School Feasibility Analysis and Grade Improvement Strategies Using the Random Forest Algorithm

Abstrak

Background of Study: Educational disparities across Indonesian provinces persist, particularly in infrastructure, teacher quality, and dropout rates, necessitating data-driven analysis for equitable improvements.Aims: This study investigates school feasibility and proposes strategies to enhance provincial education performance using the Random Forest algorithm.Methods: Aggregated provincial education data covering student numbers, dropout rates, teacher qualifications, and classroom conditions were transformed into derivative indicators. A binary classification (Feasible/Not Feasible) based on national dropout median was applied. The model was developed using R with six systematic steps, including training and evaluation of a Random Forest model (ntree = 100, mtry = 3) using accuracy, sensitivity, and specificity.Result: The model accurately classified school feasibility. Key predictors included teacher quality, student-teacher ratios, and classroom conditions. Several provinces were identified as “Not Feasible.”Conclusion: Machine learning proves effective for education policy support. The study offers targeted recommendations such as improving infrastructure, enhancing teacher training, and reducing dropouts to promote equitable education in Indonesia.