Auto-Regressive Integrated Moving Average (ARIMA) Model of Technical Analysis – A Study on NSE Nifty 50 Index

Abstract

Technical Analysis tries to estimate the future price of a stock or an index typical based past price action. Essentially technical analysts believe that fundamental view of the organization or the sector is already priced in, and the future prices move in certain patterns based on its own historical behaviour. There is a wide spectrum of techniques and methods in the area of technical analysis, which essentially try to reduce the forecast errors. The Auto-regressive Integrated Moving Average (ARIMA) model is among the first methods. These classes of models assume that a stock index is basically a time-series and that an index is a linear stationary random process. In this research an attempt is made to develop prediction models for stock indices using LSTM and CNN architectures – as both standalone models and with ARIMA as hybrid variants. The results indicate hybrid models perform better than standalone variants and the new hybrid models proposed in this study. These architectures are investigated in combination with ARIMA as well as in standalone settings. The results of this investigation show that hybrid models—which combine LSTM, CNN, and ARIMA—perform better and have higher prediction accuracy than standalone versions. The new hybrid models that are presented here increase financial market predictive modelling and present opportunities to improve trading and investment strategy decision-making.