Predictive Model for University Course Selection in Kenya.

Course selection for higher learning institutions is crucial yet challenging, often leading to a mismatch between what learners have chosen and their interests and abilities. This project aims to use historical data to create a predictive model to assist high school students in Kenya in selecting the right courses for high-level education at the university. This study assumes that students have different intelligence capabilities ([Howard Gardner Theory](verywellmind.com)) and have taken part in subject assessments before. Therefore, course selection will be affected by their intelligence level and subject assessment scores. After reviewing the existing systems, the researcher realized that: The systems have limited specializations/ study paths. The studies have represented the models as black boxes (there is no clear understanding to humans). Using a wide variety of specializations, explainable AI (XAI) techniques (LIME and SHAP) are employed for better transparency of the best-performing model, in this research study. It is noted that all the features have an impact, and that means learners should focus on improving their intelligence and performance in particular subjects. While modifying the input values, a shift in the model’s confidence in predicting a specific course is observed.