Scholarlink

<p>Predicting student dropouts has always been crucial for traditional educational institutions, but it is even more so for distance learning platforms. A comprehensive approach incorporating diverse student data sources is necessary to make accurate predictions. In this paper, we introduce a new model utilizing a multi-modal fusion of sentiment analysis, performed on student comment data using the Bidirectional Encoder Representations from Transformers (BERT) model, with socio-demographic and behavioral data examined using the Extreme Gradient Boosting (XGBoost) model. Multi-modal data fusion improves the precision of student dropout prediction models, offering a deeper understanding of student dropout risks. Using the dataset obtained from the ChallengeU online platform, our model demonstrated remarkable success in identifying at-risk students, achieving an impressive 84% accuracy. Compared with the baseline model, this shows a noteworthy improvement, underlining the effectiveness of our method. What distinguishes our approach is using sentiment analysis alongside socio-demographic and behavioral data to predict school dropouts. For the first time, a dropout prediction model uses multi-modal data sources. The proposed approach could be vital in developing personalized strategies to reduce dropout rates and encourage perseverance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predicting Online Education Dropout: A new Machine Learning Model based on Sentiment Analysis, Socio-demographic, and Behavioral Data

  • 期刊论文

  • 发表日期：

    2025年4月11日

查看全文
  • Meriem Zerkouk,
  • Miloud Mihoubi,
  • Belkacem Chikhaoui,
  • Shengrui Wang

摘要

Predicting student dropouts has always been crucial for traditional educational institutions, but it is even more so for distance learning platforms. A comprehensive approach incorporating diverse student data sources is necessary to make accurate predictions. In this paper, we introduce a new model utilizing a multi-modal fusion of sentiment analysis, performed on student comment data using the Bidirectional Encoder Representations from Transformers (BERT) model, with socio-demographic and behavioral data examined using the Extreme Gradient Boosting (XGBoost) model. Multi-modal data fusion improves the precision of student dropout prediction models, offering a deeper understanding of student dropout risks. Using the dataset obtained from the ChallengeU online platform, our model demonstrated remarkable success in identifying at-risk students, achieving an impressive 84% accuracy. Compared with the baseline model, this shows a noteworthy improvement, underlining the effectiveness of our method. What distinguishes our approach is using sentiment analysis alongside socio-demographic and behavioral data to predict school dropouts. For the first time, a dropout prediction model uses multi-modal data sources. The proposed approach could be vital in developing personalized strategies to reduce dropout rates and encourage perseverance.