Predicting the Programming Skills of Undergraduate Computer Science Students Using Machine Learning Techniques

Abstract

Computer science schools are adapting to the changing landscape of artificial intelligence and data science by developing new ways to train students to be proficient programmers. For the need mentioned before, programming is crucial. Gaining these skills can set students up for success in the IT industry and beyond. Multiple machine learning algorithms, including Logistic Regression, Decision Tree, Random Forest, Support Vector Machine (SVM) and K-Nearest Neighbour (KNN) were applied and assessed. A dataset of 113 student records was used, containing academic and behavioural features that were the most influential predictors of program competency. Experimental analysis showed that the Random Forest has achieved a high accuracy rate of 98.3%. This study proposes a machine learning prediction technique for assessing the coding ability of undergraduate students. Forecasting programming proficiency at an early stage allows the instructor to provide timely intervention and customised support to the students. Future enhancements include explainable AI techniques and graph based deep learning models to improve prediction precision and translation.