Computational Intelligence Approaches for Stock Price Forecasting

Abstract

Computational intelligence (CI) approaches such as neural networks (NNs) and neuro-fuzzy approaches have been used for stock price forecasting. Robust and efficient stock market models can achieve more accurate predictions and decision making for individual investors or stock fund managers. This work thus surveys individual and hybrid CI methods, including a self-organizing polynomial neural network (SOPNN) based on statistical learning algorithm, cerebellar model articulation controller NN, standard back propagation NN (BPNN) with the steepest descent method (BPNN-GD), BPNN with scaled conjugate gradient (SCG) method, artificial immune algorithm-based BPNN (AIA-BPNN), advanced simulated annealing-based BPNN (ASA-BPNN) and adaptive network based fuzzy inference system (ANFIS) method. The performances of these methods are evaluated by using the Taiwan Stock Exchange Capitalization Weighted Stock Index (TAIEX) dataset collected from the Taipei Stock Exchange, and root mean square error (RMSE), mean absolute difference (MAD) and mean absolute percent error (MAPE) are used as the performance indices. Experimental results show that the best SOPNN, CMAC NN, BPNN-SCG, AIA-BPNN, ASA-BPNN and ANFIS obtain identical training and test accuracies. Particularly, hybrid CI approaches such as AIA-BPNN and ASA-BPNN are recommended for stock price forecasting, since these methods have the lowest test RMSE, MAD and MAPE.