A Comparison of Different Hybrid ARIMA - Neural Network Models for Stock Index Return Forecasting and Trading Strategy

Abstract

This study presents the application and development of three different hybrid methodology that combines both the ARIMA and ANN model to take advantage of the unique strengths of both linear and nonlinear modeling to model and predict the stock market index returns. The input to the hybrid ARIMA - neural network models, includes (a) past returns, the forecast and residual (results from ARIMA model), (b) forecast and residual value (results from ARIMA model) - ARIMABP model and (c) the forecast (results from ARIMA model) and the forecast residual (results of ARIMA model) of neural network. On the other hand, the benchmark model includes, ARIMA and neural network models. The performance of the models are evaluated in terms of widely used statistical metrics, correctness of sign and direction change, and various trading performance measures like annualized return, Sharpe ratio, maximum drawdown, annualized volatility, average gain/loss ratio etc via a trading strategy. The findings of the study reveals that ARIMABP model achieve greater accuracy based on the traditional forecasting accuracy measures, sign and directional change and trading experiments and add value as a forecasting and quantitative trading tool.