Recommendation Algorithm using Weighted User Behavior Score-Based Similarity

One of the most widely used recommendation system approaches is neighborhood-based collaborative filtering. This approach uses the power of similarity between users to generate recommendations. Recently, the developed similarity model considers not only explicit rating scores but also implicit rating scores. The problem of this similarity model is to use similarity weighting based on the threshold value. The error rate for rating predictions is still needed to improve for a better recommendation. This study aims to develop a similarity model that considers similarity weighting by paying attention to the number of items rated by users to increase the recommendation performance. The proposed similarity model is called Weighted user Behavior score-based Similarity (WeBSim). Our experiment used the MovieLens 100k dataset to test the model performance. The results showed that the proposed similarity model outperforms the previous similarity (UPCF) by reducing the MAE and RMSE values by 0.0148 and 0.0123.