Hybrid recommender system model for digital library from multiple online publishers

Area of Science:

• Digital Libraries
• Recommender Systems
• Information Science

Background:

• Growing demand for online education platforms and digital library systems.
• Existing digital libraries lack data integration across different agencies.
• Limited connectivity hinders comprehensive e-book access and discovery.

Purpose of the Study:

• To develop a hybrid recommender system for digital libraries.
• To integrate diverse knowledge sources from multiple publishers and institutions.
• To improve e-book recommendation accuracy and user experience.

Main Methods:

• Developed a prototype digital library system using API-based linking.
• Integrated e-books from educational, governmental, and religious organizations.
• Implemented a hybrid recommender system combining Collaborative Filtering (CF) and Content-Based Filtering (CB).
• Considered book category, user reading habits, and information sources.

Main Results:

• Compared hybrid models (50:50, 20:80, 80:20) with CF and CB scores.
• The Hybrid Score 80:20 model achieved the highest average Normalized Discounted Cumulative Gain (NDCG) score.
• Evaluated system performance through user feedback and comparative analysis.

Conclusions:

• A hybrid recommender system model combining 80% Collaborative Filtering and 20% Content-Based Filtering improves recommendation efficiency.
• This approach offers superior referral efficiency and overall performance compared to traditional methods.
• The developed system enhances digital library functionality for users.