Comparative Analysis of Global and Local Probabilistic Time Series Forecasting for Contiguous Spatial Demand Regions

Jiahe Ling; Wei Biao Wu

REVIEW 5 major objections 5 minor 47 references

Comparative Analysis of Global and Local Probabilistic Time Series Forecasting for Contiguous Spatial Demand Regions

T0 review · 5 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A single globally pooled LightGBM model that includes a station identifier outperforms cluster-level and station-level alternatives for probabilistic demand forecasting across synthetic and real bike-share data.

desk verdict A clean but narrowly scoped empirical comparison: global LightGBM with station IDs beats cluster and local fits on Divvy and four synthetic processes, yet the ranking rests on one fixed hyperparameter configuration and the paper contradicts itself on the Divvy MSE result. read the letter →

arxiv 2509.08214 v1 pith:4ZBW6LPU submitted 2025-09-10 stat.AP

Jiahe Ling , Wei Biao Wu This is my paper

classification stat.AP

keywords probabilisticforecastingLightGBMglobaltimeseriesclusteringpredictionintervalsbike-sharedemandspatialquantileregression

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper asks how to produce probabilistic demand forecasts for many spatial locations—here, Chicago bike-share stations—when the stations behave differently. It compares three pooling strategies inside one forecasting engine: a single global model trained on all stations, a middle-ground approach that clusters stations and fits one model per cluster, and fully local station-by-station models. Across four simulated data-generating processes and one real-world year of bike-share data, the paper finds that the global model, when given a station identifier as a categorical covariate, gives the best-calibrated and narrowest prediction intervals and the lowest median forecast error in most settings. Cluster-level pooling performs worst because residual within-cluster heterogeneity hurts both interval width and point accuracy. The practical upshot is that explicit local identity inside a pooled model can beat both extremes of grouping and isolation.

What carries the argument

The load-bearing mechanism is the station identifier as a categorical covariate inside a globally pooled LightGBM quantile-regression model. Instead of deciding a priori whether series are the same or different, the model lets the tree splits decide: most structure is learned from all stations pooled together, while the category splits re-center the quantiles to each station. Clustering is the alternative mechanism that tries to approximate this local adjustment by partitioning stations before fitting, and it is the comparison that exposes the paper's core claim—when clusters retain heterogeneity, they widen intervals and degrade forecasts relative to both full pooling and local fitting.

What would settle it

Run the same three families on a new, completed year of Chicago bike-share data (or on a synthetic battery with per-family hyperparameter optimization) and compare PICP and PINAW. If a cluster-level or station-level model delivers higher coverage with equal or narrower intervals than the global model, the paper's central ordering is refuted; if the global model's coverage falls substantially below 95% in a year with different demand patterns, its calibration is not robust.

Watch

Extended reading notes

Core claim

The central claim is that a fully pooled LightGBM quantile-regression model with the station identifier included as a categorical covariate is the strongest of the three probabilistic forecasting strategies tested. In the real Chicago bike-share data it reaches 0.9885 prediction-interval coverage with a normalized width of about 0.006, and in the four synthetic scenarios its coverage stays within roughly 0.947–0.950, near the nominal 95% level, while achieving the lowest median-forecast MSE in every synthetic case. The paper attributes this to the station identifier letting the single global model adjust the conditional quantiles to each location's baseline while still borrowing strength fro

Load-bearing premise

The ranking is conditional on giving every model family the same fixed LightGBM settings (learning rate 0.05, 64 leaves, 500 iterations, two lag features) and on treating 2024 as exchangeable with 2023; if one family were tuned to its own optimal hyperparameters, or if 2024 conditions were not representative, the global model's lead could shrink or reverse.

