Pith. sign in

REVIEW 4 major objections 6 minor 79 references

A Novel Hybrid Approach to Contraceptive Demand Forecasting: Integrating Point Predictions with Probabilistic Distributions

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper claims that weighting a machine-learning forecast distribution toward a single human point forecast produces contraceptive demand forecasts that are more accurate and better calibrated than either input alone, and statistically…

desk verdict Useful empirical benchmark, but the hybrid's headline numbers are internally inconsistent with the method as described, and the expert-proxy design leaves the central claim untested. read the letter →

arxiv 2502.09685 v2 pith:TUXBROYV submitted 2025-02-13 cs.LG stat.APstat.ME

Harsha Chamara Hewage , Bahman Rostami-Tabar , Aris Syntetos , Federico Liberatore , Glenn Milano This is my paper · ORCID
classification cs.LGstat.APstat.ME
keywords contraceptivedemandforecastingfamilyplanningsupplychainhybridprobabilisticforecastcombinationquantileregressionaveragingmachinelearningexpertjudgment
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

The paper sets out to fix a gap in family-planning supply chains: forecasts for contraceptive demand are usually single numbers, made with simple methods, and nobody quantifies their uncertainty. It proposes a hybrid scheme that merges a point forecast, the kind a human planner can produce, with a probabilistic forecast distribution from machine learning, so the output is a distribution whose center follows the human input. Tested on 1,360 site-product time series from Côte d'Ivoire with a three-month horizon, the Hybrid Weighted Averaging variant ranks first in mean CRPS (9.868) and second in mean MASE (0.775), statistically indistinguishable from the best Random Forest model. If correct, the result means field planners who only produce single numbers can still feed the system and receive calibrated uncertainty bands, which is what managers need to set stock levels that avoid stockouts.

What carries the argument

The load-bearing object is the CQRA (Constrained Quantile Regression Averaging) combination scheme, adapted from Wang et al. (2018). It takes a probabilistic forecast, extracts quantiles at levels 0.01–0.99, and assigns each quantile a weight $w_i$ solved by linear programming that minimizes pinball loss against observed demand plus the absolute difference between the point forecast and the weighted quantile mean $\bar{y}_t$; a final adjustment factor $adj_t = \text{PointForecast}_t / \bar{y}_t$ rescales the quantiles so the distribution's mean coincides with the point forecast. The variant called Hybrid Weighted Averaging uses this end-to-end; Hybrid Bias Adjustment adds a free bias term $b_t$ and relaxes constraints (weights in $[0,5]$, no sum-to-one constraint), following Granger and Ramanathan (1984). The point forecast itself enters via a demographic formula, modern contraceptive prevalence rate times women population times method mix times couple-years of protection times brand mix times source share, spread over months, which the paper explicitly assumes is a proxy for what a site-level demand planner would judge.

What would settle it

Reproduce the hybrid on the released data and check the reported arithmetic: an equal-weight average of the demographic forecast, with mean MASE of about 16.1, and the ML combined forecast, with mean MASE of about 0.847, cannot under any convex weighting yield the hybrid's reported mean MASE of 0.775, so the construction must differ from the description. The decisive test is to substitute actual forecasts made by Côte d'Ivoire site-level demand planners for the demographic proxy and rerun the cross-validation to see whether the hybrid still matches or beats Random Forest on CRPS.

Watch

Extended reading notes

Core claim

The paper's central claim is that a forecast distribution built by pulling an ML probabilistic forecast toward a human point forecast beats either input used alone, and matches the best pure algorithm on both point and probabilistic accuracy. The mechanism is a Constrained Quantile Regression Averaging (CQRA) scheme in which each quantile of the probabilistic forecast receives an optimized weight, chosen by linear programming to minimize pinball loss plus the absolute gap between the point forecast and the weighted quantile mean, followed by an adjustment factor that sets the distribution's mean exactly on the point forecast. Across the full evaluation, the Hybrid Weighted Averaging variant posts the best mean CRPS (9.868), second-best mean MASE (0.775), and the Nemenyi test shows no significant difference from the top-ranking Random Forest method on either metric. The paper additionally reports that machine-learning methods as a group outperform classical time series, Bayesian, and foundational transformer-based methods in this setting, and that the demographic formula used as the expert proxy is the worst performer of all.

Load-bearing premise

The whole claim that the hybrid blends human expertise with algorithms rests on one assumption the paper states outright: the demographic formula, computed from national family-planning indicators and site-level population, stands in for what real site-level planners would forecast; if that proxy fails, the hybrid's edge over machine learning alone is not evidence about human-algorithm blending.

Editorial extensions

If this is right

  • A site-level planner who can only produce a single judgment number can feed it to the hybrid and receive a full forecast distribution, so probabilistic planning does not require probabilistic training for field staff.
  • Sites using the hybrid would get the calibration of the top ML method plus a distribution whose center reflects local knowledge, directly supporting stockout-risk decisions based on upper prediction bounds.
  • Machine-learning methods (Random Forest, LightGBM, XGBoost) generalize across smooth, erratic, lumpy, and intermittent site-level demand better than classical time series, Bayesian, and foundational methods, which argues for moving beyond moving averages and demographic projections at the site level.
  • Because the hybrid only needs one point forecast and one probabilistic forecast, the same recipe transfers to other humanitarian supply chains with noisy site-level data, such as food aid or medical supplies, as the paper itself notes.

Reading between the lines

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

  • The reported numbers imply an undocumented normalization: averaging the demographic forecast (mean MASE ≈ 16.1) with the ML combined forecast (mean MASE ≈ 0.847) at equal weight cannot produce the hybrid's reported mean MASE of 0.775, so the demographic input must have been rescaled or down-weighted before combination, a step the paper does not describe.
  • Because the demographic formula is the only stand-in for expert judgment, what is actually demonstrated is the mechanics of combining a poor point forecast with a good distribution; the human-algorithm claim needs a field trial with genuine planner forecasts to be fully tested.
  • The smoother right tail of the hybrid's error density at horizon 3, relative to Random Forest and LightGBM, suggests the adjustment step acts as a shrinkage device that reins in extreme quantiles toward the point forecast, a concrete, testable mechanism that would make the hybrid valuable precisely on the erratic series where the pure ML methods overreact.
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.

