Precision in Predicting the Stock Prices – An Empirical Approach to Accuracy in Forecasting

Abstract

Forecasting the future prices of stock by analyzing the past and current price movements in determining the trend are always areas of interest of chartists who believe in studying the action of the market itself rather than the past and current performances of the company.  Stock price prediction has ignited the interest of researchers who strive to develop better predictive models with a fair degree of accuracy. The autoregressive integrated moving average (ARIMA)model introduced by Box and Jenkins in 1970has been in the limelight in econometrics literature for time series prediction, which has been at the core of explaining many economic and finance phenomena. ARIMA models in the research domain of finance and economics, especially stock markets, have shown efficient capability to generate short-term forecasts and have hence been able to outperform complex structural models in short-term prediction.

This paper presents stock price predictive model using the ARIMA model to analyze the sensitivity of such models to different time horizons used in estimation of trends and verifies the validity of such forecasts in terms of their degree of precision. Published historical stock data, on an actively traded public sector bank's share and historical movements in the banking sector index in which the selected bank is a constituent, obtained from National Stock Exchange(NSE), India and websites of Yahoo finance are used to build and develop stock price forecasts and index movement predictive models. The experiments with dynamic as well as static forecasting methods used revealed that the ARIMA model has a strong potential for short-term prediction and can offer better precision than from long term trend estimates. As a stock price prediction or index movement forecast tool, it can be relied extensively in deciding entry and exit to and from the volatile markets, notwithstanding the fact the risk the investor faces on account of noise or shocks still can be erroneous making the entire prediction irrespective of its degree of precision irrelevant.