UsCoTc: Improved Collaborative Filtering (CFL) recommendation methodology using user confidence, time context with...

Area of Science:

• Information Science
• Computer Science
• Bibliometrics

Background:

• Researchers face challenges in efficiently identifying relevant scientific papers due to time constraints.
• Existing personalized suggestion systems using collaborative filtering struggle with sparse data and dynamic user interests.

Purpose of the Study:

• To develop an improved similarity measure for personalized recommendation systems.
• To enhance the accuracy and quality of research paper suggestions by incorporating user confidence and temporal dynamics.

Main Methods:

• Proposed a novel similarity measure that integrates user confidence and time context into collaborative filtering.
• Evaluated the approach using sparse rating data to assess its effectiveness in user similarity computation.

Main Results:

• The new similarity measure demonstrated robust performance with sparse data.
• Achieved a 16.2% improvement in prediction accuracy compared to existing models.
• Enhanced the overall quality of recommended research articles.

Conclusions:

• The proposed method effectively addresses limitations of traditional collaborative filtering in recommendation systems.
• Incorporating user confidence and time context significantly improves recommendation accuracy and relevance.
• Offers a more efficient solution for researchers seeking timely access to pertinent literature.