How can I improve my recommender?

I have a recommendation system which uses user based and item-based collaborative filtering , I am unsure what the best way to combine these two techniques is however and I would appreciate any insight on improving each individuals component. User-based: Pearson similarity Reznick prediction KNN-neighbourhood selection

Item-based: Pearson similarity weightedSum predictiion kNN-neighbourhood selection

Any advice would be greatly appreciated thank you