RecMem: Time Aware Recommender Systems Based on Memetic Evolutionary Clustering Algorithm.

Area of Science:

• Computer Science
• Artificial Intelligence
• Data Mining

Background:

• Recommender systems are crucial for navigating large, dynamic item spaces.
• Existing systems struggle with evolving user preferences and temporal dynamics.
• User behavior is influenced by social network effects and temporal factors.

Purpose of the Study:

• To develop a time-aware recommender system that addresses the limitations of static clustering in dynamic environments.
• To enhance recommendation accuracy by evolving clusters over time.
• To improve handling of cold-start problems for both users and items.

Main Methods:

• A memetic evolutionary clustering algorithm (RecMem) was developed for time-aware recommendations.
• Clusters were evolved dynamically at each timestamp using a memetic evolutionary algorithm.
• The memetic algorithm was enhanced with a chaos criterion for improved optimization.
• Item attributes and demographic information were used to address cold-start issues.

Main Results:

• The RecMem system demonstrated high accuracy in providing timely and relevant recommendations.
• The proposed method achieved an approximate accuracy of 0.95.
• RecMem outperformed existing recommender systems in terms of accuracy and adaptability.

Conclusions:

• Time-aware recommender systems based on evolutionary clustering offer significant improvements over traditional methods.
• RecMem effectively handles dynamic user preferences and evolving data.
• The approach provides a robust solution for cold-start scenarios in recommendation engines.