pith. machine review for the scientific record. sign in

arxiv: 2602.12120 · v3 · submitted 2026-02-12 · 💻 cs.AI

Forecasting Commencing Enrolments Under Data Sparsity: A Zero-Shot Time Series Foundation Models Framework for Higher Education Planning

Jittarin Jetwiriyanon , Teo Susnjak , Surangika Ranathunga This is my paper

Pith reviewed 2026-05-16 02:29 UTC · model grok-4.3

classification 💻 cs.AI
keywords enrolment forecastingzero-shot learningtime series foundation modelsdata sparsityhigher education planningcovariate conditioningGoogle Trendsstructural shifts
0
0 comments X

The pith

Covariate-conditioned zero-shot time series models match classical methods for forecasting university enrolments with sparse data.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper tests whether zero-shot Time Series Foundation Models can generate reliable annual forecasts for commencing university enrolments when only limited historical data is available and structural changes disrupt the series. It introduces a protocol that adds covariates from Google Trends and a new Institutional Operating Conditions Index while preventing any future information from entering the model at decision time. Backtests against standard operational methods show the conditioned models are competitive and sometimes more accurate without any institution-specific retraining. The result matters for university planners who must allocate staff, facilities and budgets under data constraints and sudden shifts in student behaviour.

Core claim

The paper establishes that zero-shot TSFMs conditioned on a leakage-safe covariate protocol that integrates feature-engineered Google Trends with the Institutional Operating Conditions Index achieve accuracy competitive with or better than classical statistical baselines in expanding-window backtests that respect real decision timing, thereby supplying usable enrolment forecasts without requiring bespoke model training at each institution.

What carries the argument

Zero-shot Time Series Foundation Models conditioned through a leakage-safe covariate protocol that combines Google Trends features with the Institutional Operating Conditions Index to detect environmental regime shifts.

If this is right

  • University administrators can obtain enrolment forecasts without assembling long internal time series or training custom models.
  • The same protocol supplies auditable forecasts under structural instability such as policy changes or economic shocks.
  • Forecast quality varies with cohort type and the precise design of the covariates, so institutions must still validate the approach locally.
  • The method reduces the barrier to using advanced time-series tools for operational planning in data-poor settings.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The protocol could transfer to other domains that forecast sparse event counts such as course registrations or campus facility usage.
  • Institutions might embed the conditioned models inside existing planning dashboards to produce scenario-based enrolment ranges.
  • The Institutional Operating Conditions Index itself may function as a reusable regime indicator for forecasting in adjacent sectors.
  • Further tests on multi-year horizons or international datasets would clarify the limits of zero-shot performance under different sparsity levels.

Load-bearing premise

The Google Trends and Institutional Operating Conditions Index covariates reflect genuine environmental changes without any future data leaking into the forecast at the time of decision.

What would settle it

An expanding-window backtest on a fresh collection of institutions in which the zero-shot models produce materially higher forecast errors than the classical baselines across multiple cohorts would refute the claim of competitiveness.

Figures

Figures reproduced from arXiv: 2602.12120 by Jittarin Jetwiriyanon, Surangika Ranathunga, Teo Susnjak.

Figure 1
Figure 1. Figure 1: TSFMs architecture for multivariate time series forecasting [57]. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Experiment workflow. 12 [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Actual and forecast annual new domestic-student enrolments. Panel (a) shows unconditional forecasts. Panel (b) shows the [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Actual and forecast annual new international-student enrolments. Panel (a) shows unconditional forecasts. [PITH_FULL_IMAGE:figures/full_fig_p016_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Empirical CDF of PIT values. The incorporation of IOCI reveals a heterogeneous response. The Chronos-Bolt family derives substantial benefit from this covariate: Chronos-Bolt-Tiny and Chronos-Bolt-Small improved under IOCI, with ∆MAE values of 50.22 and 47.49 relative to Persistence, respectively. Conversely, TimesFM and ARIMAX degraded sharply after the introduction of the index. This divergence suggests … view at source ↗
read the original abstract

Effective resource allocation in higher education depends on reliable enrolment forecasts, yet institutional planners frequently face data series disrupted by structural shifts. This paper investigates whether zero-shot Time Series Foundation Models (TSFMs) can provide rigorous decision support for annual enrolment forecasting under severe data sparsity. We benchmark multiple TSFMs against classical operational baselines using an expanding-window backtest that mirrors decision-time constraints. To capture environmental shifts without look-ahead bias, we introduce a leakage-safe covariate protocol that integrates feature-engineered Google Trends with the Institutional Operating Conditions Index (IOCI), a transferable regime measure extracted from historical narrative evidence. Our evaluation demonstrates that covariate-conditioned TSFMs are competitive with classical methods and can improve accuracy without requiring bespoke institutional training. However, the operational benefits depend on cohort characteristics and covariate design. This study provides an auditable and transferable forecasting protocol for operational researchers and university administrators, helping institutions determine when context-aware forecasting adds practical value under limited data and structural instability.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper claims that zero-shot Time Series Foundation Models (TSFMs) conditioned on a leakage-safe covariate protocol—combining feature-engineered Google Trends with the Institutional Operating Conditions Index (IOCI) derived from historical narratives—can deliver competitive annual enrolment forecasts for higher education under severe data sparsity. It supports this via an expanding-window backtest that mirrors decision-time constraints, showing accuracy gains over classical baselines without requiring bespoke institutional training, while noting that benefits vary by cohort and covariate design.

