OBE-Constrained Personalized Knowledge Graphs for Micro-Course Path Recommendation with Group-Domain Adaptation

Micro-course recommendations typically optimize click-through or content similarity, rarely guaranteeing Outcome-Based Education (OBE) competency coverage or prerequisite constraints. I propose a unified framework integrating OBE competency mapping, multi-relational knowledge graph construction, dynamic graph updating via Graph Attention Network, constrained path generation, and group-domain adaptive transfer. On a single self-constructed dataset (January 2024–June 2025, 480 learners, 486 knowledge units, 82,760 interactions), under a learner-disjoint split with ten seeds, the method achieves HR@10 0.872 ± 0.009 and NDCG@10 0.731 ± 0.012, outperforming KGCN (0.841 ± 0.011; 0.698 ± 0.014). Cross-group domain adaptation is evaluated under a leave-one-group-domain-out protocol, in which three group-domains are used for training and validation and the remaining group-domain is used as an unseen target for testing. Across the four target folds, the proposed method attains a macro-averaged target-domain adaptation accuracy of 0.863 ± 0.007, with fold-wise results of 0.868 ± 0.009 (East Asian target), 0.856 ± 0.010 (Southeast Asian target), 0.859 ± 0.008 (South Asian target), and 0.869 ± 0.009 (European target); the worst-group target accuracy is 0.856 ± 0.010. Relative to direct transfer, macro and worst-group gains are 8.9 and 8.2 percentage points. These findings are preliminary offline results from one dataset and do not establish causal educational effects, group fairness, or cross-platform generalizability; independent replication with external datasets and controlled studies is required.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-19
DOI
https://doi.org/10.21203/rs.3.rs-11183354/v1
Primary Topic
Recommender Systems and Techniques
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

OBE-Constrained Personalized Knowledge Graphs for Micro-Course Path Recommendation with Group-Domain Adaptation

Zhizhang He
Research Square
Recommender Systems and Techniques
preprint

OBE-Constrained Personalized Knowledge Graphs for Micro-Course Path Recommendation with Group-Domain Adaptation

Zhizhang He
preprint en

Abstract

Micro-course recommendations typically optimize click-through or content similarity, rarely guaranteeing Outcome-Based Education (OBE) competency coverage or prerequisite constraints. I propose a unified framework integrating OBE competency mapping, multi-relational knowledge graph construction, dynamic graph updating via Graph Attention Network, constrained path generation, and group-domain adaptive transfer. On a single self-constructed dataset (January 2024–June 2025, 480 learners, 486 knowledge units, 82,760 interactions), under a learner-disjoint split with ten seeds, the method achieves HR@10 0.872 ± 0.009 and NDCG@10 0.731 ± 0.012, outperforming KGCN (0.841 ± 0.011; 0.698 ± 0.014). Cross-group domain adaptation is evaluated under a leave-one-group-domain-out protocol, in which three group-domains are used for training and validation and the remaining group-domain is used as an unseen target for testing. Across the four target folds, the proposed method attains a macro-averaged target-domain adaptation accuracy of 0.863 ± 0.007, with fold-wise results of 0.868 ± 0.009 (East Asian target), 0.856 ± 0.010 (Southeast Asian target), 0.859 ± 0.008 (South Asian target), and 0.869 ± 0.009 (European target); the worst-group target accuracy is 0.856 ± 0.010. Relative to direct transfer, macro and worst-group gains are 8.9 and 8.2 percentage points. These findings are preliminary offline results from one dataset and do not establish causal educational effects, group fairness, or cross-platform generalizability; independent replication with external datasets and controlled studies is required.

Research Square
Yan'an University (CN)
Recommender Systems and Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon