Implementing Maysat method multi-model ensemble - Phase 1:
- Created InsuranceTextProcessor class with transformer pipeline
- Extracts linguistic features: urgency, emotional, vague language
- Analyzes claim text for fraud indicators
- Returns fraud score (0-1) with confidence level
- Includes test cases for validation
Next: Integrate with fraud_engine.py for hybrid scoring (40% Transformer + 40% ML + 20% Rules)
Implementing Maysat method for multi-model ensemble:
- Transformers for NLP-based fraud detection
- PyTorch for deep learning models
- Scikit-learn for traditional ML (Random Forest, XGBoost)
- Sentence-transformers for text embeddings
This enables hybrid approach: 40% Transformer + 40% ML + 20% Rules