Significance. If the central claim holds after verification, the work offers practical value for operational researchers and university administrators by providing an auditable, transferable forecasting protocol that leverages foundation models in zero-shot settings. This addresses a real gap in higher-education planning where data series are short and disrupted by structural shifts, potentially improving resource allocation without extensive custom modeling.

major comments (2)
  1. [Abstract and Evaluation] Abstract and Evaluation section: The expanding-window backtest description does not explicitly document the precise lag structure for Google Trends and IOCI alignment to each forecast origin (e.g., truncation of Google Trends to the last date known at origin, enforcement of IOCI narrative extraction windows, and whether any smoothing or feature engineering uses post-origin data). This documentation is load-bearing for the no-look-ahead-bias claim and the reported accuracy improvements.
  2. [Covariate Protocol] Covariate Protocol subsection: The claim that the protocol 'captures environmental shifts without look-ahead bias' requires concrete verification steps (e.g., pseudocode or explicit rules for origin-aligned feature construction) because any inadvertent post-origin information would invalidate the competitiveness result relative to classical methods.
minor comments (2)
  1. [Abstract] Abstract: Specify the exact TSFMs and classical baselines compared, along with the performance metrics used (e.g., MAE, RMSE) to allow readers to assess the 'competitive' claim quantitatively.
  2. [Introduction] Notation: Define IOCI more formally on first use, including its construction from narrative evidence, to improve transferability for other institutions.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which correctly identify areas where additional methodological transparency is needed to fully substantiate our no-look-ahead-bias claims. We address each point below and will revise the manuscript to incorporate the requested documentation.

read point-by-point responses

  1. Referee: [Abstract and Evaluation] Abstract and Evaluation section: The expanding-window backtest description does not explicitly document the precise lag structure for Google Trends and IOCI alignment to each forecast origin (e.g., truncation of Google Trends to the last date known at origin, enforcement of IOCI narrative extraction windows, and whether any smoothing or feature engineering uses post-origin data). This documentation is load-bearing for the no-look-ahead-bias claim and the reported accuracy improvements.

    Authors: We agree that the current description of the expanding-window backtest is insufficiently detailed on lag structures and alignment. In the revised manuscript we will expand the Evaluation section with explicit rules and pseudocode for origin-aligned feature construction. Google Trends series will be truncated to the last date known at each forecast origin; IOCI narrative extraction will be restricted to historical windows ending at the origin; and all smoothing or feature engineering steps will be confirmed to use only pre-origin data. These additions will make the no-look-ahead-bias claim directly verifiable. revision: yes

  2. Referee: [Covariate Protocol] Covariate Protocol subsection: The claim that the protocol 'captures environmental shifts without look-ahead bias' requires concrete verification steps (e.g., pseudocode or explicit rules for origin-aligned feature construction) because any inadvertent post-origin information would invalidate the competitiveness result relative to classical methods.

    Authors: We concur that concrete verification steps are required. The revised Covariate Protocol subsection will include pseudocode together with step-by-step rules for origin-aligned construction. These will specify truncation of Google Trends to information available at the origin, restriction of IOCI extraction to pre-origin narratives, and explicit checks confirming that no post-origin data enters feature engineering. This documentation will directly support the validity of the reported competitiveness results. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical benchmark protocol is self-contained

full rationale

The paper describes an expanding-window backtest and a leakage-safe covariate protocol (Google Trends + IOCI) for zero-shot TSFM evaluation. No equations, fitted parameters renamed as predictions, or self-citation chains appear in the provided text. The central claim rests on external benchmark comparisons whose inputs (historical enrolment series, narrative-derived IOCI) are independent of the reported accuracy numbers. This is the expected non-finding for an applied forecasting study.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

Based on abstract only; the framework rests on the assumption that IOCI provides a transferable regime signal and that the covariate protocol avoids leakage.

axioms (1)
  • domain assumption IOCI is a transferable regime measure extracted from historical narrative evidence that captures environmental shifts
    Invoked to justify covariate use without look-ahead bias
invented entities (1)
  • Institutional Operating Conditions Index (IOCI) no independent evidence
    purpose: To integrate contextual regime information into the forecasting model
    Newly introduced as a summary measure derived from narrative evidence

pith-pipeline@v0.9.0 · 5482 in / 1160 out tokens · 94149 ms · 2026-05-16T02:29:53.361659+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

102 extracted references · 102 canonical work pages · 2 internal anchors

  1. [1]

    An improved method for forecasting enrollments based on fuzzy time series and particle swarm optimization.Expert Systems with Applications, 36(3, Part 2):6108–6117, 2009

    I-Hong Kuo, Shi-Jinn Horng, Tzong-Wann Kao, Tsung-Lieh Lin, Cheng-Ling Lee, and Yi Pan. An improved method for forecasting enrollments based on fuzzy time series and particle swarm optimization.Expert Systems with Applications, 36(3, Part 2):6108–6117, 2009. ISSN 0957-4174. doi:10.1016/j.eswa.2008.07.043. URL https://www.sciencedirect.com/science/article/...

    work page doi:10.1016/j.eswa.2008.07.043 2009

  2. [2]

