Haoliang Sheng

Pass Rate & Overdue Rate Control

• Monitored anomalies in business reports and proposed improvements.
• Identified fraud patterns and formulated strategy rules.
• Managed projects in e-commerce, loans, and phone leasing.

Segmentation

• Classified customers based on post-loan behavior and models.
• Designed risk strategies tailored to customer segments.

Third-Party Data Validation

• Evaluated data coverage, hit rates, and lift performance.
• Selected key features based on business logic.

Post-Loan Behavior Modeling

• Conducted feature engineering and selection (PSI, KS, IV, lift).
• Handled imbalanced data (random sampling, SMOTE).
• Trained and validated models (decision trees, scorecards, ROC, OOT).

Address Standardization

• Developed modules for text entity recognition (CRF), classification (fastText), and data cleaning (regex).
• Built address completion and error correction tools using a five-level address parser.
• Implemented a retrieval module using trie structures.

Address Value Assessment

• Designed CRF-based entity recognition and data cleaning modules for address-related data.
• Derived 92 features (e.g., POI, geographic levels) and built a retrieval module (tantivy).

Address Anti-Fraud

• Designed the technical framework and constructed a POI database.
• Built modules for text inspection, mining (error correction, classification, NER), and scoring (LightGBM).

Feature Engineering Project

• Processed billion-scale data with distributed computing and shell scripts.
• Detected data coverage, distribution, and PSI.
• Derived features (min, max, median, quantiles, variance, time windows) in batches.