Referee Report

4 major / 6 minor

Summary. The paper proposes a hybrid forecasting method, Constrained Quantile Regression Averaging (CQRA), that combines point forecasts (provided by a demographic model used as a proxy for expert judgment) with probabilistic forecasts from machine learning and other methods, with the goal of producing calibrated forecast distributions for contraceptive demand at health-site level. Using Côte d'Ivoire LMIS data, the authors benchmark 20 candidate methods, including time-series, Bayesian, ML, foundational models, and two hybrid variants. They report that the Hybrid Weighted Averaging variant achieves the best mean CRPS (9.868) and the second-best mean MASE (0.775), and that a Nemenyi test shows no significant difference from the best Random Forest method. The paper concludes that combining expert point forecasts with probabilistic ML forecasts improves both accuracy and calibration.

Significance. If the hybrid method worked as described, it would address a genuine gap in humanitarian supply-chain forecasting: systematically fusing point forecasts from field experts with probabilistic algorithm outputs. The paper also provides a broad empirical comparison of modern forecasting families on a real site-level dataset, which is a useful contribution. However, the central claim is undermined by a mathematical inconsistency between the described combination pipeline and the reported hybrid performance, and by an unvalidated proxy for expert judgment. The manuscript does not yet provide reproducible evidence for its headline result, so its significance currently rests on an unverifiable implementation detail.

major comments (4)
  1. [Section 4.4.6 / Remark 1 / Table 3] The reported hybrid MASE is inconsistent with the described equal-weight averaging of the demographic forecast and the ML combined forecast. Under Remark 1 and Section 4.4.6, the hybrid point forecast is the simple average of the demographic forecast (mean MASE 16.072, Table 3) and the ML combined forecast (mean MASE 0.847). For any observation, |(e_dem + e_ML)/2| ≥ (|e_dem| − |e_ML|)/2, so the mean MASE of the average must be at least (16.072 − 0.847)/2 ≈ 7.61. Table 3 reports the hybrid's mean MASE as 0.775, which is mathematically impossible for the pipeline as described. The authors must disclose the normalization or other transformation applied to the demographic forecast before it enters the combination, or rerun the experiments with the method stated in the paper.
  2. [Section 4.4] The demographic method is used as a proxy for expert-driven forecasts, but this assumption is not validated. The paper's central claim is that the hybrid method successfully blends human expertise with algorithmic forecasts; however, if the demographic formula—computed from national indicators and site-level population—does not approximate actual site-level planners' judgments, the experiment does not test the claimed human–algorithm fusion. The authors acknowledge the assumption in Section 4.4, but it is absent from the limitations in Section 6.3, and no sensitivity analysis or comparison with real expert forecasts is provided. At minimum, the paper should discuss the conditions under which this proxy is plausible and how violations would affect the conclusions.
  3. [Section 3] The adjustment step in the modified CQRA method is not internally consistent. The final quantiles are defined as tilde_y_t(qi) = w_i * ProbForecast_t(qi) * adj_t, dropping the bias term b_t introduced in step 1 of the modified approach. The mean of this adjusted distribution equals PointForecast_t only when b_t = 0; otherwise, the mean is PointForecast_t * (m_t / (m_t + b_t)), where m_t is the weighted mean of the probabilistic quantiles. The claim that adj_t 'ensures that the mean of the forecast distribution aligns with the point forecast' is therefore not guaranteed. The role of b_t in the final distribution needs to be clarified, or the method description must be amended.
  4. [Section 3 / Tables 3–4] The loss function used to optimize the quantile weights treats PointForecast_t as the observed value in both the absolute-error and pinball-loss terms, and the subsequent rescaling forces the distribution's mean to match that same point forecast. Consequently, the reported CRPS partly measures how well the distribution centers on the point forecast, rather than how well it predicts actual demand. If the point forecast is the demographic–ML average from Major Comment 1, the reported CRPS advantage of the hybrid is unexplained and likely reflects an undeclared normalization. The authors should report the exact point forecast series used in the optimization, the resulting adj_t values, and ideally evaluate the method with an out-of-sample point forecast that is not derived from the same ML quantiles.
minor comments (6)
  1. [Section 4.3.1] Typo: 'bootstraping' should be 'bootstrapping'.
  2. [Data availability statement] The Data availability statement says code/data 'will be made available upon acceptance', but the Introduction (contribution 4) states that code and data are already publicly accessible. Please align these statements.
  3. [Section 4.4.5] The demographic forecasting equation is formatted with unmatched parentheses and is hard to read; please rewrite it cleanly and ensure the notation for the monthly weight w_t is defined consistently.
  4. [Section 4.5] In the CRPS formula, F_t(x) is described as the 'true probability distribution function'; for observations it is the empirical CDF of the realized value. Please correct the terminology.
  5. [Section 5.1 / Figures 5 and 6] The text states that a grey area represents the 95% confidence interval for the top-ranking method, but the figures (as rendered) do not show such a grey area. Please clarify the graphical representation of the Nemenyi intervals.
  6. [Section 3] The total loss L_t sums the constant absolute-error term n times (once per quantile), which is a scaling choice that should be noted explicitly to avoid confusion.

Circularity Check

3 steps flagged · score 6.0 of 10

The hybrid's output mean is the expert point forecast by construction, and its reported MASE cannot be reproduced from the demographic-based pipeline described.

  1. self definitional [Section 3, Modified Approach, step 3]
    "After optimization, we apply an adjustment factor to ensure that the mean of the forecast distribution aligns with the point forecast: ˜y(qi)t = w∗i · ProbForecast(qi)t · adjt. The adjustment factor adjt is defined as: adjt = PointForecastt / ¯yt. These adjustments ensure that the combined probabilistic forecast aligns with the central tendency of the point forecast while still capturing the uncertainty in the prediction."