    A hybrid forecasting model for enrollments based on aggregated fuzzy time series and particle swarm optimization.Expert Systems with Applications, 38(7):8014–8023, 2011

    Yao-Lin Huang, Shi-Jinn Horng, Mingxing He, Pingzhi Fan, Tzong-Wann Kao, Muhammad Khurram Khan, Jui-Lin Lai, and I-Hong Kuo. A hybrid forecasting model for enrollments based on aggregated fuzzy time series and particle swarm optimization.Expert Systems with Applications, 38(7):8014–8023, 2011. ISSN 0957-

    work page 2011

  3. [3]

    URL https://www.sciencedirect.com/science/article/pii/ S0957417410014909

    doi:10.1016/j.eswa.2010.12.127. URL https://www.sciencedirect.com/science/article/pii/ S0957417410014909

    work page doi:10.1016/j.eswa.2010.12.127 2010

  4. [4]

    Examining the relationship between institutional financial stress and stu- dent enrollment

    Robert Kelchen and Holly Evans. Examining the relationship between institutional financial stress and stu- dent enrollment. Working Paper 24-05, Federal Reserve Bank of Philadelphia, July 2024. URL https: //robertkelchen.com/wp-content/uploads/2024/07/hcm_enrollment_jul24.pdf

    work page 2024

  5. [5]

    Johns Hopkins University Press, Baltimore, MD, 2018

    Nathan D Grawe.Demographics and the Demand for Higher Education. Johns Hopkins University Press, Baltimore, MD, 2018

    work page 2018

  6. [6]

    Predictive analytics in higher education: The promises and challenges of using machine learning to improve student success.AIR Professional File, (161):1–18, 2023

    Kelli A Bird. Predictive analytics in higher education: The promises and challenges of using machine learning to improve student success.AIR Professional File, (161):1–18, 2023. URL https://www.airweb.org/ docs/default-source/aha/apf-161-predictive-analytics-in-higher-education.pdf?sfvrsn= f31dcce4_1

    work page 2023

  7. [7]

    Towards redefining strategic enrollment management: Response to emerging challenges of the 21st century.Journal of Advancing Education Practice, 5(2):45–58, 2024

    K Willis and P Yang. Towards redefining strategic enrollment management: Response to emerging challenges of the 21st century.Journal of Advancing Education Practice, 5(2):45–58, 2024. URL https://openriver. winona.edu/jaep/vol5/iss2/7/

    work page 2024

  8. [8]

    Wald,General Relativity, Chicago Univ

    Bridget Terry Long.The Financial Crisis and College Enrollment: How Have Students and Their Families Responded?, pages 209–233. University of Chicago Press, January 2014. doi:10.7208/chicago/9780226201979.003.0007. URLhttp://www.nber.org/chapters/c12862

    work page doi:10.7208/chicago/9780226201979.003.0007 2014

  9. [9]

    Intermittent demand forecasting for spare parts with little historical informa- tion.Journal of the Operational Research Society, 76(2):294–309, 2025

    Thomas C. Sharkey, Steven Foster, Sudeep Hegde, Mary E. Kurz, and Emily L. Tucker. A categoriza- tion of observed uses of operational research models for fundamental surprise events: Observations from university operations during covid-19.Journal of the Operational Research Society, 76(2):254–266, 2025. doi:10.1080/01605682.2024.2346117. URLhttps://doi.or...

    work page doi:10.1080/01605682.2024.2346117 2025

  10. [10]

    Analyzing and forecasting enrolment patterns using time series models.International Journal of Advanced Computer Science, 12(1):102–115, 2025

    J M Mendez et al. Analyzing and forecasting enrolment patterns using time series models.International Journal of Advanced Computer Science, 12(1):102–115, 2025

    work page 2025

  11. [11]

    The financial sustainability of higher educa- tion: Insights from policy in oecd countries, 2025

    Organisation for Economic Co-operation and Development. The financial sustainability of higher educa- tion: Insights from policy in oecd countries, 2025. URL https://www.oecd.org/en/publications/ the-financial-sustainability-of-higher-education_f544ccfe-en.html . Published 24 November 2025

    work page 2025

  12. [12]

    Education at a glance 2024: Oecd indi- cators, 2024

    Organisation for Economic Co-operation and Development. Education at a glance 2024: Oecd indi- cators, 2024. URL https://www.oecd-ilibrary.org/education/education-at-a-glance-2024_ c00cad36-en. Published 10 September 2024

    work page 2024

  13. [13]

    Maltz, Kenneth E

    Elliot N. Maltz, Kenneth E. Murphy, and Michael L. Hand. Decision support for university enrollment man- agement: Implementation and experience.Decision Support Systems, 44(1):106–123, 2007. ISSN 0167-

    work page 2007

  14. [14]

    URL https://www.sciencedirect.com/science/article/pii/ S016792360700053X

    doi:10.1016/j.dss.2007.03.008. URL https://www.sciencedirect.com/science/article/pii/ S016792360700053X. 21 Zero-Shot Time Series Foundation Models for Annual Institutional Forecasting Under Data SparsityA PREPRINT

    work page doi:10.1016/j.dss.2007.03.008 2007

  15. [15]

    Adapt to small-scale and long-term time series forecasting with enhanced multidimensional correlation.Expert Systems with Applications, 238:122203, 2024

