Mutsa30/gdm-risk-chatbot

AI-powered system for early prediction of Gestational Diabetes Mellitus (GDM) using machine learning and WhatsApp chatbot integration. Developed to support maternal health in low-resource settings. Machine Learning-Driven Prediction of Gestational Diabetes Mellitus (GDM) with WhatsApp Chatbot Integration ## Project Overview This project uses machine learning to predict the risk of **Gestational Diabetes Mellitus (GDM)** among pregnant women and integrates the predictive model into a **WhatsApp chatbot** for early risk detection in low-resource settings such as Zimbabwe. The chatbot allows users to input health data (like BMI, blood pressure, and family history) and instantly receive a GDM risk score generated by a trained machine learning model. ## Objectives - Develop a predictive model for early GDM detection using non-invasive features. - Identify key maternal risk factors contributing to GDM. - Integrate the model into a WhatsApp chatbot for real-time risk assessment. ## Machine Learning Approach Dataset:Kaggle — 3,525 records of pregnant women with 17 clinical and demographic features. Algorithms Tested: Random Forest, Logistic Regression, Decision Tree, Support Vector Machine, XGBoost | Model | Accuracy | Precision | Recall | F1-score | |--------|-----------|-----------|---------|-----------| | Random Forest | 92% | 91% | 90% | 90% | | Logistic Regression | 87% | 85% | 86% | 85% | | Decision Tree | 88% | 86% | 88% | 87% | | SVM | 90% | 91% | 89% | 90% | | XGBoost (Best) | 96% | 94% | 95% | 94% | --- ## Technologies Used - Programming: Python - Libraries:Scikit-Learn, Pandas, NumPy, Matplotlib, Seaborn, XGBoost - Frameworks: FastAPI (for model deployment) - Database:SQLite - Integration:WhatsApp API - Environment: Google Colab, Visual Studio Code --- ## 💬 WhatsApp Chatbot Integration - Built a FastAPI backend for model prediction endpoints. - Integrated with WhatsApp API to collect user input and return GDM risk predictions in real time. - Used SQLite databaseto log user data and model outcomes for monitoring. ## Results - The XGBoost model achieved 96% accuracy. - The chatbot successfully provided real-time, accurate risk assessments duri …