    The final quantiles are rescaled by PointForecast_t / ȳ_t, so the mean of the reported forecast distribution is the point forecast by construction. The paper's claim that the hybrid 'integrates expert point forecasts with probabilistic distributions' is therefore true by definition: the central tendency of the output is the input point forecast. Any improvement in MASE or central CRPS over the ML-only baseline is inherited from the point forecast rather than being evidence that the combination mechanism adds new information.

  2. fitted input called prediction [Section 3, Proposed hybrid approach, loss function]
    "The key concept in our proposed method is to generate quantiles from a given probabilistic forecast and adjust each quantile using weights. These weights are determined by treating the point forecast as the 'new reality' and formulating a linear programming (LP) problem that minimizes both the pinball loss and the absolute error between point forecast and mean of the weighted quantile forecast."

    The weights w_i are fit by minimizing |PointForecast_t - ȳ_t| plus a pinball loss evaluated with PointForecast_t as the observation. The point forecast is thus used as the target in the fitting procedure, and the subsequent adjustment factor forces the final distribution's mean onto that same target. The hybrid's mean is therefore not a prediction derived from combining two sources of evidence; it is the fitted value of the point forecast input, with the probabilistic forecast only supplying the shape around a pre-imposed center.

1 more flagged steps
  1. other [Section 3, Remark 1; Table 3]
    "Remark 1: When providing point forecasts to the method, they should first be combined with the mean forecasts from the probabilistic forecast using a simple averaging method. This combined point forecast will serve as the new central tendency (e.g., mean or median) for the overall forecast. Table 3: Demographic 16.072; Hybrid weighted averaging 0.775; ML combined 0.847."

    The point forecast entering the hybrid is specified as a simple average of the demographic forecast (mean MASE 16.072) and the ML combined mean (mean MASE 0.847). No equal-weight average of these two components can plausibly produce the hybrid's reported MASE of 0.775, which is even better than the ML-only component, unless the demographic forecast is rescaled, normalized, or effectively discarded before combination. The paper reports no such transformation or error-cancellation analysis. Consequently, the headline 'expert + ML' result is not reproducible from the described pipeline; the evaluation appears to rely on an unstated modification of the expert input, making the central comparison untestable.

full rationale

The central derivation contains a genuine by-construction step: the adjustment factor adj_t = PointForecast_t / ȳ_t rescales every quantile so that the final distribution's mean equals the point forecast, and the weights are fitted with the point forecast treated as the 'new reality' in the loss. Thus the hybrid's central tendency is the input by definition, and the claimed benefit of combining expert point forecasts with probabilistic distributions is partly tautological. The shape of the distribution still carries independent content from the fitted quantile weights, so the paper is not wholly circular. However, the numerical results are inconsistent with the stated pipeline: Remark 1 requires the point forecast to be a simple average of the demographic forecast (MASE 16.072) and the ML combined forecast (MASE 0.847), yet the hybrid reports MASE 0.775, which cannot be obtained from that equal-weight average without an unreported normalization or replacement of the demographic input. This makes the headline MASE/CRPS comparison untraceable. There is no load-bearing self-citation chain or imported uniqueness theorem; the listed self-citations (e.g., Syntetos-Boylan) are standard externally validated benchmarks. Overall, the paper's main claim reduces in part to a fitted input called a prediction, and its reported performance is not derivable from the described method, warranting a score of 6.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The hybrid method depends on a set of fitted weights and bias factors (one per quantile and time point), a hand-chosen weight bound, and a monthly demographic weighting factor. It also rests on domain assumptions about the target variable, the expert proxy, error stationarity, and the calibration and linearity of the probabilistic inputs. No new physical or conceptual entities are introduced.

free parameters (4)
  • quantile weights w_i = optimized per quantile, values not reported
    Weights for each quantile of the probabilistic forecast are optimized in the CQRA loss (Section 3) and change per series and per forecast origin; they are fitted to the point forecast treated as the observed outcome.
  • bias factor b_t = optimized per time point, values not reported
    Introduced in the modified CQRA (Section 3, Modified Approach step 1) and optimized alongside the weights; although it affects adj_t through ybar_t, its fitted values are not reported.
  • upper bound on weights (5) = 5
    Sensitivity analysis showed no improvement beyond 5; chosen by hand and imposed as a constraint 0 <= w_i <= 5 (Section 3).
  • monthly demographic weight w_t = not reported
    Introduced to distribute annual demographic estimates across months in Section 4.4.5; the fitting or calibration procedure is not described, and it materially affects the demographic point forecast that serves as the expert proxy.
assumptions (6)
  • domain assumption Stock distributed is a proxy for consumption at health sites.
    Section 4.1 and footnote 9 state the authors assume stock distributed proxies consumption because stockout cases were removed; if this fails, the target variable is not true demand.
  • domain assumption The demographic forecasting formula proxies expert judgment.
    Section 4.4 states 'we assumed that the demographic-based method serves as a proxy for expert-driven forecasts'; this is load-bearing because the hybrid method's point input is demographic, not real expert input.
  • domain assumption Future forecast errors resemble past errors.
    Section 4.3.1 bootstrap uncertainty construction samples from past errors; if error distributions are non-stationary, the probabilistic forecasts are miscalibrated.
  • ad hoc to paper The probabilistic forecasts are well-calibrated.
    Section 3 lists this as an underlying assumption of CQRA; if violated, the quantile combination can be biased.
  • ad hoc to paper A linear combination of quantiles, point forecast and bias is sufficient.
    Section 3 assumption 4; restricts the hypothesis space of the hybrid distribution.
  • ad hoc to paper Linear interpolation between quantiles reflects the true distribution.
    Section 3 assumption 5; needed to turn discrete quantiles into a continuous forecast distribution for CRPS evaluation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Novel Hybrid Approach to Contraceptive Demand Forecasting: Integrating Point Predictions with Probabilistic Distributions." pith.science (2026). https://pith.science/paper/TUXBROYV

@misc{pith2026250209685,
  author       = {Pith},
  title        = {Pith review of: A Novel Hybrid Approach to Contraceptive Demand Forecasting: Integrating Point Predictions with Probabilistic Distributions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TUXBROYV}},
  note         = {Machine review of arXiv:2502.09685}
}
read the original abstract

Accurate demand forecasting is vital for ensuring reliable access to contraceptive products, supporting key processes like procurement, inventory, and distribution. However, forecasting contraceptive demand in developing countries presents challenges, including incomplete data, poor data quality, and the need to account for multiple geographical and product factors. Current methods often rely on simple forecasting techniques, which fail to capture demand uncertainties arising from these factors, warranting expert involvement. Our study aims to improve contraceptive demand forecasting by combining probabilistic forecasting methods with expert knowledge. We developed a hybrid model that combines point forecasts from domain-specific model with probabilistic distributions from statistical and machine learning approaches, enabling human input to fine-tune and enhance the system-generated forecasts. This approach helps address the uncertainties in demand and is particularly useful in resource-limited settings. We evaluate different forecasting methods, including time series, Bayesian, machine learning, and foundational time series methods alongside our new hybrid approach. By comparing these methods, we provide insights into their strengths, weaknesses, and computational requirements. Our research fills a gap in forecasting contraceptive demand and offers a practical framework that combines algorithmic and human expertise. Our proposed model can also be generalized to other humanitarian contexts with similar data patterns.

Figures

Figures reproduced from arXiv: 2502.09685 by the authors.

Figure 1
Figure 1. Contraceptive stock distribution in Côte d’Ivoire by healthcare site location. The size of the circles represents the quantity of [PITH_FULL_IMAGE:figures/full_fig_p014_1.png] view at source ↗
Figure 2
Figure 2. Time series of contraceptive product stock distributed (Jan 2016 – Dec 2019) at various levels. The x-axis represents the [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗
Figure 3
Figure 3. Time series of contraceptive product stock distributed in selected sites for specific products(Jan 2016 – Dec 2019). To [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Trend strength and seasonality in the time series of stock distribution. Each point in the scatter plot represents one of the 1,360 [PITH_FULL_IMAGE:figures/full_fig_p016_4.png]
Figure 5
Figure 5. Figure 5: Average ranks of forecasting methods with 95% confidence intervals based on the Nemenyi test for MASE values. Lower [PITH_FULL_IMAGE:figures/full_fig_p027_5.png]
Figure 6
Figure 6. Figure 6: Average ranks of forecasting methods with 95% confidence intervals based on the Nemenyi test for CRPS values. Lower [PITH_FULL_IMAGE:figures/full_fig_p029_6.png]
Figure 7
Figure 7. Figure 7: The distribution of MASE values for the top three and bottom three forecasting methods across the horizons is presented. The [PITH_FULL_IMAGE:figures/full_fig_p030_7.png]
Figure 8
Figure 8. Figure 8: The distribution of CRPS values for the top three and bottom three forecasting methods across the horizons is presented. The [PITH_FULL_IMAGE:figures/full_fig_p031_8.png]
Figure 9
Figure 9. Figure 9: Runtime vs. forecast performance (The X-axis shows the runtime speed for each method as fast, moderate, or slow). [PITH_FULL_IMAGE:figures/full_fig_p033_9.png]
Figure 10
Figure 10. Figure 10: The first panel shows the distribution of MASE values for forecasting methods across different horizons. The boxplots are [PITH_FULL_IMAGE:figures/full_fig_p042_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

79 extracted references · 55 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  2. [2]

    , author Choi, Y

    author Ahmed, S. , author Choi, Y. , author Rimon, J.G. , author Alzouma, S. , author Gichangi, P. , author Guiella, G. , author Kayembe, P. , author Kibira, S.P. , author Makumbi, F. , author OlaOlorun, F. , author Omoluabi, E. , author Otupiri, E. , author Oumarou, S. , author Seme, A. , author Shiferaw, S. , author Anglewicz, P. , author Radloff, S. , ...

    work page 2019

  3. [3]

    , author Serumaga, B

    author Akhlaghi, L. , author Serumaga, B. , author Smith, A. , year 2013 . title Botswana: Condom programming; national quantification and supply chain strengthening

    work page 2013

  4. [4]

    , author Narayanan, A

    author Altay, N. , author Narayanan, A. , year 2022 . title Forecasting in humanitarian operations: Literature review and research needs . journal International Journal of Forecasting volume 38 , pages 1234--1244 . https://linkinghub.elsevier.com/retrieve/pii/S0169207020301151, :10.1016/j.ijforecast.2020.08.001

    work page doi:10.1016/j.ijforecast.2020.08.001 2022

  5. [5]

    , author Escobar Ortiz, J

    author Amat Rodrigo, J. , author Escobar Ortiz, J. , year 2023 . title skforecast . publisher Zenodo . https://zenodo.org/record/8382788, :10.5281/ZENODO.8382788. note dOI: 10.5281/ZENODO.8382788

    work page arXiv 2023

  6. [6]

    , author Stella, L

    author Ansari, A.F. , author Stella, L. , author Turkmen, C. , author Zhang, X. , author Mercado, P. , author Shen, H. , author Shchur, O. , author Rangapuram, S.S. , author Arango, S.P. , author Kapoor, S. , author Zschiegner, J. , author Maddix, D.C. , author Wang, H. , author Mahoney, M.W. , author Torkkola, K. , author Wilson, A.G. , author Bohlke-Sch...

    arXiv 2024

  7. [7]

    , author Fahimnia, B

    author Arvan, M. , author Fahimnia, B. , author Reisi, M. , author Siemsen, E. , year 2019 . title Integrating human judgement into quantitative forecasting methods: A review . journal Omega volume 86 , pages 237--252 . https://linkinghub.elsevier.com/retrieve/pii/S0305048317311155, :10.1016/j.omega.2018.07.012

    work page doi:10.1016/j.omega.2018.07.012 2019

  8. [8]

    , author De Baets, S

    author Baecke, P. , author De Baets, S. , author Vanderheyden, K. , year 2017 . title Investigating the added value of integrating human judgement into statistical demand forecasting systems . journal International Journal of Production Economics volume 191 , pages 85--96 . https://linkinghub.elsevier.com/retrieve/pii/S0925527317301652, :10.1016/j.ijpe.20...

    work page doi:10.1016/j.ijpe.2017.05.016 2017

Show all 79 references
  1. [9]

    , author Sedgh, G

    author Baker, D. , author Sedgh, G. , author Keogh, S. , author Luchsinger, G. , author Roseman, M. , author Solo, J. , author Alberto, Y. , author Salvador, E. , year 2022 . title Seeing the Unseen The case for action in the neglected crisis of unintended pregnancy . type Tec...

    2022

  2. [10]

    , author Hewamalage, H

    author Bandara, K. , author Hewamalage, H. , author Liu, Y.H. , author Kang, Y. , author Bergmeir, C. , year 2021 . title Improving the accuracy of global forecasting models using time series data augmentation . journal Pattern Recognition volume 120 , pages 108148 . https://l...

    2021

  3. [11]

    , author Popinchalk, A

    author Bearak, J. , author Popinchalk, A. , author Ganatra, B. , author Moller, A.B. , author Tuncalp , O. , author Beavin, C. , author Kwok, L. , author Alkema, L. , year 2020 . title Unintended pregnancy and abortion by income, region, and the legal status of abortion: estim...

    2020 doi

  4. [12]

    , author Van Wassenhove, L.N

    author Besiou, M. , author Van Wassenhove, L.N. , year 2020 . title Humanitarian operations: A world of opportunity for relevant and impactful research . journal Manufacturing and Service Operations Management volume 22 , pages 135--145 . https://pubsonline.informs.org/doi/10....

    2020

  5. [13]

    , author Goodwin, P

    author Boylan, J.E. , author Goodwin, P. , author Mohammadipour, M. , author Syntetos, A.A. , year 2015 . title Reproducibility in forecasting research . journal International Journal of Forecasting volume 31 , pages 79--90 . https://linkinghub.elsevier.com/retrieve/pii/S01692...

    2015 doi

  6. [14]

    , author Aloysius, J

    author Brau, R. , author Aloysius, J. , author Siemsen, E. , year 2023 . title Demand planning for the digital supply chain: How to integrate human judgment and predictive analytics . journal Journal of Operations Management volume 69 , pages 965--982 . https://onlinelibrary.w...

    2023 doi

  7. [15]

    , year 2001

    author Breiman, L. , year 2001 . title Random forests . journal Machine Learning volume 45 , pages 5--32 . http://link.springer.com/10.1023/A:1010933404324, :10.1023/A:1010933404324

    2001 doi

  8. [16]

    , author Pettenuzzo, D

    author Carriero, A. , author Pettenuzzo, D. , author Shekhar, S. , year 2024 . title Macroeconomic forecasting with large language models . https://arxiv.org/abs/2407.00890, http://arxiv.org/abs/2407.00890 arXiv:2407.00890

    2024

  9. [17]

    , author Cleveland, W.S

    author Cleveland, R.B. , author Cleveland, W.S. , author McRae, J.E. , author Terpenning, I.J. , year 1990 . title STL : A seasonal-trend decomposition procedure based on loess . journal Journal of Official Statistics volume 6 , pages 3--73

    1990

  10. [18]

    , author Fildes, R

    author Davydenko, A. , author Fildes, R. , year 2013 . title Measuring forecasting accuracy: The case of judgmental adjustments to sku-level demand forecasts . journal International Journal of Forecasting volume 29 , pages 510--522 . https://linkinghub.elsevier.com/retrieve/pi...

    2013 doi

  11. [19]

    , author Herlands, W

    author De-Arteaga, M. , author Herlands, W. , author Neill, D.B. , author Dubrawski, A. , year 2018 . title Machine learning for the developing world . journal ACM Transactions on Management Information Systems volume 9 , pages 1--14 . https://dl.acm.org/doi/10.1145/3210548, :...

    2018 doi

  12. [20]

    , author Besiou, M

    author De Boeck, K. , author Besiou, M. , author Decouttere, C. , author Rafter, S. , author Vandaele, N. , author Van Wassenhove, L.N. , author Yadav, P. , year 2023 . title Data, analytical techniques and collaboration between researchers and practitioners in humanitarian he...

    2023 doi

  13. [21]

    , author Van Wassenhove, L.N

    author De Vries, H. , author Van Wassenhove, L.N. , year 2020 . title Do optimization models for humanitarian operations need a paradigm shift? journal Production and Operations Management volume 29 , pages 55--61 . https://journals.sagepub.com/doi/10.1111/poms.13092, :10.1111...

    2020 doi

  14. [22]

    title XGBoost Python Package

    author xgboost Developers , year 2021 . title XGBoost Python Package . https://xgboost.readthedocs.io/en/stable/python/index.html

    2021

  15. [23]

    , author Goodwin, P

    author Fildes, R. , author Goodwin, P. , year 2007 . title Against your better judgment? how organizations can improve their use of management judgment in forecasting . journal Interfaces volume 37 , pages 570--576 . https://pubsonline.informs.org/doi/10.1287/inte.1070.0309, :...

    2007

  16. [24]

    , author Goodwin, P

    author Fildes, R. , author Goodwin, P. , author Lawrence, M. , author Nikolopoulos, K. , year 2009 . title Effective forecasting and judgmental adjustments: an empirical evaluation and strategies for improvement in supply-chain planning . journal International Journal of Forec...

