Forecasting the Demand for Freshmen Alma Mater Jackets and Sports T-Shirts by Size Using a Hybrid Bayesian–Machine Learning Approach
DOI:
https://doi.org/10.30871/jaic.v10i3.12679Keywords:
Admissions Operations, Bayesian Dirichlet– Multinomial, Demand Forecasting, Multiclass Classification, Uniform Size PredictionAbstract
This study addresses an operational procurement problem in university admissions, where alma mater jackets and sports T-shirts for incoming students must be ordered several months before complete size information becomes available. In the case of ITERA admissions, procurement decisions are typically made in March or April, whereas actual student size data are only gradually collected during the re-registration period from April to July. To support earlier and more reliable procurement planning, this study formulates the problem as a size-demand forecasting task covering six categories: S, M, L, XL, XXL, and XXXL. Historical data from 2015 to 2025 were analyzed, with reliable size records concentrated in the 2019–2025 period. The main novelty of this study lies in formulating freshman uniform procurement as a staged forecasting problem that follows the actual admissions workflow. Specifically, the study proposes a hybrid framework that combines: (i) a time-weighted Bayesian Dirichlet–Multinomial model for early-stage aggregate forecasting when current-year size data are not yet available, and (ii) a CatBoostClassifier-based multiclass machine learning model for prediction updates when student attributes become available. Model performance was evaluated using an expanding-window rolling/forward chaining scheme with a one-year forecasting horizon. In addition to conventional historical baselines, the study also included Simple Exponential Smoothing (SES) as a time-series benchmark. Performance was assessed using cross-entropy for size-distribution accuracy, MAE/size and WAPE for quantity prediction, and stockout/overstock for operational impact. The results show that the previous-year proportion remains a strong baseline, while the Bayesian model provides competitive performance and yields posterior uncertainty estimates that are useful for determining safety-oriented order quantities. The statistical analysis further confirms that gender is the most influential predictor of size, while study program, admission track, and province provide complementary but weaker signals. The findings indicate that the proposed framework can support more adaptive and evidence-based procurement planning, reduce the risk of size shortages and excess inventory, and provide a transferable forecasting workflow that may be adapted to other institutions after local recalibration.
Downloads
References
[1] R. J. Hyndman and G. Athanasopoulos, Forecasting: Principles and Practice, 3rd ed. Melbourne, Australia: OTexts, 2021.
[2] S. Chopra, Supply Chain Management: Strategy, Planning, and Operation, 7th ed., Global ed. Pearson, 2019.
[3] J. Heizer, B. Render, and C. Munson, Operations Management: Sustainability and Supply Chain Management, 13th ed. Pearson, 2020.
[4] A. Géron, Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent Systems, 3rd ed. Sebastopol, CA, USA: O’Reilly Media, 2022.
[5] K. Swaminathan and R. Venkitasubramony, “Demand forecasting for fashion products: A systematic review,” Int. J. Forecast., vol. 40, no. 1, pp. 247–267, 2024, doi: 10.1016/j.ijforecast.2023.02.005.
[6] J. Huber and H. Stuckenschmidt, “Daily retail demand forecasting using machine learning with emphasis on calendric special days,” Int. J. Forecast., vol. 36, no. 4, pp. 1420–1438, 2020, doi: 10.1016/j.ijforecast.2020.02.005.
[7] H. Katz, K. T. Brusch, and R. E. Weiss, “A Bayesian Dirichlet auto-regressive moving average model for forecasting lead times,” Int. J. Forecast., vol. 40, no. 4, pp. 1556–1567, 2024, doi: 10.1016/j.ijforecast.2024.01.004.
[8] C. Hanretty, “Forecasting multiparty by-elections using Dirichlet regression,” Int. J. Forecast., vol. 37, no. 4, pp. 1666–1676, 2021, doi: 10.1016/j.ijforecast.2021.03.007.
[9] W. D. Wadsworth, R. Argiento, M. Guindani, J. Galloway-Peña, S. A. Shelburne, and M. Vannucci, “An integrative Bayesian Dirichlet-multinomial regression model for the analysis of taxonomic abundances in microbiome data,” BMC Bioinformatics, vol. 18, Art. no. 94, 2017, doi: 10.1186/s12859-017-1516-0.
[10] L. Prokhorenkova, G. Gusev, A. Vorobev, A. V. Dorogush, and A. Gulin, “CatBoost: unbiased boosting with categorical features,” in Advances in Neural Information Processing Systems 31 (NeurIPS 2018), 2018, pp. 6639–6649.
[11] A. V. Dorogush, V. Ershov, and A. Gulin, “CatBoost: gradient boosting with categorical features support,” arXiv preprint arXiv:1810.11363, 2018, doi: 10.48550/arXiv.1810.11363.
[12] P. Meyer, B. Birregah, P. Beauseroy, E. Grall, and A. Lauxerrois, “Missing body measurements prediction in fashion industry: a comparative approach,” Fash. Text., vol. 10, Art. no. 37, 2023, doi: 10.1186/s40691-023-00357-5.
[13] C. Bergmeir, R. J. Hyndman, and B. Koo, “A note on the validity of cross-validation for evaluating autoregressive time series prediction,” Comput. Stat. Data Anal., vol. 120, pp. 70–83, 2018, doi: 10.1016/j.csda.2017.11.003.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Miranti Verdiana, Eko Dwi Nugroho, Leslie Anggraini, Radhinka Bagaskara
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
Authors who publish with this journal agree to the following terms:
- Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution License (Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) ) that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (See The Effect of Open Access).
