@article{ART003372637},
author={Heejin Jung and Gyuejeong Lee},
title={A Study on Machine Learning-Based User Segmentation and Overdue Prediction for Public Libraries},
journal={Journal of the Korean Society for Library and Information Science},
issn={1225-598X},
year={2026},
volume={60},
number={3},
pages={315-338},
doi={10.4275/KSLIS.2026.60.3.315}
TY - JOUR
AU - Heejin Jung
AU - Gyuejeong Lee
TI - A Study on Machine Learning-Based User Segmentation and Overdue Prediction for Public Libraries
JO - Journal of the Korean Society for Library and Information Science
PY - 2026
VL - 60
IS - 3
PB - 한국문헌정보학회
SP - 315
EP - 338
SN - 1225-598X
AB - This study explores the feasibility of a preventive overdue management strategy for public libraries by segmenting user types and optimizing type-specific thresholds for an overdue prediction model based on machine learning. Based on the actual borrowing data from adult users of a C public library (2,148 users, 75,377 transactions), an overdue prediction model was trained with LightGBM. K-Prototypes clustering was then applied to segment users into five distinct types, which showed clearly different overdue rates ranging from 5% to 25%. For each segment, model performance was compared using a single optimal threshold versus segment-specific optimal thresholds. Results showed substantial variation in recall (34.6%-94.2%) and F1-Scores (0.257-0.513) across segments when a single optimal threshold was applied. In contrast, segment-specific thresholds kept recall above 75% for all segments and improved F1-Scores in most cases. One exception was the Senior Excellent segment, where the low overdue rate limited predictive performance, indicating the need for a supplementary management approach. These findings suggest that segment-specific prediction and management strategies can serve as an alternative to standard overdue response practices in public libraries.
KW - Book Overdue;User Segmentation;Cluster Analysis;Overdue Prediction;Public Library
DO - 10.4275/KSLIS.2026.60.3.315
ER -
Heejin Jung and Gyuejeong Lee. (2026). A Study on Machine Learning-Based User Segmentation and Overdue Prediction for Public Libraries. Journal of the Korean Society for Library and Information Science, 60(3), 315-338.
Heejin Jung and Gyuejeong Lee. 2026, "A Study on Machine Learning-Based User Segmentation and Overdue Prediction for Public Libraries", Journal of the Korean Society for Library and Information Science, vol.60, no.3 pp.315-338. Available from: doi:10.4275/KSLIS.2026.60.3.315
Heejin Jung, Gyuejeong Lee "A Study on Machine Learning-Based User Segmentation and Overdue Prediction for Public Libraries" Journal of the Korean Society for Library and Information Science 60.3 pp.315-338 (2026) : 315.
Heejin Jung, Gyuejeong Lee. A Study on Machine Learning-Based User Segmentation and Overdue Prediction for Public Libraries. 2026; 60(3), 315-338. Available from: doi:10.4275/KSLIS.2026.60.3.315
Heejin Jung and Gyuejeong Lee. "A Study on Machine Learning-Based User Segmentation and Overdue Prediction for Public Libraries" Journal of the Korean Society for Library and Information Science 60, no.3 (2026) : 315-338.doi: 10.4275/KSLIS.2026.60.3.315
Heejin Jung; Gyuejeong Lee. A Study on Machine Learning-Based User Segmentation and Overdue Prediction for Public Libraries. Journal of the Korean Society for Library and Information Science, 60(3), 315-338. doi: 10.4275/KSLIS.2026.60.3.315
Heejin Jung; Gyuejeong Lee. A Study on Machine Learning-Based User Segmentation and Overdue Prediction for Public Libraries. Journal of the Korean Society for Library and Information Science. 2026; 60(3) 315-338. doi: 10.4275/KSLIS.2026.60.3.315
Heejin Jung, Gyuejeong Lee. A Study on Machine Learning-Based User Segmentation and Overdue Prediction for Public Libraries. 2026; 60(3), 315-338. Available from: doi:10.4275/KSLIS.2026.60.3.315
Heejin Jung and Gyuejeong Lee. "A Study on Machine Learning-Based User Segmentation and Overdue Prediction for Public Libraries" Journal of the Korean Society for Library and Information Science 60, no.3 (2026) : 315-338.doi: 10.4275/KSLIS.2026.60.3.315
