raeeynega/HDSS-HHR-Record-Linkage

🏥 HDSS-HHR Linkage System - AI-powered record linkage for Ethiopian health data. Matches Household Registration records with Health and Demographic Surveillance System data using Microsoft MiniLM AI + Fellegi-Sunter probabilistic scoring. # 🏥 HDSS-HHR Linkage System An **intelligent record linkage system** that matches Household Registration (HHR) records with Health and Demographic Surveillance System (HDSS) data. Designed for Ethiopian health data with special handling for name variations (QIMI = KIMI, MOHAMED = MUHAMMAD). ## 📖 What is This System? The **HDSS-HHR Linkage System** helps match people's records across different health databases. It can find the same person even when names are spelled differently. ### Real-World Example - **HHR Record**: "KIMI MUSSA" from household registration - **HDSS Record**: "QIMI MUSA" from health surveillance The system understands that "KIMI" and "QIMI" are the same name! ## 🎯 Key Features | Feature | What It Does | |---------|--------------| | **Smart Name Matching** | Understands name variations like "KIMI" = "QIMI" | | **Family Linkage** | Uses father's names to connect family members | | **Location Filtering** | Searches by village (Kebele), cluster, or site | | **Household View** | Shows all family members after finding a match | | **Easy Review System** | Lets you confirm or reject potential matches | | **Works Anywhere** | No internet needed | ## 📊 Matching Accuracy | Name Pair | Without AI | With MiniLM | |-----------|------------|-------------| | KIMI / QIMI | 70% | **95%** ✓ | | MUSSA / MUSA | 85% | **92%** ✓ | | MOHAMED / MUHAMMAD | 65% | **88%** ✓ | | ABRAHAM / ABRHAM | 80% | **92%** ✓ | | TIGIST / TIGISTU | 75% | **91%** ✓ | | HAILE / HAYLE | 80% | **94%** ✓ | | **Overall Accuracy** | **75-80%** | **85-90%** | ## 🛠️ Technologies | Category | Technology | |----------|------------| | **Web Framework** | Flask 3.1.0 | | **Database** | SQLite | | **Name Matching** | RapidFuzz, Phonetics | | **AI / Semantic** | Microsoft MiniLM (Sentence-Transformers) | | **Deep Learning** | PyTorch (CPU) | | **Data Processing** | Pandas, NumPy | | **Frontend** | HTML/CSS/JavaScript, Bootstrap 5 | | **Authentication** | Flask-Login, Flask-Bcrypt | ## 🤖 How …