RIDGE REGRESSION MODEL FOR THE PREDICTION OF STOCK INDEX WHEN MULTICOLLINEARITY OCCURS

Abstract

Stock prices are very dynamic and susceptible to quick changes because of the underlying nature of the financial domain. Difficulty in prediction of stock prices comes from the complexities associated with the market dynamics when parameters are constantly shifting and are not fully defined. Recently a lot of interesting work has been done in area of applying Machine Learning Algorithm for analyzing and predicting stock prices. In this paper an attempt is made to predict the daily closing prices of BSE sensex data using the daily opening price, high price and low price. Due to the volatile nature of the closing prices, the basic assumption of normality for parametric modeling of the data is not met. There by nor parametric neural network models are proposed. In addition the prediction variables considered are multi collinear and hence classical Ridge Regression model is fitted. Computation nonparametric neural network model surpassed classical statistical model in predicting the daily prices. Standard error measures are used to validate the prediction ability of the proposed models.