Early prediction of student performance in CS1 programming courses.

Area of Science:

• Computer Science Education
• Machine Learning Applications

Background:

• Programming courses often exhibit high student failure rates and low academic performance.
• Early identification of at-risk students is essential for timely academic intervention and support.

Purpose of the Study:

• To develop and evaluate a machine learning model for predicting student performance in an introductory programming course (CS1).
• To identify key predictive features and optimal algorithms for early performance forecasting.

Main Methods:

• Utilized student data including grades, lab/exam delivery times, and number of attempts.
• Employed eight distinct classification algorithms for model training and evaluation.
• Assessed model performance using accuracy, recall, F1 score, and Area Under the Curve (AUC).

Main Results:

• The Gradient Boosting Classifier (GBC) achieved the highest F1 score of 86% in week 3.
• The Random Forest Classifier (RFC) demonstrated strong performance with an 83% F1 score in week 3.
• The model showed significant potential for accurate early prediction of student success.

Conclusions:

• Machine learning models can effectively predict student performance in programming courses.
• Early prediction enables targeted interventions to improve student learning outcomes.
• Key features like grades, delivery time, and attempts are valuable predictors.