    2009 doi

  17. [25]

    , author Petropoulos, F

    author Fildes, R. , author Petropoulos, F. , year 2015 . title Improving forecast quality in practice . journal Foresight: The International Journal of Applied Forecasting , pages 5--12 https://EconPapers.repec.org/RePEc:for:ijafaa:y:2015:i:36:p:5-12

    2015

  18. [26]

    , author Challu, C

    author Garza, A. , author Challu, C. , author Mergenthaler-Canseco, M. , year 2024 . title Timegpt-1 . https://arxiv.org/abs/2310.03589, http://arxiv.org/abs/2310.03589 arXiv:2310.03589

    2024 arXiv

  19. [27]

    , author Katzfuss, M

    author Gneiting, T. , author Katzfuss, M. , year 2014 . title Probabilistic forecasting . journal Annual Review of Statistics and Its Application volume 1 , pages 125--151 . https://www.annualreviews.org/doi/10.1146/annurev-statistics-062713-085831, :10.1146/annurev-statistics...

    2014 doi

  20. [28]

    , author Bandara, K

    author Godahewa, R. , author Bandara, K. , author Webb, G.I. , author Smyl, S. , author Bergmeir, C. , year 2021 . title Ensembles of localised models for time series forecasting . journal Knowledge-Based Systems volume 233 , pages 107518 . https://linkinghub.elsevier.com/retr...

    2021

  21. [29]

    , year 2000

    author Goodwin, P. , year 2000 . title Correct or combine? mechanically integrating judgmental forecasts with statistical methods . journal International Journal of Forecasting volume 16 , pages 261--275 . https://linkinghub.elsevier.com/retrieve/pii/S0169207000000388, :10.101...

    2000 doi

  22. [30]

    , author Fildes, R

    author Goodwin, P. , author Fildes, R. , year 1999 . title Judgmental forecasts of time series affected by special events: does providing a statistical forecast improve accuracy? journal Journal of Behavioral Decision Making volume 12 , pages 37--53 . https://onlinelibrary.wil...

    1999 doi

  23. [31]

    , author Ramanathan, R

    author Granger, C.W.J. , author Ramanathan, R. , year 1984 . title Improved methods of combining forecasts . journal Journal of Forecasting volume 3 , pages 197--204 . https://onlinelibrary.wiley.com/doi/10.1002/for.3980030207, :10.1002/for.3980030207

    1984 doi

  24. [32]

    , author Angelino, O

    author Haakenstad, A. , author Angelino, O. , author Irvine, C.M.S. , author Bhutta, Z.A. , author Bienhoff, K. , author Bintz, C. , author Causey, K. , author Dirac, M.A. , author Fullman, N. , author Gakidou, E. , author Glucksman, T. , author Hay, S.I. , author Henry, N.J. ...

    2022

  25. [33]

    , author Ewerling, F

    author Hellwig, F. , author Ewerling, F. , author Coll, C.V. , author Barros, A. , year 2022 . title The role of female permanent contraception in meeting the demand for family planning in low- and middle-income countries, . journal Contraception volume 114 , pages 41--48 . ht...

    2022 doi

  26. [34]

    , author Lamberson, P

    author Hong, L. , author Lamberson, P. , author Page, S.E. , year 2021 . title Hybrid predictive ensembles: Synergies between human and computational forecasts . journal Journal of Social Computing volume 2 , pages 89--102 . https://ieeexplore.ieee.org/document/9520815/, :10.2...

    2021

  27. [35]

    , author Athanasopoulos, G

    author Hyndman, R.J. , author Athanasopoulos, G. , year 2021 . title Forecasting: principles and practice . edition 3rd ed., publisher OTexts

    2021

  28. [36]

    , author Wang, Y

    author Januschowski, T. , author Wang, Y. , author Torkkola, K. , author Erkkilä , T. , author Hasson, H. , author Gasthaus, J. , year 2022 . title Forecasting with trees . journal International Journal of Forecasting volume 38 , pages 1473--1481 . https://linkinghub.elsevier....

    2022 doi

  29. [37]

    , author Wheldon, M.C

    author Kantorová , V. , author Wheldon, M.C. , author Dasgupta, A.N.Z. , author Ueffing, P. , author Castanheira, H.C. , year 2021 . title Contraceptive use and needs among adolescent women aged 15 19: Regional and global estimates and projections from 1990 to 2030 from a baye...

    2021 doi

  30. [38]

    , author Mbugua, L

    author Karanja, P. , author Mbugua, L. , author Riungu, J. , author Mulaku, M. , author Okalebo, F. , year 2019 . title Time series analysis of consumption and short term forecasting of female contraceptives in the kenyan public health sector

    2019

  31. [39]

    , author Mishra, A

    author Karimi, A. , author Mishra, A. , author Natarajan, K.V. , author Sinha, K.K. , year 2021 . title Managing commodity stock - outs in public health supply chains in developing countries: An empirical analysis . journal Production and Operations Management volume 30 , page...

    2021 doi

  32. [40]

    , author Grady, B

    author Khan, S. , author Grady, B. , author Tifft, S. , year 2015 . title Estimating demand for a new contraceptive method: Projections for the introduction of sayana press . journal International Journal of Gynecology and Obstetrics volume 130 . https://obgyn.onlinelibrary.wi...

    2015 doi

  33. [41]

    , author Bhattacharjee, A

    author Kohns, D. , author Bhattacharjee, A. , year 2023 . title Nowcasting growth using google trends data: A bayesian structural time series model . journal International Journal of Forecasting volume 39 , pages 1384--1412 . https://linkinghub.elsevier.com/retrieve/pii/S01692...

    2023 doi

  34. [43]

    , author Clark, T.E

    author Krüger , F. , author Clark, T.E. , author Ravazzolo, F. , year 2017 . title Using entropic tilting to combine bvar forecasts with external nowcasts . journal Journal of Business & Economic Statistics volume 35 , pages 470--485 . https://www.tandfonline.com/doi/full/10.1...

