TY  - JOUR
AU  - Mansoor, Muhammad Talha
PY  - 2026
DA  - 2026/03/16
TI  - Deployment Challenges of Bayesian Network–Based Learning Path Recommendation in Real Classroom Settings
JO  - ICCK Journal of Software Engineering
T2  - ICCK Journal of Software Engineering
JF  - ICCK Journal of Software Engineering
VL  - 2
IS  - 1
SP  - 71
EP  - 84
DO  - 10.62762/JSE.2025.490810
UR  - https://www.icck.org/article/abs/JSE.2025.490810
KW  - bayesian networks
KW  - learning path recommendation
KW  - real-world deployment
KW  - curriculum dependency
KW  - adaptive learning systems
KW  - educational data mining
KW  - computational feasibility
AB  - Bayesian Network (BN)–based Learning Path Recommendation (LPR) systems are widely adopted in personalized education for modeling uncertainty and providing interpretable learner representations. However, existing studies predominantly evaluate these systems under controlled settings that assume balanced data, simplified curricula, and unconstrained resources. Consequently, limited empirical understanding exists regarding their performance in authentic classrooms. This study addresses this gap by examining the real-world deployment of a BN-based LPR system using naturally occurring classroom data. The system is evaluated using 426,004 quiz responses collected across 19 formative assessments, reflecting routine instructional practice. Two curricular configurations were examined: an independent-topic structure with no explicit dependencies, and a prerequisite-based structure where mastery depended on prior knowledge. Rather than focusing on predictive accuracy, the study investigates operational feasibility, including structural complexity and memory consumption. From a software-engineering perspective, the central problem addressed is the scalability and inference feasibility of BN-based LPR systems under curriculum dependencies and real-world resource constraints. Results show that BN inference remains stable and computationally tractable for independent topics. In contrast, prerequisite-based modeling substantially increases network density, memory requirements, and inference variability, in some cases causing deployment failures under standard resources. Sparse assessment coverage at higher cognitive levels further undermines recommendation stability. This work provides a deployment-oriented perspective that complements performance-centric research and emphasizes the need for deployment-sensitive evaluation beyond accuracy-focused assessment in real classrooms.
SN  - 3069-1834
PB  - Institute of Central Computation and Knowledge
LA  - English
ER  - 

@article{Mansoor2026Deployment,
  author = {Muhammad Talha Mansoor},
  title = {Deployment Challenges of Bayesian Network–Based Learning Path Recommendation in Real Classroom Settings},
  journal = {ICCK Journal of Software Engineering},
  year = {2026},
  volume = {2},
  number = {1},
  pages = {71-84},
  doi = {10.62762/JSE.2025.490810},
  url = {https://www.icck.org/article/abs/JSE.2025.490810},
  abstract = {Bayesian Network (BN)–based Learning Path Recommendation (LPR) systems are widely adopted in personalized education for modeling uncertainty and providing interpretable learner representations. However, existing studies predominantly evaluate these systems under controlled settings that assume balanced data, simplified curricula, and unconstrained resources. Consequently, limited empirical understanding exists regarding their performance in authentic classrooms. This study addresses this gap by examining the real-world deployment of a BN-based LPR system using naturally occurring classroom data. The system is evaluated using 426,004 quiz responses collected across 19 formative assessments, reflecting routine instructional practice. Two curricular configurations were examined: an independent-topic structure with no explicit dependencies, and a prerequisite-based structure where mastery depended on prior knowledge. Rather than focusing on predictive accuracy, the study investigates operational feasibility, including structural complexity and memory consumption. From a software-engineering perspective, the central problem addressed is the scalability and inference feasibility of BN-based LPR systems under curriculum dependencies and real-world resource constraints. Results show that BN inference remains stable and computationally tractable for independent topics. In contrast, prerequisite-based modeling substantially increases network density, memory requirements, and inference variability, in some cases causing deployment failures under standard resources. Sparse assessment coverage at higher cognitive levels further undermines recommendation stability. This work provides a deployment-oriented perspective that complements performance-centric research and emphasizes the need for deployment-sensitive evaluation beyond accuracy-focused assessment in real classrooms.},
  keywords = {bayesian networks, learning path recommendation, real-world deployment, curriculum dependency, adaptive learning systems, educational data mining, computational feasibility},
  issn = {3069-1834},
  publisher = {Institute of Central Computation and Knowledge}
}