Editorial extensions

If this is right

A single global quantile-regression model with an identity covariate can serve as a default for station-level or store-level demand forecasting, replacing hundreds of local fits with one model.
Cluster-based partial pooling should be used with caution: unless clusters are genuinely homogeneous, it can be worse than both full pooling and local models on every metric.
Local station models remain competitive only for point forecasts in highly heterogeneous real data; their prediction intervals can badly under-cover in more homogeneous or short-series settings.
The global model's PINAW advantage on heterogeneous data is partly mechanical (pooling widens the normalization range), so interval sharpness comparisons across pooling schemes should be read with that denominator effect in mind.
The fixed LightGBM configuration means the ranking is conditional on hyperparameters; per-family tuning could alter the ordering.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the mechanism generalizes, the principle extends beyond tree ensembles: encode locality as a feature rather than as a separate model, letting the learner decide which series share parameters.
Because the global model has no volatility-related features, adding station-level exogenous variables such as weather or local events could shrink its intervals further in heteroskedastic regimes—an extension the paper does not test.
The cluster-level failure may be as much about cluster count and quality as about the clustering algorithm itself; a spatially constrained grouping that preserves geographic contiguity could change the middle-ground verdict.
The exchangeability assumption behind the 2023-train/2024-test split (correlation 0.735 between mean hourly profiles) can be stress-tested by training on 2024 and testing on a completed 2025 year or by adding drift terms.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

You should know two things up front. First, this is a legitimate extension of the global-vs-local forecasting literature to probabilistic LightGBM with a station-ID feature, run on genuinely held-out 2024 Divvy data plus four synthetic DGPs. Second, the headline claim—global with station ID wins on coverage, sharpness, and mostly MSE—is supported by the tables under the exact configuration tested, but that configuration is a single fixed hyperparameter setting applied to all model families, and the paper contains a direct internal contradiction about which model wins the MSE comparison on Divvy.

What is actually new: no prior work in the cited literature reports a probabilistic-forecast comparison of global, cluster, and local LightGBM on a contiguous bike-share network with a station-ID covariate. That narrow result is useful guidance for applied demand forecasting. The paper also does several things well: it uses a clean out-of-sample split (2023 train / 2024 test), builds four sensible synthetic processes with parameter selection via AIC/BIC, reports PICP and PINAW alongside point MSE, and explicitly acknowledges in §6 that the fixed hyperparameters and static clustering may not generalize beyond Divvy. That limitation statement is not window dressing—it is the key caveat on the whole enterprise.

The soft spots are proportional. Most importantly, the fixed LightGBM config (learning rate 0.05, 64 leaves, 500 iterations, two lags) is likely to penalize station-level models, which train on only 8,760 rows per station and get 500 boosting rounds. The synthetic PICP gap between station-level (0.83–0.85) and global (0.95) is large, and the paper itself notes that tuning could change the ranking. That makes the global superiority claim load-bearing on the configuration, not a model-family property. Second, the paper says in §5.2 that the station-level model “marginally outperforms” the global on Divvy MSE, while Table 3 shows global MSE 0.82696 and Table 5 shows station mean 0.51728 (median 0.00546). The text and abstract imply global wins, but the tables clearly favor station-level on Divvy. That inconsistency should be fixed before this is citable. Third, there are no error bars on the primary metrics and no code released, so the stability of the ordering across seeds or repeated evaluations is unknown. Fourth, the synthetic data parameters are estimated from the same Divvy series used for the heterogeneous benchmark—a mild circularity, not fatal, but worth stating.

Bottom line: this paper deserves a serious referee, but the referee should ask for a sensitivity analysis with per-family hyperparameter tuning, at least one alternative split or bootstrap-resampled evaluation, and reconciliation of the Divvy MSE numbers. The central finding is plausible and useful, but as written it is a conditional result about one configuration, not a universal ordering of pooling strategies.

Referee Report

5 major / 5 minor

Summary. The paper compares three LightGBM-based probabilistic forecasting strategies—fully pooled global modeling, cluster-level pooling, and station-level local modeling—for hourly demand at 1,782 Chicago Divvy bike-share stations over 2023–2024, plus four synthetic homogeneous data-generating processes (SARIMA, heavy-tailed AR, MLP-AR, AR-GARCH). Models are evaluated on held-out 2024 data using PICP, PINAW, and median-forecast MSE. The headline finding is that a global LightGBM with a station identifier as a categorical covariate achieves the highest PICP, the lowest PINAW, and lowest median MSE in most scenarios, while cluster-level models perform worst and station-level models under-cover severely in synthetic settings. The tabulated results support this ordering within the exact configuration tested, but several load-bearing methodological choices and reporting gaps currently limit the strength of the conclusions.

