KT-Devv/student-dropout-prediction-ghana

# Student Dropout Prediction in Ghanaian Basic Schools Using Machine Learning **KNUST – Department of Computer Science** **2025–2026 Academic Year** **Group 5** A comprehensive machine learning framework for the early prediction of student dropout in Ghanaian basic schools using institutional records, baseline machine learning models, class imbalance handling (SMOTE and CTGAN), hyperparameter optimization, and explainable artificial intelligence (SHAP). --- # Project Overview Student dropout remains a significant challenge affecting educational outcomes in Ghana. Early identification of students at risk enables timely interventions by teachers, school administrators, and policymakers. This project develops and evaluates multiple machine learning models to predict student dropout using demographic, academic, attendance, behavioural, and socioeconomic data collected from Ghanaian basic schools. The study follows a reproducible end-to-end machine learning pipeline, beginning with data cleaning and exploratory analysis through model development, optimization, explainability, and comparative evaluation. --- # Research Objectives The project aims to: - Develop machine learning models for early dropout prediction. - Compare the performance of multiple classification algorithms. - Investigate the effect of class imbalance handling using SMOTE and CTGAN. - Optimize the best-performing model through hyperparameter tuning. - Explain model predictions using SHAP. - Produce a reproducible machine learning workflow suitable for educational research. --- # Machine Learning Pipeline ``` Raw Data │ ▼ Data Cleaning │ ▼ Exploratory Data Analysis │ ▼ Feature Engineering │ ▼ Baseline Models │ ▼ SMOTE vs CTGAN Experiments │ ▼ Hyperparameter Optimization │ ▼ Final Model │ ▼ SHAP Explainability │ ▼ Model Comparison & Evaluation ``` --- # Machine Learning Models The following supervised learning algorithms are evaluated: - Logistic Regression - Decision Tree - Random Fores …