It's been a while. Takes so long for an engineer to actually do it to come up with common sense conclusions. Be it linear regression or A.I. or other type of statistics based pattern-matching/classification methods, the tool is only as good as the data. If the data has predictive value to future price movement, most likely from sentiment and supply/demand, even naked eyes would be able to do the forecast. Btw, look at http://www.pyfia.com, choose your own major indices to perform a short term prediction on your favorite stock. Some of the predictors to choose from may have causal connection to your target.