Significance. If the findings withstand scrutiny, the paper makes a useful empirical contribution to the global-versus-local forecasting debate in a probabilistic setting, with direct relevance to bike-share and mobility demand forecasting. The study has clear strengths: a genuinely held-out 2024 test period, four transparent synthetic DGPs whose parameters are estimated from the Divvy series, consistent use of LightGBM quantile regression, and detailed tabular results. The paper also explicitly acknowledges important limitations. However, the headline ordering is currently conditional on a single untuned LightGBM configuration, a single synthetic realization per DGP, and metric-aggregation choices that are not fully consistent. These issues must be addressed before the comparative claims can be regarded as robust.

major comments (5)

[§4.1, §5.2, §6] All three model families use the same LightGBM configuration (learning rate 0.05, 64 leaves, 500 iterations, two lag features) with no per-family tuning or early stopping. Station-level models fit only ~8,760 hourly observations per station; their synthetic PICP values of 0.834–0.854 (Table 5) are consistent with overfitting rather than with a structural disadvantage of local estimation. Since the paper's headline claim rests on the global model's PICP advantage in the synthetic scenarios, the comparison should include tuned configurations for each family (e.g., fewer trees, larger min_child_samples, validation-based early stopping) or explicitly limit the claim to untuned LightGBM. Section 6's own caveat ('fixed hyperparameter settings ... may not generalize beyond the Divvy context') makes this issue load-bearing.
[§3.4, Tables 3–5] The synthetic evaluation uses one generated realization per DGP, with parameters estimated on the pooled average demand series and station-specific starting values. No description is given of how discrete hourly counts are generated (rounding? integerization?) nor whether multiple Monte Carlo replicates were used. Consequently, the differences among close PICP values (e.g., 0.9469 vs 0.9483 vs 0.9500) and the MSE comparisons are presented without sampling error. Please provide the generation algorithm, state the number of replicates, and report standard errors or confidence intervals for all three metrics; otherwise the synthetic ordering is not statistically assessable.
[§5.2, §5.3, Tables 3 and 5] For Divvy, Table 5 reports station-level median-forecast MSE as mean 0.51728 and median 0.00546, while global MSE is 0.82696. The text calls the station-level model 'marginally' better and later quotes 'approximately 0.005'. If the mean row is the relevant aggregate, 0.517 vs 0.827 is not marginal; if the median is used, the text should say so and justify using the cross-station median rather than the mean. The metric in Eq. (6) is defined over time observations, not over stations; the aggregation level must be clarified and applied consistently.
[§4.3, §5.2, §5.3] PINAW normalizes by the observed range, but the range used differs by construction between global and local models. For the heterogeneous Divvy data, pooling all stations inflates the denominator and automatically lowers global PINAW; the station-level PINAW uses a station-specific denominator. The paper acknowledges this in §5.3 but still concludes that global models 'produce the narrowest prediction intervals' (§5.2). This conclusion is not supported by PINAW alone. Report absolute average interval widths or a sharpness metric independent of the normalizing range before making a narrowest-interval claim.
[§4.2, §5.1] The cluster-level comparison uses a single K selected by KneeLocator on WSS and silhouette cross-check. No sensitivity analysis is reported; the conclusion that clustering 'often suffer[s] from residual within-group heterogeneity' is based on one partition per DGP. Since K=134 for Divvy but K=9 for MLP-AR, the cluster-level results may be driven by the chosen K. Report results for a range of K (e.g., K/2, K, 2K) or otherwise show that the ordinal conclusions are robust to the clustering resolution.

minor comments (5)

[Table 2] The demand-count minimum is reported as 1.0, but §3.1 says zero-demand hours were imputed with zeros; reconcile the summary statistics.
[§5.1] 'read Divvy Bike Data' should read 'real Divvy Bike Data' (typo).
[Eq. (5)] State explicitly whether PINAW is computed per station (with station-specific range) or pooled; the text's comparisons mix both interpretations.
[Tables 4 and 5] The rows labeled 'Mean', 'Med', etc. are cross-cluster/cross-station summaries, while Eq. (6) defines MSEmedian as a single time-series average; add a sentence defining the aggregation.
[§3.4] '17,544 hourly trajectories' equals the combined length of 2023–2024; clarify whether this is 365+366 days and whether leap day is included.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the comparison is a held-out empirical evaluation with independently simulated synthetic data.

full rationale

This paper is an empirical benchmark rather than a derivation, and no step reduces an output to an input by construction. The global, cluster-level, and station-level LightGBM models are trained on 2023 hourly data and evaluated on held-out 2024 data (Sec. 3.5), with PICP, PINAW, and MSE computed from standard definitions (Eqs. 4–6). The synthetic experiments are generated from SARIMA/AR-t/MLP-AR/AR-GARCH parameters estimated on the pooled Divvy average series (Sec. 3.4); these DGPs are not the forecasting models, and the simulated test sequences are not fitted quantities, so the finding that global pooling performs well under homogeneous data is an outcome of the simulation, not a reduction to the fitted parameters. The main caveats—fixed LightGBM hyperparameters (Sec. 4.1) and the range-normalization artifact in PINAW acknowledged in Sec. 5.3—are validity and generalizability concerns, not circularity. There are no self-citations by the authors, and no uniqueness theorem is imported from prior work. The paper's comparative claims therefore have independent empirical content.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The paper introduces no new entities. The main input choices are the fixed LightGBM configuration, the K-means K via WSS knee, the PCA threshold, and the stationarity/exchangeability of 2023-2024. These are modeling choices rather than fitted parameters in the physics sense, but they are load-bearing for the ranking.

free parameters (4)

LightGBM hyperparameters (learning rate 0.05, 64 leaves, 500 iterations) = 0.05 / 64 / 500
Fixed uniformly across model families; no tuning shown, so the comparison may not reflect each family's best achievable performance.
K in K-means clustering = 9, 14, 20, 41, 134 by DGP
Chosen by knee-point detection on WSS and cross-checked via silhouette; it is a fitted modeling choice that determines cluster quality.
PCA threshold (90% explained variance) = about 130 components
Hand-selected threshold that determines the clustering input space.
Station buffer radius (100 m) for CTA covariates = 100 meters
Chosen without sensitivity analysis; affects the near-zero-variance transit covariates.

assumptions (3)

domain assumption 2024 is an exchangeable out-of-sample replicate of 2023 demand patterns
The paper argues via r = 0.735 between 2023 and 2024 mean hourly profiles and visual inspection that no major service changes occurred, then uses 2023 train/2024 test.
domain assumption LightGBM quantile regression at the 2.5th and 97.5th percentiles gives valid empirical prediction intervals without further calibration
Section 4.1 treats the fitted quantiles as the interval directly; no split-sample recalibration or coverage correction is applied.
domain assumption The four fitted DGP templates capture enough of the demand structure to serve as meaningful homogeneous benchmarks
Section 3.4 estimates each DGP once on the pooled average series and generates all station series from that single parameter vector.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Comparative Analysis of Global and Local Probabilistic Time Series Forecasting for Contiguous Spatial Demand Regions." pith.science (2026). https://pith.science/paper/4ZBW6LPU

