Modeling user rating preference behavior to improve the performance of the collaborative filtering based recommender...

Area of Science:

• Computer Science
• Information Retrieval
• Artificial Intelligence

Background:

• Online shopping websites require efficient, customized recommendations for numerous users.
• Collaborative filtering (CF) is a primary method for personalized recommendations, relying on user-item rating matrices.
• Traditional similarity measures like Pearson correlation coefficient (PCC) often overlook user rating preference behavior (RPB).

Purpose of the Study:

• To develop a novel similarity measure that incorporates user rating preference behavior (RPB) into collaborative filtering (CF).
• To address the limitations of standard PCC by accounting for individual user rating patterns.

Main Methods:

• A new similarity measure, improved PCC weighted with RPB (IPWR), was developed.
• IPWR models user RPB using average rating, variance, or standard deviation.
• The proposed method combines RPB with an enhanced PCC model.

Main Results:

• IPWR demonstrated superior performance compared to existing state-of-the-art similarity measures.
• Improvements were observed across key metrics including Mean Absolute Error (MAE), Root Mean Square Error (RMSE), precision, recall, and F-measure.
• The method was validated on five diverse datasets: Epinions, MovieLens-100K, MovieLens-1M, CiaoDVD, and MovieTweetings.

Conclusions:

• The proposed IPWR similarity measure effectively enhances CF-based recommender systems.
• Incorporating user rating preference behavior (RPB) leads to more accurate and personalized recommendations.
• IPWR offers a significant advancement over traditional similarity measures in recommendation systems.