DEVELOPMENT OF INTELLIGENT ADVISORY SYSTEM WITH COGNITIVE TECHNOLOGY
4
Issued Date
2025-01-01
Resource Type
ISSN
19928645
eISSN
18173195
Scopus ID
2-s2.0-105028179086
Journal Title
Journal of Theoretical and Applied Information Technology
Volume
103
Issue
24
Start Page
10281
End Page
10292
Rights Holder(s)
SCOPUS
Bibliographic Citation
Journal of Theoretical and Applied Information Technology Vol.103 No.24 (2025) , 10281-10292
Suggested Citation
Suaprae P., Nilsook P., Wannapiroon P., Nittayathammakul V. DEVELOPMENT OF INTELLIGENT ADVISORY SYSTEM WITH COGNITIVE TECHNOLOGY. Journal of Theoretical and Applied Information Technology Vol.103 No.24 (2025) , 10281-10292. 10292. Retrieved from: https://repository.li.mahidol.ac.th/handle/123456789/114103
Title
DEVELOPMENT OF INTELLIGENT ADVISORY SYSTEM WITH COGNITIVE TECHNOLOGY
Corresponding Author(s)
Other Contributor(s)
Abstract
Existing research on student retention mainly focuses on risk prediction, with few studies implementing advisory processes that translate predictions into timely, personalized interventions. This study develops and evaluates the intelligent advisory system with cognitive technology (referred to as the IAS-CT system) to improve student retention in higher education. The persisting gap in the literature is that most retention studies stop at risk prediction and rarely operationalize a closed-loop advisory workflow that converts predictions into timely, personalized interventions. Using de-identified institutional records of 2,973 undergraduates from academic years 2019–2022 with 25 academic and socio-demographic features, we trained and compared Decision Trees, Logistic Regression, Random Forest, K-Nearest Neighbors, and Naive Bayes. Preprocessing comprised imputation, normalization, and categorical encoding/selection; evaluation used a stratified split and standard metrics (accuracy, precision, recall, and F1) with confusion matrices. Correlation analysis indicated that GPA (r = 0.55), absenteeism (r = 0.48), father’s income (r = 0.45), year of study (r = 0.38), and field of study (r = 0.20) were the most associated factors with retention. Decision Trees achieved the best predictive performance (accuracy = 98.90%), exceeding Logistic Regression (97.40%), Random Forest (86.10%), K-Nearest Neighbors (85.90%), and Naive Bayes (85.80%). The selected model was integrated into an advisory architecture that issues early-warning alerts, generates personalized study recommendations, and supports advisor–student communication. An expert panel rated the system’s suitability at an overall high level. Consequently, the system operationalizes prediction into intervention, providing actionable retention support with practical implications for data governance and institutional scaling.