@misc{pith2026250908214,
  author       = {Pith},
  title        = {Pith review of: Comparative Analysis of Global and Local Probabilistic Time Series Forecasting for Contiguous Spatial Demand Regions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4ZBW6LPU}},
  note         = {Machine review of arXiv:2509.08214}
}

read the original abstract

This study evaluates three probabilistic forecasting strategies using LightGBM: global pooling, cluster-level pooling, and station-level modeling across a range of scenarios, from fully homogeneous simulated data to highly heterogeneous real-world Divvy bike-share demand observed during 2023 to 2024. Clustering was performed using the K-means algorithm applied to principal component analysis transformed covariates, which included time series features, counts of nearby transportation infrastructure, and local demographic characteristics. Forecasting performance was assessed using prediction interval coverage probability (PICP), normalized interval width (PINAW), and the mean squared error (MSE) of the median forecast. The results show that global LightGBM models incorporating station identifiers consistently outperform both cluster-level and station-level models across most scenarios. These global models effectively leverage the full cross-sectional dataset while enabling local adjustments through the station identifier, resulting in superior prediction interval coverage, sharper intervals, and lower forecast errors. In contrast, cluster-based models often suffer from residual within group heterogeneity, leading to degraded accuracy. Station-level models capture fine-grained local dynamics in heterogeneous settings. These findings underscore that global LightGBM models with embedded station identifiers provide a robust, scalable, and computationally efficient framework for transportation demand forecasting. By balancing global structure with local specificity, this approach offers a practical and effective solution for real-world mobility applications.

Figures

Figures reproduced from arXiv: 2509.08214 by the authors.

**Figure 2.** Spatial distribution of Divvy bike-share demand at four peak hours [PITH_FULL_IMAGE:figures/full_fig_p014_2.png] view at source ↗

**Figure 3.** Daily and monthly Divvy bike-share demand time series [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: Spatial clustering comparison of Divvy bike demand time series across DGPs [PITH_FULL_IMAGE:figures/full_fig_p021_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

47 extracted references · 46 canonical work pages

[1]

K. Aas. The Generalized Hyperbolic Skew Student’s t-Distribution.Journal of Financial Econometrics, 4(2):275–309, March 2006

work page 2006
[2]

Time-series clustering – A decade review.Information Systems, 53:16–38, October 2015

Saeed Aghabozorgi, Ali Seyed Shirkhorshidi, and Teh Ying Wah. Time-series clustering – A decade review.Information Systems, 53:16–38, October 2015

work page 2015
[3]

Ali Alqahtani, Mohammed Ali, Xianghua Xie, and Mark W. Jones. Deep Time-Series Clustering: A Review.Electronics, 10(23):3001, December 2021

work page 2021
[4]

Assimakopoulos and K

V. Assimakopoulos and K. Nikolopoulos. The theta model: a decomposition approach to forecasting.International Journal of Forecasting, 16(4):521–530, October 2000

work page 2000
[5]

Kasun Bandara, Christoph Bergmeir, and Slawek Smyl. Forecasting across time series databases using recurrent neural networks on groups of similar series: A clustering approach.Expert Systems with Applications, 140:112896, February 2020

work page 2020
[6]

George E. P. Box, Gwilym M. Jenkins, and Gregory C. Reinsel.Time Series Analysis. Wiley Series in Probability and Statistics. Wiley, 1 edition, June 2008

work page 2008
[7]

Carney, P

J.G. Carney, P. Cunningham, and U. Bhagwan. Confidence and prediction intervals for neural network ensembles. InIJCNN’99. International Joint Conference on Neural Networks. Proceedings (Cat. No.99CH36339), volume 2, pages 1215–1218, Washington, DC, USA, 1999. IEEE

work page 1999
[8]

Model-based clustering with Hidden Markov Model regression for time series with regime changes

Faicel Chamroukhi, Allou Samé, Patrice Aknin, and Gérard Govaert. Model-based clustering with Hidden Markov Model regression for time series with regime changes. In The 2011 International Joint Conference on Neural Networks, pages 2814–2821, July

work page 2011

Show all 47 references

[9]

Calculating Interval Forecasts.Journal of Business & Economic Statistics, 11(2):121–135, April 1993

Chris Chatfield. Calculating Interval Forecasts.Journal of Business & Economic Statistics, 11(2):121–135, April 1993

1993
[10]

