Sequential recommendation via agent-based irrelevancy skipping.

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Sequential Recommendation models user interaction sequences to predict future behavior.
• Current methods struggle with irrelevant historical data, hindering accurate next-interaction predictions.
• User interests and behaviors are diverse and uncertain, complicating recommendation accuracy.

Purpose of the Study:

• To propose a novel Sequential Recommendation method, Dynamic-Skip for Sequential Recommendation (DyS4Rec).
• To enhance recommendation accuracy by adaptively filtering irrelevant historical user interactions.
• To improve personalization in sequential recommendations.

Main Methods:

• Utilized a Long Short-Term Memory (LSTM) network with dynamic skip connections.
• Implemented a Personalized Module (PM) to guide the interaction skipping process.
• Developed an adaptive learning mechanism to exclude irrelevant historical data.

Main Results:

• DyS4Rec demonstrated superior performance over state-of-the-art methods on five real-world datasets.
• Achieved performance improvements ranging from 1% to 12% compared to existing approaches.
• Visualization analyses confirmed DyS4Rec's ability to selectively skip irrelevant interactions.

Conclusions:

• DyS4Rec effectively models long-term dependencies by dynamically skipping irrelevant interactions.
• The method enhances personalization and accuracy in sequential recommendation systems.
• DyS4Rec offers a robust solution for improving user-centric recommendations.