    2017

  35. [44]

    , author Jackson, A

    author LaCroix, E. , author Jackson, A. , author McGovern, S. , author Rademacher, K.H. , author Rothschild, C.W. , year 2023 . title Demand forecasting approaches for new contraceptive technologies: A landscape review and recommendations for alignment . journal Global Health:...

    2023 doi

  36. [45]

    , author Spiliotis, E

    author Makridakis, S. , author Spiliotis, E. , author Assimakopoulos, V. , year 2022 . title M5 accuracy competition: Results, findings, and conclusions . journal International Journal of Forecasting volume 38 , pages 1346--1364 . https://linkinghub.elsevier.com/retrieve/pii/S...

    2022 doi

  37. [46]

    , author Frazier, D.T

    author Martin, G.M. , author Frazier, D.T. , author Maneesoonthorn, W. , author Loaiza-Maya, R. , author Huber, F. , author Koop, G. , author Maheu, J. , author Nibbering, D. , author Panagiotelis, A. , year 2024 . title Bayesian forecasting in economics and finance: A modern ...

    2024 doi

  38. [47]

    , author Smith, A

    author Metaxoglou, K. , author Smith, A. , year 2016 . title Option-implied equity premium predictions via entropic tilting . journal SSRN Electronic Journal http://www.ssrn.com/abstract=2754246, :10.2139/ssrn.2754246

    2016 doi

  39. [48]

    title LightGBM

    author Microsoft Corporation , year 2022 . title LightGBM . https://github.com/Microsoft/LightGBM

    2022

  40. [49]

    , author Rostami-Tabar, B

    author Mircetic, D. , author Rostami-Tabar, B. , author Nikolicic, S. , author Maslaric, M. , year 2022 . title Forecasting hierarchical time series in supply chains: an empirical investigation . journal International Journal of Production Research volume 60 , pages 2514--2533...

    2022

  41. [50]

    , author Ali, M

    author Mukasa, B. , author Ali, M. , author Farron, M. , author Van De Weerdt, R. , year 2017 . title Contraception supply chain challenges: a review of evidence from low- and middle-income countries . journal The European Journal of Contraception and Reproductive Health Care ...

    2017

  42. [51]

    , author Cahill, N

    author New, J.R. , author Cahill, N. , author Stover, J. , author Gupta, Y.P. , author Alkema, L. , year 2017 . title Levels and trends in contraceptive prevalence, unmet need, and demand for family planning for 29 states and union territories in india: a modelling study using...

    2017 doi

  43. [52]

    , author Hyndman, R

    author O'Hara-Wild, M. , author Hyndman, R. , author Wang, E. , author Caceres, G. , year 2022 . title fable : Forecasting models for tidy time series . note R package version 0.3.2

    2022

  44. [53]

    , author Varoquaux, G

    author Pedregosa, F. , author Varoquaux, G. , author Gramfort, A. , author Michel, V. , author Thirion, B. , author Grisel, O. , author Blondel, M. , author Prettenhofer, P. , author Weiss, R. , author Dubourg, V. , author Vanderplas, J. , author Passos, A. , author Cournapeau...

    2011

  45. [54]

    , author Hurley, J

    author Perera, H.N. , author Hurley, J. , author Fahimnia, B. , author Reisi, M. , year 2019 . title The human factor in supply chain forecasting: A systematic review . journal European Journal of Operational Research volume 274 , pages 574--600 . https://linkinghub.elsevier.c...

    2019 doi

  46. [55]

    , author Apiletti, D

    author Petropoulos, F. , author Apiletti, D. , author Assimakopoulos, V. , author Babai, M.Z. , author Barrow, D.K. , author Ben Taieb, S. , author Bergmeir, C. , author Bessa, R.J. , author Bijak, J. , author Boylan, J.E. , author Browell, J. , author Carnevale, C. , author C...

    2022

  47. [56]

    , author Kourentzes, N

    author Petropoulos, F. , author Kourentzes, N. , author Nikolopoulos, K. , author Siemsen, E. , year 2018 . title Judgmental selection of forecasting models . journal Journal of Operations Management volume 60 , pages 34--46 . https://onlinelibrary.wiley.com/doi/10.1016/j.jom....

    2018 doi

  48. [57]

    title Pma data

    author PMA , year 2024 . title Pma data . https://www.pmadata.org/. note accessed: 2024-01-10

    2024

  49. [58]

    , year 2016

    author Raftery, A.E. , year 2016 . title Use and communication of probabilistic forecasts . journal Statistical Analysis and Data Mining: The ASA Data Science Journal volume 9 , pages 397--410 . https://onlinelibrary.wiley.com/doi/10.1002/sam.11302, :10.1002/sam.11302

    2016 doi

  50. [59]

    , author Gneiting, T

    author Ranjan, R. , author Gneiting, T. , year 2010 . title Combining Probability Forecasts . journal Journal of the Royal Statistical Society Series B: Statistical Methodology volume 72 , pages 71--91 . https://doi.org/10.1111/j.1467-9868.2009.00726.x, :10.1111/j.1467-9868.20...

    2010

  51. [60]

    , author Ashok, A

    author Rasul, K. , author Ashok, A. , author Williams, A.R. , author Ghonia, H. , author Bhagwatkar, R. , author Khorasani, A. , author Bayazi, M.J.D. , author Adamopoulos, G. , author Riachi, R. , author Hassen, N. , author Biloš, M. , author Garg, S. , author Schneider, A. ,...

    2024 arXiv

  52. [61]

    , author Browell, J

    author Rostami-Tabar, B. , author Browell, J. , author Svetunkov, I. , year 2024 . title Probabilistic forecasting of hourly emergency department arrivals . journal Health Systems volume 13 , pages 133--149 . https://www.tandfonline.com/doi/full/10.1080/20476965.2023.2200526, ...

