Multiagent-System-Based Attention Mechanism for Predicting Product Popularity: Handling Positive-Negative Diffusion...

Area of Science:

• Computer Science
• Artificial Intelligence
• Social Network Analysis

Background:

• Product popularity prediction is crucial for marketing and business strategy.
• Social networks significantly influence product diffusion dynamics.
• Existing models often struggle to capture complex user interactions and diffusion patterns.

Purpose of the Study:

• To propose a novel model for predicting product popularity on social networks considering positive-negative diffusion (PND).
• To develop an efficient feature extraction method for user representation in diffusion prediction.
• To establish a multi-agent system (MAS) model that simulates and predicts product diffusion.

Main Methods:

• A positive-negative diffusion (PND) model was developed to simulate product spread.
• A multi-agent-system-based attention mechanism (MASAM) was devised for optimal user feature vector extraction.
• A distributed learning algorithm was used to train the MASAM's shared weight matrix.
• An MAS model for product diffusion was established using MASAM feature representations.
• Agent interaction rules were suggested to accelerate simulation.

Main Results:

• The proposed PND model and MASAM effectively simulate product diffusion on social networks.
• The MASAM significantly improves the precision of user feature extraction for prediction.
• The MAS model demonstrates high effectiveness and efficiency in product popularity prediction.
• Experimental results show superior performance compared to baseline methods.
• A case study validates the algorithm's applicability and extendibility.

Conclusions:

• The developed PND and MAS models, powered by MASAM, offer a robust solution for product popularity prediction in social networks.
• The approach provides accurate and efficient predictions by effectively modeling user behavior and diffusion dynamics.
• The findings have practical implications for marketing strategies and understanding information spread.