Abstract:
This paper used machine learning in assessing the mitigates of primary school dropout risk in Narok West Sub- County, Kenya. Although the use of bursaries, school feeding and community sensitization has been long held, current interventions are reactive meaning that they deal with dropouts once they stop attending school. The predictive modeling to predict dropout and inform preventative action developed using structured field survey (n= 1,000) with Monte Carlo simulation extending to 10,000 records. Three classifiers, namely, Random Forest, XGBoost, and Support Vector Machine were trained on an 80/20 split with five-fold cross-validation and measured in terms of accuracy, precision, recall, F1-score, and ROC-AUC. XGBoost has obtained the best results (AUC = 0.804; F1 = 0.771), which makes it the model that has been validated. The findings of Chapter Four have revealed that financial considerations prevailed in risk dynamics: bursary receipt and bursary amount had a significant negative effect on dropout whereas monthly fees donations and traveling a long distance contributed to the level of dropout. The welfare programs like school feeding, meals per day and community participation were identified as the important protective factors. To make sure the results could be interpreted, explainable AI methods such as permutation importance, SHAP values, and partial dependence plots revealed both the importance and direction of the influence of every factor, not just in prediction but actionable insights. Its results show that financial strain mediated by structural and social supports is the leading cause of dropout, and that predictive analytics can offer policy-makers evidence-based drops to intervene. The integration of such models into the education planning provides a proactive channel of maintaining learner retention and enhancing equity in the rural schooling. The evaluation work helps construct an education system that stands resilient along with technology development and social fairness in Narok West Sub County.
