Evaluation of Accuracy between Item-Based and Matrix Factorization Recommender System

摘要

Recommender systems is used by e-commerce websites and streaming services to predict user opinion about products. This work examined two specific recommender algorithms, matrix factorization collaborative filtering algorithm and Item-based collaborative filtering, which utilizes item similarity. This study to compared the prediction accuracy of the algorithms using Mean Square Error and Root Mean Square Error criteria. The work yielded a results which indicated that the matrix factorization collaborative filtering algorithm is more accurate than the Item-based collaborative filtering algorithm. From the study, the results of evaluation metrics showed that matrix factorization method having RMSE of 0.916250 and MAE 0.708731 scales slightly better than the item-based method having RMSE and MAE of 0.937089 and 0.719434. The study concluded that the matrix factorization method is more accurate than the item-based method when evaluating their prediction accuracy using RMSE and MAE.