Odwa-Yamile/student-dropout-early-warning-system

An explainable machine learning system for predicting student dropout risk at South African universities using educational data and AI-driven early warning indicators. # 🎓 Student Dropout Early Warning System > An AI-powered Early Warning System that predicts student dropout risk using Machine Learning, Explainable AI (SHAP), and an interactive Streamlit dashboard. # 📌 Project Highlights * Built an end-to-end Machine Learning pipeline for student dropout prediction. * Compared Logistic Regression and Random Forest classifiers. * Achieved an ROC-AUC score of **92.7%**. * Applied Explainable AI using SHAP. * Developed an interactive Streamlit dashboard for real-time risk assessment. * Implemented a Train / Validation / Test split strategy (50% / 25% / 25%). * Identified key academic and financial factors associated with student dropout. --- # 📑 Table of Contents - 📌 Executive Summary - 🎯 Business Problem - 🎯 Project Objectives - 📚 Dataset Overview - 🛠️ Tools & Technologies - 📂 Repository Structure - 📈 Model Evaluation - 🧠 Explainable AI (SHAP) - 🖥️ Interactive Dashboard - 🔑 Key Findings - 🚀 How to Run the Project - 🔮 Future Improvements - 👨‍💻 Author --- # 🎓 Student Dropout Early Warning System ## 📌 Executive Summary Student dropout remains one of the most significant challenges facing higher education institutions. Students who discontinue their studies affect graduation rates, institutional performance, and long-term educational outcomes. This project develops an AI-powered Early Warning System capable of identifying students who may be at risk of dropping out using demographic, academic, financial, and economic information. The solution combines: * Machine Learning * Explainable AI (SHAP) * Interactive Analytics * Streamlit Dashboarding The final model achieved strong predictive performance and was deployed through an interactive dashboard that allows stakeholders to assess student dropout risk in real time. --- # 🎯 Business Problem Universities often struggle to identify at-risk students before it becomes too late for effective intervention. Traditional approaches are usually reactive and rely heavily on ma …