A Hybrid Neural Network and Box-Jenkins Models for Time Series Forecasting

Time series forecasting plays a significant role in numerous applications, including but not limited to, industrial planning, water consumption, medical domains, exchange rates and consumer price index. The main problem is insufficient accuracy. present study proposes hybrid methods to address this need. proposed method includes three models. first model based on the autoregressive integrated moving average (ARIMA) statistical model; second back propagation neural network (BPNN) with adaptive slope momentum parameters; third hybridization between ARIMA BPNN (ARIMA/BPNN) artificial networks (ARIMA/ANN) gain benefits of linear nonlinear modeling. models are used predict indices index (CPI), expected number cancer patients Ibb Province Yemen. Statistical standard measures evaluate include (i) mean square error, (ii) absolute (iii) root (iv) percentage error. Based computational results, improvement rate CPI dataset was 5%, 71%, 4% for ARIMA/BPNN model, ARIMA/ANN respectively; while result patients’ 7%, 200%, 19% respectively. Therefore, it obvious that reduced randomness degree, alterations affected time data non-linearity. outperformed each its components when applied separately terms increasing accuracy decreasing overall errors forecasting.