Machine Learning Algorithms Used in Recommendation Systems

Recommendation systems help people by recommending items which suit their preferences. In collaborative filtering approach, methods recommend items to the users from other users whose past activities are similar to the recommendee users.In this project a review of collaborative filtering (cf) area on recommendation systems are given. After the review three models named SVD, Biased SVD and NNMF that are categorized under dimensionality reduction topic of model-based collaborative filtering models are implemented along with a simple naive bayes probabilistic model. The performance of the models are measured by three metrics which are RMSE, MAE and AUC. As data MovieLens 100k dataset has been used. To increase the validity of the measurements k-fold method with k = 10 has been applied to the data. According to the results of the measurements SVD gave the best results on RMSE and MAE metrics while Naive Bayes model has the best performance on AUC compared to other models.