Enhancing Student Retention in Introductory Programming Courses: Leveraging Advanced Learning Validation Tools and Educational Data Mining

Student retention in introductory programming courses remains a persistent challenge in higher education, with high failure and dropout rates impacting both learners and institutions. This article presents a novel, behavior-based approach to addressing this issue through Educational Data Mining (EDM) and a custom-built learning validation framework called AssessMe. Developed by the authors in collaboration with SmoothSoft Ltd., AssessMe is an advanced software tool that monitors the real-time development of programming assignments. It captures detailed behavioral data—including active coding time, code changes (added, modified, removed lines), and submission timelines—to generate learning indicators reflecting how students approach problem-solving tasks. Unlike traditional assessment methods that focus solely on final code correctness, AssessMe emphasizes the coding process, offering deeper insights into student engagement, effort, and learning strategies. This study focuses on Programming I and II courses within the Web & Mobile Computing program at RIT Croatia. The dataset includes 3,537 student submissions from the 2024/2025 academic year, covering homework, practicals, and in-class activities, all enriched with AssessMe indicators. We apply traditional machine learning models combined with the TSFRESH library to extract meaningful time-series features from students’ coding activity. This enables the identification of temporal learning patterns and supports early prediction of academic outcomes. Our models achieve over 93% accuracy in forecasting pass/fail status by the fifth week of a 15-week semester, demonstrating a strong correlation between AssessMe indicators and final grades. This behavior-based assessment approach enhances early intervention strategies and provides actionable insights for improving student retention and learning outcomes.