COC Competence Performance Evaluation for Ethiopian TVET Schools

The development of a nation's human capital is intrinsically linked to the quality of its educational systems. In Ethiopia, Technical and Vocational Education and Training (TVET) program is a crucial of this development. It is the source for the much-required middle level highly skilled oriented professionals. In this respect the Certificate of Competence (COC) assessment serves as a critical gateway for graduates to succeeds. This study explores the application of machine learning (ML) techniques to predict student performance in COC examinations. The aim of the paper is to identify key predictive factors in COC that helps enhance educational outcomes. The study utilizes a dataset of 19,680 student records obtained from Addis Ababa Occupational Competency Assessment and Certification Center. Various ML classification models employed including Random Forest, K-Nearest Neighbors (KNN), Neural Networks, and Support Vector Machines (SVM). The main activities of the study pertaining on the model are rigorous data preprocessing and feature engineering and model training and evaluation using 10-fold cross-validation. Regarding the result most models score a high result – on average scoring around 95% accuracy. The actual predication of the model involves predicting whether a student will be "Competent" or "Not Yet Competent". The feature Practical result is identified as the most significant predictor of success. This study demonstrates the robust potential of ML to transform raw educational data into actionable insights. This enables TVET institutions to implement timely interventions, optimize resource allocation, and ultimately improve student success rates, thereby contributing to Ethiopia's broader socioeconomic development goals.