A Decision-Tree Based Model to Identify Student Drop-Out Predictors in an Online Course: A Case Study
Bildiriler > Tam Metin Bildiri · 2025
Yazarlar: İnanç KABASAKAL
The 14th InTraders International Conference On Multidisciplinary Studies, Türkiye, Sakarya, Uluslararası, Sözlü Sunum, Tam Metin Bildiri, 2025-2025, ss 125-134
DOI: 10.66793/intraders14proceeding
Özet
<p>Predicting dropout students is one of the most popular problems in educational data mining. Identifying which students are likely to quit is useful in order to take action to prevent dropouts. Exploratory data mining identifies patterns that can be used in predicting such cases, and those patterns might involve valuable inputs for root cause analysis. This study presents a decision-tree-based short-term dropout prediction model for a speed-reading course, which is organized as 28 stages, each involving two or three reading activities and subsequent questions for assessment. The objective of the study is to predict short-term dropouts, which corresponds to the end of the next stage determined by the current progress for each participant. The dataset used in model training involves reading speeds and comprehension scores achieved throughout the most recent 10 stages completed by 244 participants. The CART Algorithm was used to train decision trees that are evaluated using k-fold cross-validation (k=5) in 100 iterations. Recall and precision scores were 55.18% and 68.55%, respectively, on average. In addition, the model with the highest performance (recall: 83.33%, precision: 100%) is presented in the findings. The study demonstrates that decision-tree-based induction is useful in predicting short-term dropouts in our case.</p>