    Xinshuai Li, Senlin Luo, Limin Pan, and Zhouting Wu. Adapt to small-scale and long-term time series forecasting with enhanced multidimensional correlation.Expert Systems with Applications, 238:122203, 2024. ISSN 0957-

    work page 2024

  16. [16]

    URL https://www.sciencedirect.com/science/article/pii/ S0957417423027057

    doi:10.1016/j.eswa.2023.122203. URL https://www.sciencedirect.com/science/article/pii/ S0957417423027057

    work page doi:10.1016/j.eswa.2023.122203 2023

  17. [17]

    The m4 competition: 100,000 time series and 61 forecasting methods.International Journal of Forecasting, 36(1):54–74, 2020

    Spyros Makridakis, Evangelos Spiliotis, and Vassilios Assimakopoulos. The m4 competition: 100,000 time series and 61 forecasting methods.International Journal of Forecasting, 36(1):54–74, 2020. ISSN 0169-

    work page 2020

  18. [18]

    URL https://www.sciencedirect.com/science/article/ pii/S0169207019301128

    doi:10.1016/j.ijforecast.2019.04.014. URL https://www.sciencedirect.com/science/article/ pii/S0169207019301128. M4 Competition

    work page doi:10.1016/j.ijforecast.2019.04.014 2019

  19. [19]

    Hyndman, and Bonsoo Koo

    Christoph Bergmeir, Rob J. Hyndman, and Bonsoo Koo. A note on the validity of cross-validation for evaluating autoregressive time series prediction.Computational Statistics & Data Analysis, 120:70–83, 2018. ISSN 0167-

    work page 2018

  20. [20]

    URL https://www.sciencedirect.com/science/article/pii/ S0167947317302384

    doi:10.1016/j.csda.2017.11.003. URL https://www.sciencedirect.com/science/article/pii/ S0167947317302384

    work page doi:10.1016/j.csda.2017.11.003 2017

  21. [21]

    Sasd: Self-attention for small datasets—a case study in smart villages.Expert Systems with Applications, 271:126245, 2025

    Daniel Bolaños-Martinez, Alberto Durán-López, Jose Luis Garrido, Blanca Delgado-Márquez, and Maria Bermudez-Edo. Sasd: Self-attention for small datasets—a case study in smart villages.Expert Systems with Applications, 271:126245, 2025. ISSN 0957-4174. doi:10.1016/j.eswa.2024.126245. URL https: //www.sciencedirect.com/science/article/pii/S0957417424031129

    work page doi:10.1016/j.eswa.2024.126245 2025

  22. [22]

    Maddix, Michael W

    Abdul Fatir Ansari, Lorenzo Stella, Caner Turkmen, Xiyuan Zhang, Pedro Mercado, Huibin Shen, Oleksandr Shchur, Syama Syndar Rangapuram, Sebastian Pineda Arango, Shubham Kapoor, Jasper Zschiegner, Danielle C. Maddix, Michael W. Mahoney, Kari Torkkola, Andrew Gordon Wilson, Michael Bohlke-Schneider, and Yuyang Wang. Chronos: Learning the language of time se...

    work page 2024

  23. [23]

    A decoder-only foundation model for time-series forecasting

    Abhimanyu Das, Weihao Kong, Rajat Sen, and Yichen Zhou. A decoder-only foundation model for time-series forecasting. InProceedings of the 41st International Conference on Machine Learning (ICML), volume 235 of Proceedings of Machine Learning Research, pages 9510–9529. PMLR, 2024. URL https://proceedings. mlr.press/v235/das24c.html

    work page 2024

  24. [24]

    Unified training of universal time series forecasting transformers

    Gerald Woo, Chenghao Liu, Akshat Kumar, Caiming Xiong, Silvio Savarese, and Doyen Sahoo. Unified training of universal time series forecasting transformers. InForty-first International Conference on Machine Learning, 2024

    work page 2024

  25. [25]

    Hyunyoung Choi and Hal R. Varian. Predicting the present with google trends.Economic Record, 88(s1):2–9,

    work page

  26. [26]

    doi:10.1111/j.1475-4932.2012.00809.x

    work page doi:10.1111/j.1475-4932.2012.00809.x 2012

  27. [27]

    Forecasting macroeconomic data with bayesian vars: Sparse or dense? it depends!International Journal of Forecasting, 41(4):1589–1619, 2025

    Luis Gruber and Gregor Kastner. Forecasting macroeconomic data with bayesian vars: Sparse or dense? it depends!International Journal of Forecasting, 41(4):1589–1619, 2025. ISSN 0169-2070. doi:https://doi.org/10.1016/j.ijforecast.2025.02.001. URL https://www.sciencedirect.com/science/ article/pii/S0169207025000081

    work page doi:10.1016/j.ijforecast.2025.02.001 2025

  28. [28]

    Eichenauer, Ronald Indergand, Isabel Z

    Vera Z. Eichenauer, Ronald Indergand, Isabel Z. MartÃnez, and Christoph Sax. Obtaining consistent time series from google trends.Economic Inquiry, 60(2):694–705, April 2022. doi:10.1111/ecin.13049. URL https://ideas.repec.org/a/bla/ecinqu/v60y2022i2p694-705.html

    work page doi:10.1111/ecin.13049 2022

  29. [29]

