Modeling for Stock Price Forecasting in Colombo Stock Exchange: An Historical Analysis of Stock Prices

Abstract

Stock prediction with data mining techniques is one of the interesting areas being investigated in recent research. Weighted Moving Average (WMA) technique is one such widely used technique in stock forecasting, in which each historical data term can have its own weightage. One of the main drawbacks of WMA is that there is no exact base to determine those weighting factors. Because of this drawback, the investors can assign arbitrary weightages on periodical data though it is misleading the investment decisions. The present study addresses the limitations in Weighted Moving Average technique and tries to generate more reliable and statistically proven weighting factors for stock price prediction. We develop a more reliable stock predictive model by using a panel data set of quarterly closing stock prices of 41 companies for a period of ten years, approximating a sample size of 1680 observations. The Auto Regressive Moving Average model analysed in our study provides strong evidence for statistically significant impact of past stock prices on current stock prices. The study further found statistically more reliable weight factors for past four quarters which can be used for forecasting future stock prices. The findings of the present study confirm that the weight factor drops as the data become older in a linear pattern.