Comprehensive Empirical Evaluation of Deep Learning Approaches for Session-Based Recommendation in E-Commerce.

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• E-commerce success hinges on effective recommendation systems to match user interests quickly.
• Session-based recommendation systems are gaining traction due to privacy concerns and data regulations.
• These systems predict user preferences based on in-session behavior sequences.

Purpose of the Study:

• To comprehensively evaluate state-of-the-art deep learning approaches for session-based recommendation.
• To compare deep learning methods against traditional baseline techniques.
• To identify challenges and future research directions in the field.

Main Methods:

• Investigated baseline techniques: nearest neighbors and pattern mining.
• Evaluated deep learning approaches: recurrent neural networks, graph neural networks, and attention-based networks.
• Conducted extensive experiments to compare model performance across various scenarios.

Main Results:

• Advanced neural networks and session-based nearest neighbor algorithms generally outperform baseline methods.
• Deep learning models exhibit limitations with long sessions, user interest drift, and insufficient training data.
• Hybrid models combining different approaches show potential for improved results.

Conclusions:

• Deep learning offers powerful tools for session-based recommendations, but challenges remain.
• Hybrid models tailored to dataset characteristics are a promising direction.
• Further research is needed to address the limitations of current algorithms, especially for complex user behaviors.