    Justin Grimmer and Brandon M. Stewart. Text as data: The promise and pitfalls of automatic content analysis methods for political texts.Political Analysis, 21(3):267–297, 2013. doi:10.1093/pan/mps028

    work page doi:10.1093/pan/mps028 2013

  30. [30]

    Baker, Nicholas Bloom, and Steven J

    Scott R. Baker, Nicholas Bloom, and Steven J. Davis. Measuring economic policy uncertainty*.The Quarterly Journal of Economics, 131(4):1593–1636, 07 2016. ISSN 0033-5533. doi:10.1093/qje/qjw024. URL https: //doi.org/10.1093/qje/qjw024

    work page doi:10.1093/qje/qjw024 2016

  31. [31]

    William J. Webster. The cohort-survival ratio method in the projection of school attendance.The Journal of Experimental Education, 39(1):89–96, 1970. doi:10.1080/00220973.1970.11011238. URL https://doi.org/ 10.1080/00220973.1970.11011238

    work page doi:10.1080/00220973.1970.11011238 1970

  32. [32]

    Persistence modeling for sales prediction: A simple, self-contained approach.Journal of Business Research, 166:114103, 2023

    Giacomo Sbrana and Paolo Antonetti. Persistence modeling for sales prediction: A simple, self-contained approach.Journal of Business Research, 166:114103, 2023. ISSN 0148-2963. doi:10.1016/j.jbusres.2023.114103. URLhttps://www.sciencedirect.com/science/article/pii/S0148296323004617

    work page doi:10.1016/j.jbusres.2023.114103 2023

  33. [33]

    Ho and M

    S.L. Ho and M. Xie. The use of arima models for reliability forecasting and analysis.Computers & Industrial Engineering, 35(1):213–216, 1998. ISSN 0360-8352. doi:10.1016/S0360-8352(98)00066-7. URL https: //www.sciencedirect.com/science/article/pii/S0360835298000667

    work page doi:10.1016/s0360-8352(98)00066-7 1998

  34. [34]

    Undergraduate international student enrollment forecasting model: An application of time series analysis.Journal of International Students, 9(1):242–261, Jan

    Yu Chen, Ran Li, and Linda Serra Hagedorn. Undergraduate international student enrollment forecasting model: An application of time series analysis.Journal of International Students, 9(1):242–261, Jan. 2019. doi:10.32674/jis.v9i1.266. URLhttps://ojed.org/jis/article/view/266. 22 Zero-Shot Time Series Foundation Models for Annual Institutional Forecasting ...

    work page doi:10.32674/jis.v9i1.266 2019

  35. [35]

    Macroeconomic forecasting using diffusion indexes.Journal of Business & Economic Statistics, 20(2):147–162, 2002

    James H Stock and Mark W Watson. Macroeconomic forecasting using diffusion indexes.Journal of Business & Economic Statistics, 20(2):147–162, 2002. doi:10.1198/073500102317351921. URL https://doi.org/10. 1198/073500102317351921

    work page doi:10.1198/073500102317351921 2002

  36. [36]

    Forecasting school enrollments in the australian capital territory.Journal of the Royal Statistical Society Series A: Statistics in Society, 188(4):1107–1124, 10 2024

    Tianyu Shen, James Raymer, and Caroline Hendy. Forecasting school enrollments in the australian capital territory.Journal of the Royal Statistical Society Series A: Statistics in Society, 188(4):1107–1124, 10 2024. ISSN 0964-1998. doi:10.1093/jrsssa/qnae094. URLhttps://doi.org/10.1093/jrsssa/qnae094

    work page doi:10.1093/jrsssa/qnae094 2024

  37. [37]

    Intermittent demand forecasting for spare parts with little historical informa- tion.Journal of the Operational Research Society, 76(2):294–309, 2025

    Marco Caserta and Luca D’Angelo. Intermittent demand forecasting for spare parts with little historical informa- tion.Journal of the Operational Research Society, 76(2):294–309, 2025. doi:10.1080/01605682.2024.2349734. URLhttps://doi.org/10.1080/01605682.2024.2349734

    work page doi:10.1080/01605682.2024.2349734 2025

  38. [38]

    A novel fractional grey system model with non-singular exponen- tial kernel for forecasting enrollments.Expert Systems with Applications, 219:119652, 2023

    Wanli Xie, Chong Liu, and Wen-Ze Wu. A novel fractional grey system model with non-singular exponen- tial kernel for forecasting enrollments.Expert Systems with Applications, 219:119652, 2023. ISSN 0957-

    work page 2023

  39. [39]

    URL https://www.sciencedirect.com/science/article/pii/ S0957417423001537

    doi:10.1016/j.eswa.2023.119652. URL https://www.sciencedirect.com/science/article/pii/ S0957417423001537

    work page doi:10.1016/j.eswa.2023.119652 2023

  40. [40]

    Andrew Barr and Sarah E. Turner. Out of work and into school: Labor market policies and college enrollment during the great recession.Journal of Public Economics, 127:63–73, 2015. doi:10.1016/j.jpubeco.2015.04.002

    work page doi:10.1016/j.jpubeco.2015.04.002 2015

  41. [41]

    Zimmermann

    Nikos Askitas and Klaus F. Zimmermann. Google econometrics and unemployment forecasting. RatSWD Research Notes 41, German Council for Social and Economic Data (RatSWD), June 2009. URL https: //ssrn.com/abstract=1480251. Available at SSRN

    work page 2009

  42. [42]

