Learning analytics and predictive modeling in ESL education: Early detection of language acquisition barriers using big data
Year: 2026
Type: article
Abstract: The digital transformation of ESL education generates vast amounts of learner data, offering unprecedented opportunities for identifying language acquisition barriers before they impede progress.Despite the potential of big data and predictive analytics, educational practice often remains reactive, intervening only after learners have already fallen behind.Proactive, data-informed early warning systems are critically needed.This conceptual paper synthesizes interdisciplinary insights from learning analytics, educational data mining, second language acquisition theory, and predictive modeling.It proposes a multi-source framework integrating trace data from digital learning platforms, linguistic performance indicators, and contextual learner variables.Methodologically, the framework leverages machine learning classification (e.g., random forests, gradient boosting) and natural language processing to detect subtle signals of emerging difficulties.The proposed framework enables early identification of at-risk learners by analyzing engagement patterns, response times, error trajectories, and socio-affective factors.Predictive models achieve adequate accuracy in forecasting vocabulary stagnation, grammatical fossilization, and declining participation.Text-analytic techniques, including word embeddings, help uncover lexical and syntactic barriers.Integrating social network metrics further reveals interaction-based challenges in collaborative settings.The paper also discusses practical thresholds for actionable alerts and ethical safeguards for student data.
Source: Гуманитарный научный журнал
Language: English
Subfield: Computer Science Applications
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7223531636