Enhancing Online Analytical Processing and Recommender System for User Consideration

Abstract

MovieLens dataset is used for experiments. To evaluate the proposed system, there are two testings have been implemented. In the first test, when a take is 10 closest and 10 furthest about target user from neighbors users, the results showed that precision is about 92%, recall about 91%, average V diversity score about 7.66, Discounted Cumulative Gain(DCG), and Normalized Discounted Cumulative Gain (NDCG) progresses as the list of recommendations decreases. When the number of neighbor users in the second test is increased to 50, the results showed that precision is about 91%, recall still the same and the diversity becomes 8. The results also showed that the DCG does not differ from the first test and this is one of the system advantages. However, the NDCG is decreasing as the recommended list progressed. This refers, to the closeness between DCG and Ideal order (IDCG), which means that the recommendation list has a good organization. Finally, after using clustering techniques, the Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) have been measured and the results showed that the MAE has been decreasing from 0.54 to 0.39 and the RMSE is about 1.16