    Adam Hale Shapiro, Moritz Sudhof, and Daniel J. Wilson. Measuring news sentiment.Journal of Econo- metrics, 228(2):221–243, 2022. ISSN 0304-4076. doi:10.1016/j.jeconom.2020.07.053. URL https://www. sciencedirect.com/science/article/pii/S0304407620303535

    work page doi:10.1016/j.jeconom.2020.07.053 2022

  43. [43]

    Ilin et al

    C. Ilin et al. Public mobility data enables covid-19 forecasting and management at local and global scales. Scientific Reports, 11:13531, 2021. doi:10.1038/s41598-021-92866-z

    work page doi:10.1038/s41598-021-92866-z 2021

  44. [44]

    Hush, Tushar Ojha, and Terry Babbitt

    Ahmad Slim, Don R. Hush, Tushar Ojha, and Terry Babbitt. Predicting student enrollment based on student and college characteristics. InEducational Data Mining, 2018. URL https://api.semanticscholar.org/ CorpusID:52173545

    work page 2018

  45. [45]

    Relative performance of methods for forecasting special events.Journal of Business Research, 68(8):1785–1791,

    Konstantinos Nikolopoulos, Akrivi Litsa, Fotios Petropoulos, Vasileios Bougioukos, and Marwan Khammash. Relative performance of methods for forecasting special events.Journal of Business Research, 68(8):1785–1791,

    work page

  46. [46]

    doi:https://doi.org/10.1016/j.jbusres.2015.03.037

    ISSN 0148-2963. doi:https://doi.org/10.1016/j.jbusres.2015.03.037. URL https://www.sciencedirect. com/science/article/pii/S0148296315001551. Special Issue on Simple Versus Complex Forecasting

    work page doi:10.1016/j.jbusres.2015.03.037 2015

  47. [47]

    Zheqi Wang, Jonathan Crook, and Galina Andreeva. Improving the accuracy of credit scoring models using an innovative bayesian informative prior specification method.Journal of the Operational Research Society, 76(2): 229–253, 2025. doi:10.1080/01605682.2024.2339510. URL https://doi.org/10.1080/01605682.2024. 2339510

    work page doi:10.1080/01605682.2024.2339510 2025

  48. [48]

    Özgür Arik, Nicolas Loeff, and Tomas Pfister

    Bryan Lim, S. Özgür Arik, Nicolas Loeff, and Tomas Pfister. Temporal fusion transformers for inter- pretable multi-horizon time series forecasting.International Journal of Forecasting, 37(4):1748–1764, 2021. doi:10.1016/j.ijforecast.2021.03.012

    work page doi:10.1016/j.ijforecast.2021.03.012 2021

  49. [49]

    McGraw-Hill, New York, 1959

    Robert Goodell Brown.Statistical Forecasting for Inventory Control. McGraw-Hill, New York, 1959

    work page 1959

  50. [50]

    Kempa-Liehr

    Maximilian Christ, Nils Braun, Julius Neuffer, and Andreas W. Kempa-Liehr. Time series feature extraction on basis of scalable hypothesis tests (tsfresh – a python package).Neurocomputing, 307:72–77, 2018. ISSN 0925-

    work page 2018

  51. [51]

    URL https://www.sciencedirect.com/science/article/ pii/S0925231218304843

    doi:10.1016/j.neucom.2018.03.067. URL https://www.sciencedirect.com/science/article/ pii/S0925231218304843

    work page doi:10.1016/j.neucom.2018.03.067 2018

  52. [52]

    Andrew Halterman and Katherine A. Keith. Codebook llms: Evaluating llms as measurement tools for political science concepts.Political Analysis, page 1–17, September 2025. ISSN 1476-4989. doi:10.1017/pan.2025.10017. URLhttp://dx.doi.org/10.1017/pan.2025.10017

    work page doi:10.1017/pan.2025.10017 2025

  53. [53]

    Lgtime: Leveraging llms with feature-aware processing and multi-granularity fusion for zero-shot time series forecasting

    Shujie Wu, Yao Zhang, YiWei Yang, Zhen Li, GuangYu Yu, Long Chen, Jianwei Xu, and Aziguli Wulamu. Lgtime: Leveraging llms with feature-aware processing and multi-granularity fusion for zero-shot time series forecasting. Expert Systems with Applications, 297:129483, 2026. ISSN 0957-4174. doi:10.1016/j.eswa.2025.129483. URL https://www.sciencedirect.com/sci...

    work page doi:10.1016/j.eswa.2025.129483 2026

  54. [54]

    Do ais know what the most important issue is? using language models to code open-text social survey responses at scale.Research & Politics, 11(1):20531680241231468, 2024

    Jonathan Mellon, Jack Bailey, Ralph Scott, James Breckwoldt, Marta Miori, and Phillip Schmedeman. Do ais know what the most important issue is? using language models to code open-text social survey responses at scale.Research & Politics, 11(1):20531680241231468, 2024. doi:10.1177/20531680241231468. URL https://doi.org/10.1177/20531680241231468. 23 Zero-Sh...

    work page doi:10.1177/20531680241231468 2024

  55. [55]

    From text to quantified insights: A large-scale llm analysis of central bank communication, 2025

