Netflix type recommender system competition project on Flixer data from 2005

This is basically an experimentation on how to improve recommendation predictions for movies and users. 

I tried a number of more basic techniques

I incrementally improve on the prediction over time and at this moment have a hybrid user-item based collaborative filtering
system.  The primary similarity algorightm is a mixture of Cosine and Pearson and the prediction algorightm is Resnick we a small
enhancement on user based to include standard deviation. 

This might be useful for others exploring collborative filtering.