Efficient clustering in collaborative filtering recommender system: Hybrid method based on genetic algorithm and...

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Recommender Systems (RS) aim to predict user preferences accurately.
• Data clustering is a key technique to improve RS performance.
• Existing methods may not fully optimize item clustering for collaborative filtering.

Purpose of the Study:

• To propose a single-objective hybrid evolutionary approach for item clustering in offline collaborative filtering recommender systems.
• To enhance the accuracy of recommendations by improving data clustering.

Main Methods:

• A hybrid meta-heuristic algorithm combining Genetic Algorithm (GA) and Gravitational Emulation Local Search (GELS).
• The algorithm iteratively improves a population of randomized solutions for item clustering.
• Application to offline collaborative filtering recommender systems.

Main Results:

• The proposed algorithm achieves more appropriate data clustering compared to existing methods.
• Improvements were observed in key performance metrics: Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Coverage.
• The method requires a relatively high runtime but yields superior clustering results.

Conclusions:

• Hybrid evolutionary algorithms, like the proposed GA-GELS method, can effectively improve item clustering in recommender systems.
• Enhanced data clustering directly translates to more accurate user recommendations.
• The trade-off between runtime and improved accuracy metrics (MAE, RMSE, Coverage) is a key consideration.