    Thiago Christiano Silva, Kei Moriya, and Romain M Veyrune. From text to quantified insights: A large-scale llm analysis of central bank communication, 2025. URL https://www.elibrary.imf.org/view/journals/ 001/2025/109/001.2025.issue-109-en.xml

    work page 2025

  56. [56]

    Shubham Atreja, Joshua Ashkinaze, Lingyao Li, Julia Mendelsohn, and Libby Hemphill. What’s in a prompt?: A large-scale experiment to assess the impact of prompt design on the compliance and accuracy of llm-generated text annotations.Proceedings of the International AAAI Conference on Web and Social Media, 19:122–145, June

    work page

  57. [57]

    doi:10.1609/icwsm.v19i1.35807

    ISSN 2162-3449. doi:10.1609/icwsm.v19i1.35807. URL http://dx.doi.org/10.1609/icwsm.v19i1. 35807

    work page doi:10.1609/icwsm.v19i1.35807

  58. [58]

    Miranda-Belmonte, Victor Muñiz-Sánchez, and Francisco Corona

    Hairo U. Miranda-Belmonte, Victor Muñiz-Sánchez, and Francisco Corona. Word embeddings for topic modeling: An application to the estimation of the economic policy uncertainty index.Expert Systems with Applications, 211: 118499, 2023. ISSN 0957-4174. doi:10.1016/j.eswa.2022.118499. URL https://www.sciencedirect.com/ science/article/pii/S0957417422015822

    work page doi:10.1016/j.eswa.2022.118499 2023

  59. [59]

    Foundation models for time series: A survey, 2025

    Siva Rama Krishna Kottapalli, Karthik Hubli, Sandeep Chandrashekhara, Garima Jain, Sunayana Hubli, Gayathri Botla, and Ramesh Doddaiah. Foundation models for time series: A survey, 2025. URLhttps://arxiv.org/ abs/2504.04011

    work page arXiv 2025

  60. [60]

    Generalisation bounds of zero-shot economic forecasting using time series foundation models.Machine Learning and Knowledge Extraction, 7(4), 2025

    Jittarin Jetwiriyanon, Teo Susnjak, and Surangika Ranathunga. Generalisation bounds of zero-shot economic forecasting using time series foundation models.Machine Learning and Knowledge Extraction, 7(4), 2025. ISSN 2504-4990. doi:10.3390/make7040135. URLhttps://www.mdpi.com/2504-4990/7/4/135

    work page doi:10.3390/make7040135 2025

  61. [61]

    General time transformer: an encoder-only foundation model for zero-shot multivariate time series forecasting

    Cheng Feng, Long Huang, and Denis Krompass. General time transformer: an encoder-only foundation model for zero-shot multivariate time series forecasting. InProceedings of the 33rd ACM International Conference on Information and Knowledge Management, CIKM ’24, page 3757–3761, New York, NY , USA, 2024. Association for Computing Machinery. ISBN 979840070436...

    work page doi:10.1145/3627673.3679931 2024

  62. [62]

    T. Aksu, G. Woo, J. Liu, X. Liu, C. Liu, S. Savarese, C. Xiong, and D. Sahoo. GIFT-Eval: A benchmark for general time series forecasting model evaluation.arXiv preprint arXiv:2410.10393, 2024. URL https: //arxiv.org/abs/2410.10393

    work page arXiv 2024

  63. [63]

    Apress, Berkeley, CA, 2025

    Banglore Vijay Kumar Vishwas and Sri Ram Macharla.TimesFM: Time Series Forecasting Using Decoder-Only Foundation Model, pages 195–210. Apress, Berkeley, CA, 2025. ISBN 979-8-8688-1276-7. doi:10.1007/979-8- 8688-1276-7_8

    work page doi:10.1007/979-8- 2025

  64. [64]

    Chronos-2: From Univariate to Universal Forecasting

    Abdul Fatir Ansari, Oleksandr Shchur, Jaris Küken, Andreas Auer, Boran Han, Pedro Mercado, Syama Sundar Rangapuram, Huibin Shen, Lorenzo Stella, Xiyuan Zhang, Mononito Goswami, Shubham Kapoor, Danielle C. Maddix, Pablo Guerron, Tony Hu, Junming Yin, Nick Erickson, Prateek Mutalik Desai, Hao Wang, Huzefa Rangwala, George Karypis, Yuyang Wang, and Michael B...

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  65. [65]

    Wolff, Shenghao Yang, Kari Torkkola, and Michael W

    Malcolm L. Wolff, Shenghao Yang, Kari Torkkola, and Michael W. Mahoney. Using pre-trained llms for multivariate time series forecasting, 2025. URLhttps://arxiv.org/abs/2501.06386

    work page arXiv 2025

  66. [66]

    Units: A unified multi-task time series model

    Shanghua Gao, Teddy Koker, Owen Queen, Thomas Hartvigsen, Theodoros Tsiligkaridis, and Marinka Zitnik. Units: A unified multi-task time series model. In A. Globerson, L. Mackey, D. Belgrave, A. Fan, U. Paquet, J. Tomczak, and C. Zhang, editors,Advances in Neural Information Processing Systems, volume 37, pages 140589– 140631. Curran Associates, Inc., 2024...

    work page doi:10.52202/079017-4463 2024

  67. [67]

