Deep Attentive Study Session Dropout Prediction in Mobile Learning Environment (2002.11624v5)

Abstract: Student dropout prediction provides an opportunity to improve student engagement, which maximizes the overall effectiveness of learning experiences. However, researches on student dropout were mainly conducted on school dropout or course dropout, and study session dropout in a mobile learning environment has not been considered thoroughly. In this paper, we investigate the study session dropout prediction problem in a mobile learning environment. First, we define the concept of the study session, study session dropout and study session dropout prediction task in a mobile learning environment. Based on the definitions, we propose a novel Transformer based model for predicting study session dropout, DAS: Deep Attentive Study Session Dropout Prediction in Mobile Learning Environment. DAS has an encoder-decoder structure which is composed of stacked multi-head attention and point-wise feed-forward networks. The deep attentive computations in DAS are capable of capturing complex relations among dynamic student interactions. To the best of our knowledge, this is the first attempt to investigate study session dropout in a mobile learning environment. Empirical evaluations on a large-scale dataset show that DAS achieves the best performance with a significant improvement in area under the receiver operating characteristic curve compared to baseline models.

• The paper introduces the DAS model that leverages a Transformer architecture to predict mobile study session dropouts.
• The methodology employs an encoder-decoder framework with sequential masking to ensure real-time, causally valid predictions.
• Empirical results on a large-scale dataset demonstrate that DAS significantly outperforms baselines, notably improving ROC AUC scores.

Deep Attentive Study Session Dropout Prediction in Mobile Learning Environments

Introduction

The advent of mobile learning platforms has introduced new dynamics into the education sector, enabling personalized learning experiences anytime and anywhere. Despite the numerous benefits, mobile learning is susceptible to higher dropout rates due to distractions and the inherent nature of mobile interfaces. Addressing paper session dropout in mobile learning environments is crucial for enhancing student engagement and optimizing learning outcomes. This paper presents a novel approach to predicting paper session dropouts in mobile learning environments using a Transformer-based model, named DAS (Deep Attentive Study Session Dropout Prediction). The research uniquely focuses on the dropout prediction problem within the context of mobile learning, distinguishing itself from existing studies that primarily concentrate on school or course dropout predictions.

Study Session Dropout Prediction Problem

The concept of paper session dropout is meticulously defined, considering the unique characteristics of mobile learning environments. A paper session is identified by a sequence of learning activities with less than an hour's gap between consecutive actions. A dropout is marked if a student is inactive for an hour, indicating the end of a session. This operational definition lays the foundation for formulating the dropout prediction problem, aiming to compute the probability of a student dropping out after engaging with a particular learning activity.

DAS Model Architecture

The proposed DAS model leverages the Transformer architecture's power, incorporating an encoder-decoder structure tailored for the dropout prediction task. The model processes sequences of question-response pairs, extracting complex relations among student interactions through deep attentive computations. Notably, DAS applies subsequent masks across all multi-head attention layers, ensuring predictions are strictly based on previously observed interactions without peering into future activities. This approach addresses the causality concern, making the model's predictions valid and actionable in real-time learning scenarios.

Empirical Evaluation

Extensive evaluations were conducted using a large-scale dataset from an active mobile education application named Santa. The results indicate that DAS significantly outperforms baseline models, achieving notable improvements in the area under the ROC curve (AUC). This empirical validation not only confirms DAS's effectiveness in predicting paper session dropouts but also demonstrates the model's practical applicability in enhancing mobile learning platforms.

Implications and Future Directions

The implications of this research are twofold. Practically, DAS provides a robust tool for mobile learning platforms to dynamically adjust learning paths, aiming to reduce dropout rates and foster sustained engagement. Theoretically, this work opens new avenues for exploring dropout prediction in varied learning contexts, encouraging further developments in AI-driven educational interventions. Future research could delve into integrating DAS with recommendation systems for personalized learning experiences or extending the model to predict long-term dropouts for comprehensive student retention strategies.

Conclusion

This paper introduces a groundbreaking approach to tackling the challenge of paper session dropout in mobile learning environments. By creatively applying the Transformer model, DAS sets a new standard for predictive analytics in education, offering valuable insights for educators, platform developers, and AI researchers alike. As mobile learning continues to evolve, research endeavors like this will play a pivotal role in shaping the future of education, making learning more adaptive, engaging, and effective.