XGBoost: A Scalable Tree Boosting System

Tianqi Chen and Carlos Guestrin. XGBoost: A Scalable Tree Boosting System. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 785–794, San Francisco California USA, August 2016. ACM

2016
[11]

Research on the Urban Bike-sharing Usage based on ARIMA Model.Transactions on Computer Science and Intelligent Systems Research, 5:166–172, August 2024

Hao Cheng, Muze Li, and Haoting Zhang. Research on the Urban Bike-sharing Usage based on ARIMA Model.Transactions on Computer Science and Intelligent Systems Research, 5:166–172, August 2024

2024
[12]

ACS 5 Year Data by Ward, February 2025

Chicago Data Portal. ACS 5 Year Data by Ward, February 2025

2025
[13]

Chicago Transit Authority (CTA) Open data, 2022

Chicago Transit Authority. Chicago Transit Authority (CTA) Open data, 2022

2022
[14]

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, September 2018. 28

2018
[15]

Divvy Data, June 2025

Divvy. Divvy Data, June 2025

2025
[16]

Thai, Sanjay Basu, Andrew Y

Tony Duan, Anand Avati, Daisy Yi Ding, Khanh K. Thai, Sanjay Basu, Andrew Y. Ng, and Alejandro Schuler. NGBoost: Natural Gradient Boosting for Probabilistic Prediction, June 2020. arXiv:1910.03225 [cs]

2020 arXiv
[17]

Robert F. Engle. Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of United Kingdom Inflation.Econometrica, 50(4):987, July 1982

1982
[18]

Gao, S.R

J.B. Gao, S.R. Gunn, C.J. Harris, and M. Brown. A Probabilistic Framework for SVM Regression and Error Bar Estimation.Machine Learning, 46(1):71–89, January 2002

2002
[19]

Hashem Pesaran and Takashi Yamagata

M. Hashem Pesaran and Takashi Yamagata. Testing slope homogeneity in large panels. Journal of Econometrics, 142(1):50–93, January 2008

2008
[20]

Global models for time series forecasting: A Simulation study.Pattern Recognition, 124:108441, April 2022

Hansika Hewamalage, Christoph Bergmeir, and Kasun Bandara. Global models for time series forecasting: A Simulation study.Pattern Recognition, 124:108441, April 2022

2022
[21]

Model selection for count timeseries with applications in forecasting number of trips in bike-sharing systems and its volatility, November 2020

Alireza Hosseini and Reza Hosseini. Model selection for count timeseries with applications in forecasting number of trips in bike-sharing systems and its volatility, November 2020. arXiv:2011.08389 [stat]

2020 arXiv
[22]

A state space framework for automatic forecasting using exponential smoothing methods.International Journal of Forecasting, 18(3):439–454, July 2002

Rob J Hyndman, Anne B Koehler, Ralph D Snyder, and Simone Grose. A state space framework for automatic forecasting using exponential smoothing methods.International Journal of Forecasting, 18(3):439–454, July 2002

2002
[23]

Johnson, Miles Q

Alicia A. Johnson, Miles Q. Ott, and Mine Dogucu.Bayes rules! an introduction to Bayesian modeling with R. Chapman & Hall/CRC texts in statistical science. CRC Press, Boca Raton, 2022

2022
[24]

LightGBM: A Highly Efficient Gradient Boosting Decision Tree

Guolin Ke, Qi Meng, Thomas Finley, Taifeng Wang, Wei Chen, Weidong Ma, Qiwei Ye, and Tie-Yan Liu. LightGBM: A Highly Efficient Gradient Boosting Decision Tree. In Advances in Neural Information Processing Systems, volume 30. Curran Associates, Inc., 2017

2017
[25]

Regression Quantiles.Econometrica, 46(1):33, January 1978

Roger Koenker and Gilbert Bassett. Regression Quantiles.Econometrica, 46(1):33, January 1978

1978
[26]

Predict, Refine, Synthesize: Self-Guiding Diffusion Models for Probabilistic Time Series Forecasting, 2023