    Transformer-based models for probabilistic time series forecasting with explanatory variables.Mathematics, 13(5), 2025

    Ricardo Caetano, José Manuel Oliveira, and Patrícia Ramos. Transformer-based models for probabilistic time series forecasting with explanatory variables.Mathematics, 13(5), 2025. ISSN 2227-7390. doi:10.3390/math13050814. URLhttps://www.mdpi.com/2227-7390/13/5/814

    work page doi:10.3390/math13050814 2025

  68. [68]

    Medeiros and Henrique F

    Marcelo C. Medeiros and Henrique F. Pires. The proper use of google trends in forecasting models, 2021. URL https://arxiv.org/abs/2104.03065

    work page arXiv 2021

  69. [69]

    Nowcasting growth using google trends data: A bayesian struc- tural time series model.International Journal of Forecasting, 39(3):1384–1412, 2023

    David Kohns and Arnab Bhattacharjee. Nowcasting growth using google trends data: A bayesian struc- tural time series model.International Journal of Forecasting, 39(3):1384–1412, 2023. ISSN 0169-2070. doi:https://doi.org/10.1016/j.ijforecast.2022.05.002. URL https://www.sciencedirect.com/science/ article/pii/S0169207022000620

    work page doi:10.1016/j.ijforecast.2022.05.002 2023

  70. [70]

    Restoring the forecasting power of google trends with statistical preprocessing, 2025

    Candice Djorno, Mauricio Santillana, and Shihao Yang. Restoring the forecasting power of google trends with statistical preprocessing, 2025. URLhttps://arxiv.org/abs/2504.07032

    work page arXiv 2025

  71. [71]

    The m3-competition: Results, conclusions and implications.International Journal of Forecasting, 16(4):451–476, 2000

    Spyros Makridakis and Michèle Hibon. The m3-competition: Results, conclusions and implications.International Journal of Forecasting, 16(4):451–476, 2000. doi:10.1016/S0169-2070(00)00057-1. 24 Zero-Shot Time Series Foundation Models for Annual Institutional Forecasting Under Data SparsityA PREPRINT

    work page doi:10.1016/s0169-2070(00)00057-1 2000

  72. [72]

    Hamilton.Time Series Analysis

    James D. Hamilton.Time Series Analysis. Princeton University Press, 1994

    work page 1994

  73. [73]

    Time series analysis: Forecasting and control, 5th edition , by george e

    Granville Tunnicliffe Wilson. Time series analysis: Forecasting and control, 5th edition , by george e. p. box , gwilym m. jenkins , gregory c. reinsel and greta m. ljung , 2015 . published by john wiley and sons inc. , hoboken, n.Journal of Time Series Analysis, 37(5):709–711, September 2016. URL https://ideas.repec.org/a/ bla/jtsera/v37y2016i5p709-711.html

    work page 2015

  74. [74]

    Peter Kennedy. Forecasting with dynamic regression models: Alan pankratz, 1991, (john wiley and sons, new york), isbn 0-471-61528-5, [uk pound]47.50.International Journal of Forecasting, 8(4):647–648, December 1992. URLhttps://ideas.repec.org/a/eee/intfor/v8y1992i4p647-648.html

    work page 1991

  75. [75]

    Leonard J. Tashman. Out-of-sample tests of forecasting accuracy.International Journal of Forecasting, 16(4): 437–450, 2000. doi:10.1016/S0169-2070(00)00065-0

    work page doi:10.1016/s0169-2070(00)00065-0 2000

  76. [76]

    M., Box G

    Greta M. Ljung and George E. P. Box. On a measure of a lack of fit in time series models.Biometrika, 65(2): 297–303, 1978. doi:10.1093/biomet/65.2.297

    work page doi:10.1093/biomet/65.2.297 1978

  77. [77]

    Another look at measures of forecast accuracy

    Rob J. Hyndman and Anne B. Koehler. Another look at measures of forecast accuracy.International Journal of Forecasting, 22(4):679–688, 2006. ISSN 0169-2070. doi:10.1016/j.ijforecast.2006.03.001. URL https: //www.sciencedirect.com/science/article/pii/S0169207006000239

    work page doi:10.1016/j.ijforecast.2006.03.001 2006

  78. [78]

    Willmott and Kenji Matsuura

    Cort J. Willmott and Kenji Matsuura. Advantages of the mean absolute error (mae) over the root mean square error (rmse) in assessing average model performance.Climate Research, 30(1):79–82, 2005. ISSN 0936577X, 16161572. URLhttp://www.jstor.org/stable/24869236

    work page arXiv 2005

  79. [79]

    Chai and R

    T. Chai and R. R. Draxler. Root mean square error (rmse) or mean absolute error (mae)? – arguments against avoiding rmse in the literature.Geoscientific Model Development, 7(3):1247–1250, 2014. doi:10.5194/gmd-7- 1247-2014. URLhttps://gmd.copernicus.org/articles/7/1247/2014/

    work page doi:10.5194/gmd-7- 2014

  80. [80]

    Proper scoring rules for estimation and forecast evaluation

    Kartik Waghmare and Johanna Ziegel. Proper scoring rules for estimation and forecast evaluation, 2025. URL https://arxiv.org/abs/2504.01781

    work page internal anchor Pith review Pith/arXiv arXiv 2025

Showing first 80 references.