artificial Bee Colony (ABC) and Back Propagation (BP) algorithms to train artificial neural networks (ANN) for stock price forecasting. Preprocessing techniques are used on the input data starting with haar wavelet transform to eliminate noise. For illustration and evaluation purposes several stocks in Tehran Stock Exchange Market are presented. As these simulation results demonstrate, the proposed hybrid method is promising in comparison with Genetic Algorithm, standard ABC Algorithm and different variations of BP algorithm.