Which regression model is best for predicting/forecasting stock prices?
Good question but I am afraid there is no simple answer. It really does depend on what you are trying to achieve.If you are trying to predict, tomorrow’s price then you will need a lot of computing power and software that can deal with the essentially random nature of short term price movements. The same approach used by the military to develop software that enables a missile to track an aircraft and intercept it, can be used on stocks. Take a look at the Kalman Filter if you are interested to know more.Over longer time periods, stock prices will typically follow a general direction which can be estimated with regression techniques such as OLS. However the problem with regression, is that it assumes the residuals are normally distributed but that is definitely not the case with stocks. Plus there is the issue of heteroskedasticity! So an approach is required to combat these issues with stock price data such as Stock Market Prediction with Multiple Regression, Fuzzy Type-2 Clustering and Neural Networks but requires a lot of geekery and raw computing power.Stock prices can exhibit mean reversion: this means that a stock will meander around a mean value and stay within 2 or 3 standard deviations of that mean but invariably return to the mean value at some time in the future. This is an ideal application for OLS regression to identify the mean path of stock price and then buy or sell that stock when it has reached a distance of 2/3 standard deviations. There is much written on this trading strategy on the internet so I wont elaborate further.Good luck.