Marcel Kollovieh, Abdul Fatir Ansari, Michael Bohlke-Schneider, Jasper Zschiegner, Hao Wang, and Yuyang Wang. Predict, Refine, Synthesize: Self-Guiding Diffusion Models for Probabilistic Time Series Forecasting, 2023. Version Number: 3

2023
[27]

Tsoukalas

Dimitrios Kontogiannis, Dimitrios Bargiotas, Aspassia Daskalopulu, Athanasios Ioannis Arvanitidis, and Lefteri H. Tsoukalas. Structural Ensemble Regression for Cluster-Based Aggregate Electricity Demand Forecasting.Electricity, 3(4):480–504, October 2022

2022
[28]

S. Lloyd. Least squares quantization in PCM.IEEE Transactions on Information Theory, 28(2):129–137, March 1982. 29

1982
[29]

Principal components analysis (PCA)

Andrzej Maćkiewicz and Waldemar Ratajczak. Principal components analysis (PCA). Computers & Geosciences, 19(3):303–342, March 1993

1993
[30]

Quantile Regression Forests.Journal of Machine Learning Research, 7(35):983–999, 2006

Nicolai Meinshausen. Quantile Regression Forests.Journal of Machine Learning Research, 7(35):983–999, 2006

2006
[31]

Pablo Montero-Manso and Rob J. Hyndman. Principles and algorithms for forecasting groups of time series: Locality and globality.International Journal of Forecasting, 37(4):1632–1653, October 2021

2021
[32]

Nix and A.S

D.A. Nix and A.S. Weigend. Estimating the mean and variance of the target probability distribution. InProceedings of 1994 IEEE International Conference on Neural Networks (ICNN’94), pages 55–60 vol.1, Orlando, FL, USA, 1994. IEEE

1994
[33]

Note on Regression and Inheritance in the Case of Two Parents.Proceedings of the Royal Society of London, 58:240–242, 1895

Karl Pearson. Note on Regression and Inheritance in the Case of Two Parents.Proceedings of the Royal Society of London, 58:240–242, 1895. Publisher: The Royal Society
[34]

Managing Supply and Demand Balance Through Machine Learning, June 2021

Stas Sajin and Zainab Danish. Managing Supply and Demand Balance Through Machine Learning, June 2021

2021
[35]

Finding a "Kneedle" in a Haystack: Detecting Knee Points in System Behavior

Ville Satopaa, Jeannie Albrecht, David Irwin, and Barath Raghavan. Finding a "Kneedle" in a Haystack: Detecting Knee Points in System Behavior. In2011 31st International Conference on Distributed Computing Systems Workshops, pages 166–171, Minneapolis, MN, USA, June 2011. IEEE

2011
[36]

Think Globally, Act Locally: A Deep Neural Network Approach to High-Dimensional Time Series Forecasting, October 2019

Rajat Sen, Hsiang-Fu Yu, and Inderjit Dhillon. Think Globally, Act Locally: A Deep Neural Network Approach to High-Dimensional Time Series Forecasting, October 2019. arXiv:1905.03806 [stat]

2019 arXiv
[37]

Planning for Bike-sharing System: Predicting Potential Usage with Spatial Regression Models, September 2022

Sai Sun and Lingqian Hu. Planning for Bike-sharing System: Predicting Potential Usage with Spatial Regression Models, September 2022

2022
[38]

Yuru Sun, Worapree Maneesoonthorn, Ruben Loaiza-Maya, and Gael M. Martin. Optimal probabilistic forecasts for risk management, March 2023. arXiv:2303.01651 [q-fin]

2023 arXiv
[39]

Forecasting the Usage of Bike-Sharing Systems through Machine Learning Techniques to Foster Sustainable Urban Mobility.Sustainability, 16(16):6910, August 2024

Jaume Torres, Enrique Jiménez-Meroño, and Francesc Soriguera. Forecasting the Usage of Bike-Sharing Systems through Machine Learning Techniques to Foster Sustainable Urban Mobility.Sustainability, 16(16):6910, August 2024

2024
[40]

