An Incremental Knowledge-Graph Learner Digital Twin for Evidence-Grounded Early Warning

Ngoc Luyen Le; Marie-Hélène Abel; Bertrand Laforge

cs.AIOct 27, 2026

An Incremental Knowledge-Graph Learner Digital Twin for Evidence-Grounded Early Warning

Authors: Ngoc Luyen Le, Marie-Hélène Abel, Bertrand Laforge

Organizations: Gamaizer, 93340 Le Raincy, France · Université de technologie de Compiègne, CNRS, Heudiasyc (Heuristics and Diagnosis of Complex Systems), CS 60319, 60203 Compiègne Cedex, France · Sorbonne Université, CNRS UMR 7585, LPNHE (Laboratoire de Physique Nucléaire et des Hautes Énergies), 75252 Paris Cedex 05, France

Abstract

Early-warning models in learning analytics can identify learners at risk of failure or withdrawal, but the resulting alerts are often only weakly connected to inspectable learning evidence. A risk score may indicate that intervention is needed without showing which traces, assessment events, or temporal patterns justify the alert. This paper introduces an incremental Learner Digital Twin for evidence-grounded early warning, implemented as a lightweight knowledge graph. The twin is updated at temporal checkpoints by replaying learning traces and assessment records, and it serves both as a feature-extraction substrate and as an explanation-grounding substrate. Predictive features are derived through graph-mediated queries, while features, attributions, alerts, evidence items, and raw records are represented as queryable graph entities. The framework is instantiated and evaluated on a public learning-analytics benchmark under a temporal protocol. The results show that graph-mediated feature extraction preserves the predictive behavior of equivalent raw-table features across model families and checkpoints. Generated alerts are materialized as evidence chains linking risk decisions to feature attributions, evidence items, and raw platform records. The study shows that early-warning prediction can be routed through a semantic and provenance-aware representation while preserving predictive performance and adding traceability, verifiability, and evidence-grounded explanation.

Figures & tables

Figure 1: Generic workflow of the incremental ldt for evidence-grounded early warning.

Figure 2: Core ldt ontology: observation layer, analytic layer, evidence links.

Element	Setting
Dataset	OULAD; six eligible modules after preprocessing
Train/test	2013 presentations for training; 2014 presentations for testing
Checkpoints	Day 28 (Week 4), day 56 (Week 8), day 84 (Week 12)
Target	Fail/Withdrawn = at-risk; Pass/Distinction = non-risk
Observation tables	studentVle , studentAssessment , studentRegistration
Mapped observation entities	LearningTrace , AssessmentResult , RegistrationEvent , RawRecord

Table 1: OULAD protocol and ontology-mapping summary.

Checkpoint	Learners	New traces	New results	New triples	Cum. traces	Cum. triples
Week 4	2,210	148,560	1,386	1,349,804	148,560	1,367,486
Week 8	2,210	106,204	1,674	970,948	254,763	2,523,586
Week 12	2,210	66,638	950	608,326	321,401	3,278,846

Table 2: Incremental twin construction per checkpoint; mean over 6 test modules.

	Week 4		Week 8		Week 12
Method	KG F1	KG AUC	KG F1	KG AUC	KG F1	KG AUC
XGB	0.610	0.735	0.680	0.809	0.718	0.842
RF	0.597	0.725	0.653	0.795	0.719	0.840
ET	0.605	0.733	0.674	0.805	0.723	0.847
HGB	0.581	0.703	0.644	0.778	0.710	0.822
LGBM	0.588	0.700	0.645	0.774	0.702	0.815

Table 3: KG-mediated early-warning performance over six modules; raw-table results are identical and omitted for compactness.

Checkpoint	Alerts	Direct@3	Direct@5	Fid@3
Week 4	5,914	$0.764\pm 0.119$	$0.760\pm 0.120$	$0.898\pm 0.069$
Week 8	5,609	$0.706\pm 0.126$	$0.695\pm 0.088$	$0.828\pm 0.131$
Week 12	6,385	$0.648\pm 0.150$	$0.665\pm 0.079$	$0.744\pm 0.063$

Table 4: Explanation-grounding metrics for the fixed XGB-LDT alert generator; alerts are aggregated over the six test modules.

Early Warning Detection

See all 23 papers

When Can We Trust Early Warnings? Leakage-Excluded Early Outcome Prediction from LMS Interaction Logs
May 25, 2026Ngoc Luyen Le, Marie-Hélène Abel, Bertrand LaforgeData Leakage AI in Education
Gamaizer, 93340 Le Raincy, France · Université de technologie de Compiègne, CNRS, Heudiasyc (Heuristics and Diagnosis of Complex Systems), CS 60319 - 60203 Compiègne Cedex, France · Sorbonne Université, CNRS UMR 7585, LPMHE (Laboratoire de Physique Nucléaire et des Hautes Énergies), 75252 Paris cedex 05, France
Advanced Mathematics Learning Behavior Prediction and Academic Early Warning Model Based on Multimodal Data Analysis
May 31, 2026Liu Qiong, Li ZhengboEducational Assessment AI in Education
Moutai Institute,, Renhuai, 564507, Guizhou, China
A Unified Survival Benchmark for Temporal Dropout Risk Prediction in Learning Analytics
Apr 10, 2026Rafael da Silva, Jeff Eicher, Gregory LongoEducational Data Mining Survival Analysis
Applied Data Science Program, Eastern University, St. Davids, PA, USA.

May 25, 2026J/K move · Enter open · S save

When Can We Trust Early Warnings? Leakage-Excluded Early Outcome Prediction from LMS Interaction Logs

Ngoc Luyen Le, Marie-Hélène Abel, Bertrand Laforge

Gamaizer, 93340 Le Raincy, France · Université de technologie de Compiègne, CNRS, Heudiasyc (Heuristics and Diagnosis of Complex Systems), CS 60319 - 60203 Compiègne Cedex, France · Sorbonne Université, CNRS UMR 7585, LPMHE (Laboratoire de Physique Nucléaire et des Hautes Énergies), 75252 Paris cedex 05, France

Data Leakage AI in Education Educational Data Mining Temporal Data Leakage Early Warning Detection

Open paper AskPDF arXiv

Early-warning models built from Learning Management System (LMS) logs aim to predict end-of-course outcomes early enough to enable timely learner support. However, reported "early" performance is often inflated by temporal leakage. This occurs when the pipeline uses information that would not yet be available at the time of prediction. We formalize cutoff-based early outcome prediction under a temporal availability constraint and introduce LEAP (Leakage-Excluded Early-Availability Protocol), which enforces cutoff-first truncation prior to joins and aggregation and audits feature provenance to prevent post-cutoff evidence from entering the benchmark. We instantiate LEAP on the public Open University Learning Analytics Dataset (OULAD) as a multi-step protocol for leakage-controlled evaluation across weekly cutoffs. Using several standard learning methods, we evaluate performance using ROC-AUC, PR-AUC, Brier score, and [email protected]. Results show improving performance as the observation window expands, with a marked gain around week~3; Random Forest performs best at the earliest cutoffs, while Gradient Boosting dominates thereafter. Leakage ablations further show that temporal violations, especially through assessment information, can inflate apparent "early" performance.

Abstract

Early Warning Detection

Explore similar work

When Can We Trust Early Warnings? Leakage-Excluded Early Outcome Prediction from LMS Interaction Logs

Advanced Mathematics Learning Behavior Prediction and Academic Early Warning Model Based on Multimodal Data Analysis

A Unified Survival Benchmark for Temporal Dropout Risk Prediction in Learning Analytics