This project is built for preference recommendation. EM algorithm taught by our professor John is used.
Original training data (first column is user id, second column is movie id, third column is rating. To simplify the problem, only considering rating = 1 as "like", rating = -1 as "dislike") has been offered.
The goal is to predict users' preference on thoes movies they haven't watched yet and recommend highest rating movies to them.
By comparing the test data with ground truth in the same format as training data, the confusion matrix containing true/false positives/negatives can be calculated and the accuracy of used algorithm can be estimated.