Understanding Self-Regulated Learning Behavior Among High and Low Dropout Risk Students During CS1: Combining Trace Logs, Dropout Prediction and Self-Reports

TL;DR

This study analyzes behavioral patterns of CS1 students at different dropout risks using learning analytics, revealing distinct strategies and behaviors that can inform targeted interventions to reduce dropout rates.

Contribution

It combines trace logs, dropout prediction, and self-reports to identify specific learning strategies and behaviors associated with dropout risk in CS1 students.

Findings

Low dropout students show 3 distinct learning strategies.

High dropout students exhibit 9 diverse strategies, including temporary unsuccessful ones.

Behavior profiling combined with analytics can inform targeted dropout interventions.

Abstract

The introductory programming course (CS1) at the university level is often perceived as particularly challenging, contributing to high dropout rates among Computer Science students. Identifying when and how students encounter difficulties in this course is critical for providing targeted support. This study explores the behavioral patterns of CS1 students at varying dropout risks using self-regulated learning (SRL) as the theoretical framework. Using learning analytics, we analyzed trace logs and task performance data from a virtual learning environment to map resource usage patterns and used student dropout prediction to distinguish between low and high dropout risk behaviors. Data from 47 consenting students were used to carry out the analysis. Additionally, self-report questionnaires from 29 participants enriched the interpretation of observed patterns. The findings reveal distinct weekly learning strategy types and categorize course behavior. Among low dropout risk students, three learning strategies were identified that differed in how students prioritized completing tasks and reading course materials. High dropout risk students exhibited nine different strategies, some representing temporary unsuccessful strategies that can be recovered from, while others indicated behaviors of students on the verge of dropping out. This study highlights the value of combining student behavior profiling with predictive learning analytics to explain dropout predictions and devise targeted interventions. Practical findings of the study can in turn be used to help teachers, teaching assistants and other practitioners to better recognize and address students at the verge of dropping out.