Luyipai Asset & User Matching System

Project Background:

• Developed a precise asset-to-investor matching system for Luyipai, a special asset trading platform under Ping An.
• Objectives: Improve user satisfaction, asset transaction conversion, and personalized recommendation.

Data Collection & Processing:

• Asset data: Property, mortgage, and financial assets; processed for static and dynamic features.
• User data: Behavioral logs, registration info, and preference settings.
• Data cleaning and tagging for both users and assets.

Modeling & Optimization:

• User profiling: Long-term and short-term profile construction.
• Multi-channel asset recall:
  • Single-dimension label-based retrieval.
  • Similar assets retrieval.
  • Collaborative user-based recall.
• Ranking: Linear weighting model, tag relevance score.
• Workflow engineering: Redis for short-term profiles, MySQL/Redis for long-term, Elasticsearch for asset storage.

Project Workflow:

1. Construct asset and user profiles with feature scoring.
2. Multi-channel asset recall and top-K selection.
3. Merge and re-rank recalled assets.
4. Provide recommendations and generate detailed asset analysis reports.

Project Outcome:

• Built a complete matching system from scratch, significantly improving asset exposure and transaction success rates.
• Detailed reports facilitated user decision-making and increased revenue.