SentiDrop: A Multi-Modal Machine Learning model for Predicting Dropout in Distance Learning

This paper presents **SentiDrop**, a novel machine learning model designed to **predict student dropout in distance learning environments**, a significant challenge in education. The model integrates **diverse data sources** to offer a more comprehensive understanding of dropout risks, including students' **socio-demographic information**, their **behavioral data** (such as login frequency and course completion rates), and crucially, **sentiment analysis of their online comments**. By fine-tuning the **Bidirectional Encoder Representations from Transformers (BERT) model** to capture nuanced sentiments from student comments and combining this with socio-demographic and behavioral data analyzed through **Extreme Gradient Boosting (XGBoost)**, SentiDrop aims to provide early detection of at-risk students. The inclusion of sentiment data, especially early in the academic term, helps reveal emotional states that traditional metrics might miss, significantly improving the model's ability to identify students who are disengaged or dissatisfied. When tested on unseen data, the proposed model achieved an **accuracy of 84%**, demonstrating enhanced predictive performance compared to baseline models, thereby offering a **vital tool for educators to develop personalized strategies and interventions** to reduce dropout rates and encourage student perseverance.

https://arxiv.org/pdf/2507.10421

Видео SentiDrop: A Multi-Modal Machine Learning model for Predicting Dropout in Distance Learning канала AI Papers Podcast Daily