SiyasangaMudau/education-inequality-ml-project

Machine learning project predicting at-risk South African schools using socio-economic and educational data. Built with Python, Pandas and Scikit-learn, achieving 89% accuracy using Random Forest classification. # Education Inequality Risk Prediction in South African Schools ## Overview This project uses Machine Learning and predictive analytics to identify South African schools that may be at risk of underperformance based on educational, socio-economic, and infrastructure-related factors. The goal is to support data-driven decision-making by helping stakeholders identify vulnerable schools and prioritise interventions. The project combines educational and socio-economic data to develop predictive models capable of classifying school performance risk levels while generating actionable insights for policymakers and educational institutions. ## Business Problem Educational inequality remains one of South Africa's most significant challenges. Factors such as funding, infrastructure quality, teacher availability, class sizes, and socio-economic conditions often influence learner outcomes. This project aims to answer the following questions: * Which factors have the strongest impact on school performance? * Can Machine Learning accurately identify schools at risk of underperformance? * How can predictive analytics support education policy and resource allocation? ## Methodology ### Data Collection Data was sourced from publicly available educational and socio-economic datasets, including: * Statistics South Africa (Stats SA) * Department of Basic Education reports * School performance indicators * Infrastructure and resource metrics ### Data Preparation * Data cleaning and preprocessing * Missing value treatment * Feature selection and engineering * Data transformation and standardisation ### Exploratory Data Analysis * School performance trend analysis * Socio-economic factor analysis * Infrastructure and resource evaluation * Correlation analysis ### Machine Learning Models Two classification models were developed and evaluated: 1. Logistic Regression 2. Random Forest Classifier ### Model Evaluation Performance was assessed using: * Accuracy * Precision * …