Exploiting bi-directional global transition patterns and personal preferences for missing POI category identification

Area of Science:

• Computer Science
• Data Science
• Artificial Intelligence

Background:

• Location-based Social Network (LBSN) services are increasingly popular, offering opportunities for personalized Point-of-Interest (POI) recommendations.
• Existing methods for POI recommendation and location prediction primarily use unidirectional historical data.
• Identifying missing POI categories requires analyzing check-in data both before and after the missing information, posing a significant challenge.

Purpose of the Study:

• To develop a novel neural network approach for accurately identifying missing POI categories in LBSN data.
• To effectively integrate bi-directional global non-personal transition patterns with personal user preferences.
• To address the challenge of identifying POI categories in real-world mobile user check-in data.

Main Methods:

• A novel neural network architecture is proposed.
• An attention matching cell is designed to model the alignment between check-in category information and user-specific patterns.
• The model integrates bi-directional transition patterns and personal preferences for comprehensive analysis.

Main Results:

• The proposed model demonstrates significant effectiveness in identifying missing POI categories.
• Empirical evaluation on two real-world datasets shows superior performance compared to state-of-the-art baselines.
• The model's effectiveness in identifying missing POI categories is validated.

Conclusions:

• The novel neural network approach effectively identifies missing POI categories by integrating user preferences and transition patterns.
• The model offers a robust solution for a long-standing challenge in LBSN data analysis.
• The approach can be extended to enhance next POI category recommendation and prediction tasks.