Research on a Multi-Source Data Fusion–Based Early-Warning System for Online Learner Dropout Risk
Year: 2026
Type: conference-paper
Abstract: Online learning platforms have broadened educational access and expanded the supply of learning resources, yet persistent dropout remains prevalent due to highly volatile engagement patterns and heterogeneous learning trajectories. In practice, existing early-warning solutions are often constrained by the difficulty of unifying heterogeneous multi-source data, sparse and missing temporal behaviors, delayed labels, and incomplete feedback loops, which collectively undermine score stability and hinder timely intervention. This paper presents a deployable dropout-risk early-warning system built upon a layered microservice architecture spanning data ingestion, feature processing, fusion-based modeling, and warning services. Windowed features are constructed through event-stream preprocessing and time alignment; layered fusion is performed at both the feature and representation levels with learned modality reliability weights; evidence fusion is further used to output both a risk score and a conflict degree. Finally, tiered warning rules and a unified explanation format provide traceable evidence to support closed-loop refinement.
Authors: Cai Li, Zhuochen Zhu
Institution: Wuhan Huaxia University of Technology
Language: English
Subfield: Computer Science Applications
Field: Computer Science
Domain: Physical Sciences
SDG: Quality education
Open Access status: hybrid
OpenAlex ID: W7203660078