A Hybrid ARFIMA Wavelet Artificial Neural Network Model for DJIA Index Forecasting

This paper proposes a hybrid modelling approach for forecasting returns and volatilities of the stock market. The model, called ARFIMA-WLLWNN integrates advantages ARFIMA wavelet decomposition technique (namely, discrete MODWT with Daubechies least asymmetric filter) artificial neural network LLWNN network). model develops through two-phase approach. In phase one, improves accuracy network, resulting in Wavelet Local Linear Neural Network (WLLWNN) model. Back Propagation Particle Swarm Optimization (PSO) learning algorithms optimize WLLWNN structure. two, residuals an conditional mean become input to is evaluated using daily Dow Jones Industrial Average index over 01/05/2010 02/11/2020. experimental results indicate that PSO-optimized version outperforms LLWNN, WLLWNN, ARFIMA-LLWNN, ARFIMA-HYAPARCH models provides more accurate out-of-sample forecasts validation horizons five twenty-two days.