A novel neighbor selection scheme based on dynamic evaluation towards recommender systems.

Area of Science:

• Computer Science
• Information Retrieval
• Machine Learning

Background:

• Collaborative filtering (CF) is a key recommendation technique.
• Existing CF methods struggle with dynamic user preferences and data sparsity.
• Evaluating recommendation effectiveness in CF remains a challenge.

Purpose of the Study:

• To propose a novel neighbor selection scheme for collaborative filtering.
• To address limitations in dynamic user preference modeling and recommendation evaluation.
• To enhance recommendation performance in sparse data environments.

Main Methods:

• Introduced the concept of preference decay period to model user preference evolution.
• Defined dynamic decay factors to attenuate the influence of historical data.
• Developed three dynamic evaluation modules for user trustworthiness and recommendation ability.
• Implemented a hybrid selection strategy with two neighbor selection layers and adjustable thresholds.

Main Results:

• The proposed scheme effectively selects capable and trustworthy neighbors.
• Experimental results on three real datasets demonstrate superior recommendation performance.
• The method shows significant improvements over state-of-the-art techniques, especially with data sparsity.

Conclusions:

• The novel neighbor selection scheme enhances collaborative filtering by dynamically adapting to user preferences.
• The approach offers improved recommendation accuracy and effectiveness in real-world applications.
• This method provides a robust solution for challenges posed by data sparsity and preference dynamics.