Warren Liao

T. Warren Liao. Clustering of time series data—a survey.Pattern Recognition, 38(11):1857–1874, November 2005

2005
[41]

Wellens, Nikolaos Kourentzes, and Maximiliano Udenio

Arnoud P. Wellens, Nikolaos Kourentzes, and Maximiliano Udenio. When and How to Use Global Forecasting Methods on Heterogeneous Datasets.SSRN Electronic Journal, 2023

2023
[42]

Gaussian Processes for Regression

Christopher Williams and Carl Rasmussen. Gaussian Processes for Regression. In D. Touretzky, M. C. Mozer, and M. Hasselmo, editors,Advances in Neural Information Processing Systems, volume 8. MIT Press, 1995. 30

1995
[43]

Understanding the demand predictability of bike share systems: A station-level analysis.Frontiers of Engineering Management, 10(4):551–565, December 2023

Zhuoli Yin, Kendrick Hardaway, Yu Feng, Zhaoyu Kou, and Hua Cai. Understanding the demand predictability of bike share systems: A station-level analysis.Frontiers of Engineering Management, 10(4):551–565, December 2023

2023
[44]

Local vs

Zhao Yingjie and Mahdi Abolghasemi. Local vs. Global Models for Hierarchical Fore- casting, November 2024. arXiv:2411.06394 [cs]

2024 arXiv
[45]

Eddy Patuwo, and Michael Y

Guoqiang Zhang, B. Eddy Patuwo, and Michael Y. Hu. Forecasting with artificial neural networks:.International Journal of Forecasting, 14(1):35–62, March 1998

1998
[46]

Deep and Confident Prediction for Time Series at Uber

Lingxue Zhu and Nikolay Laptev. Deep and Confident Prediction for Time Series at Uber. In2017 IEEE International Conference on Data Mining Workshops (ICDMW), pages 103–110, New Orleans, LA, November 2017. IEEE. 31

2017
[2011]

arXiv:1312.7024 [stat]

arXiv

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.

[1] [1]

K. Aas. The Generalized Hyperbolic Skew Student’s t-Distribution.Journal of Financial Econometrics, 4(2):275–309, March 2006

work page 2006

[2] [2]

Time-series clustering – A decade review.Information Systems, 53:16–38, October 2015

Saeed Aghabozorgi, Ali Seyed Shirkhorshidi, and Teh Ying Wah. Time-series clustering – A decade review.Information Systems, 53:16–38, October 2015

work page 2015

[3] [3]

Ali Alqahtani, Mohammed Ali, Xianghua Xie, and Mark W. Jones. Deep Time-Series Clustering: A Review.Electronics, 10(23):3001, December 2021

work page 2021

[4] [4]

Assimakopoulos and K

V. Assimakopoulos and K. Nikolopoulos. The theta model: a decomposition approach to forecasting.International Journal of Forecasting, 16(4):521–530, October 2000

work page 2000

[5] [5]

Kasun Bandara, Christoph Bergmeir, and Slawek Smyl. Forecasting across time series databases using recurrent neural networks on groups of similar series: A clustering approach.Expert Systems with Applications, 140:112896, February 2020

work page 2020

[6] [6]

George E. P. Box, Gwilym M. Jenkins, and Gregory C. Reinsel.Time Series Analysis. Wiley Series in Probability and Statistics. Wiley, 1 edition, June 2008

work page 2008

[7] [7]

Carney, P

J.G. Carney, P. Cunningham, and U. Bhagwan. Confidence and prediction intervals for neural network ensembles. InIJCNN’99. International Joint Conference on Neural Networks. Proceedings (Cat. No.99CH36339), volume 2, pages 1215–1218, Washington, DC, USA, 1999. IEEE

work page 1999

[8] [8]

Model-based clustering with Hidden Markov Model regression for time series with regime changes

Faicel Chamroukhi, Allou Samé, Patrice Aknin, and Gérard Govaert. Model-based clustering with Hidden Markov Model regression for time series with regime changes. In The 2011 International Joint Conference on Neural Networks, pages 2814–2821, July

work page 2011