HAM: Hybrid Associations Models for Sequential Recommendation.

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Sequential recommendation systems predict user preferences based on historical interactions.
• Existing methods often struggle to capture complex user behaviors and item relationships.
• Effective recommendation requires understanding long-term preferences and short-term sequential patterns.

Purpose of the Study:

• To develop Hybrid Associations Models (HAM) for enhanced sequential recommendation.
• To incorporate users' long-term preferences, sequential association patterns, and item synergies.
• To evaluate HAM's performance against state-of-the-art methods.

Main Methods:

• Developed Hybrid Associations Models (HAM) integrating user preferences and item associations.
• Utilized simplistic pooling for item set representation and element-wise product for item synergies.
• Compared HAM with existing methods on six benchmark datasets across three experimental settings.

Main Results:

• HAM models demonstrated significant performance improvements, achieving up to 46.6% enhancement over state-of-the-art methods.
• Experimental results confirmed HAM's superiority across all tested settings.
• HAM exhibited superior runtime efficiency, with speedups reaching up to 139.7 folds.

Conclusions:

• Hybrid Associations Models (HAM) represent a significant advancement in sequential recommendation.
• HAM effectively captures complex user-item dynamics, leading to superior prediction accuracy.
• The efficiency of HAM makes it a practical solution for real-world recommendation systems.