    2024

  53. [62]

    , author Chou, D

    author Say, L. , author Chou, D. , author Gemmill, A. , author Tuncalp , O. , author Moller, A.B. , author Daniels, J. , author Gulmezoglu , A.M. , author Temmerman, M. , author Alkema, L. , year 2014 . title Global causes of maternal death: a who systematic analysis . journal...

    2014 doi

  54. [63]

    , author Ashford, L.S

    author Sedgh, G. , author Ashford, L.S. , author Hussain, R. , year 2016 . title Unmet need for contraception in developing countries: Examining women s reasons for not using a method

    2016

  55. [64]

    , author Hussain, R

    author Sedgh, G. , author Hussain, R. , year 2014 . title Reasons for contraceptive nonuse among women having unmet need for contraception in developing countries . journal Studies in Family Planning volume 45 , pages 151--169 . https://onlinelibrary.wiley.com/doi/10.1111/j.17...

    2014

  56. [65]

    , author Maddow-Zimet, I

    author Singh, S. , author Maddow-Zimet, I. , year 2016 . title Facility-based treatment for medical complications resulting from unsafe pregnancy termination in the developing world, 2012: a review of evidence from 26 countries . journal BJOG: An International Journal of Obste...

    2016

  57. [66]

    , author Biddlecom, A

    author Sully, E.A. , author Biddlecom, A. , author Darroch, J.E. , author Riley, T. , author Lori S., A. , author Lince-Deroche, N. , author Firestein, L. , author Murro, R. , year 2020 . title Adding it up: Investing in sexual and reproductive health 2019

    2020

  58. [67]

    , year 2023

    author Svetunkov, I. , year 2023 . title Forecasting and Analytics with the Augmented Dynamic Adaptive Model (ADAM) . edition 1 ed., publisher Chapman and Hall/CRC , address Boca Raton . https://www.taylorfrancis.com/books/9781003452652, :10.1201/9781003452652. note dOI: 10.12...

    2023

  59. [68]

    , author Boylan, J.E

    author Syntetos, A.A. , author Boylan, J.E. , year 2005 . title The accuracy of intermittent demand estimates . journal International Journal of Forecasting volume 21 , pages 303--314 . https://www.sciencedirect.com/science/article/pii/S0169207004000792, :https://doi.org/10.10...

    2005 doi

  60. [69]

    , author Boylan, J.E

    author Syntetos, A.A. , author Boylan, J.E. , author Disney, S.M. , year 2009 . title Forecasting for inventory planning: a 50-year review . journal Journal of the Operational Research Society volume 60 , pages S149--S160 . https://www.tandfonline.com/doi/full/10.1057/jors.200...

    2009 doi

  61. [70]

    , author Qian, Y

    author Thompson, R. , author Qian, Y. , author Vasnev, A.L. , year 2024 . title Flexible global forecast combinations . journal Omega volume 126 , pages 103073 . https://linkinghub.elsevier.com/retrieve/pii/S0305048324000409, :10.1016/j.omega.2024.103073

    2024

  62. [71]

    , author Cardós , M

    author Trapero, J.R. , author Cardós , M. , author Kourentzes, N. , year 2019 . title Quantile forecast optimal combination to enhance safety stock estimation . journal International Journal of Forecasting volume 35 , pages 239--250 . https://linkinghub.elsevier.com/retrieve/p...

    2019 doi

  63. [72]

    , author Pedregal, D.J

    author Trapero, J.R. , author Pedregal, D.J. , author Fildes, R. , author Kourentzes, N. , year 2013 . title Analysis of judgmental adjustments in the presence of promotions . journal International Journal of Forecasting volume 29 , pages 234--243 . https://linkinghub.elsevier...

    2013 doi

  64. [73]

    title World Contraceptive Use 2020 (POP/DB/CP/Rev2020)

    author United Nations , year 2021 . title World Contraceptive Use 2020 (POP/DB/CP/Rev2020) . type Technical Report . United Nations (Department of Economic and Social, Population Division

    2021

  65. [74]

    title Contraceptive Forecasting Handbook for Family Planning and HIV/AIDS Prevention Programs

    author USAID , year 2000 . title Contraceptive Forecasting Handbook for Family Planning and HIV/AIDS Prevention Programs . type Technical Report . address Arlington, Va

    2000

  66. [75]

    title Intelligent Forecasting Prize

    author USAID , year 2020 . title Intelligent Forecasting Prize . https://www.usaid.gov/global-health/health-areas/family-planning/usaids-intelligent-forecasting-competition-model-future#: :text=With Intelligent Forecasting

    2020

  67. [76]

    title Couple-years of protection (cyp)

    author USAID , year 2024 a. title Couple-years of protection (cyp) . https://www.usaid.gov/global-health/health-areas/family-planning/couple-years-protection-cyp. note accessed: 2024-01-10

    2024

  68. [77]

    title The demographic and health surveys (dhs) program

    author USAID , year 2024 b. title The demographic and health surveys (dhs) program . https://dhsprogram.com/. note accessed: 2024-01-10

    2024

  69. [78]

    , author Hyndman, R.J

    author Wang, X. , author Hyndman, R.J. , author Li, F. , author Kang, Y. , year 2023 . title Forecast combinations: An over 50-year review . journal International Journal of Forecasting volume 39 , pages 1518--1547 . https://linkinghub.elsevier.com/retrieve/pii/S01692070220014...

    2023 doi

  70. [79]

    , author Zhang, N

    author Wang, Y. , author Zhang, N. , author Tan, Y. , author Hong, T. , author Kirschen, D. , author Kang, C. , year 2018 . title Combining Probabilistic Load Forecasts . http://arxiv.org/abs/1803.06730. note arXiv:1803.06730 [stat]

    2018 arXiv

  71. [80]

    title Worldpop project categories

    author WorldPop , year 2024 . title Worldpop project categories . https://hub.worldpop.org/project/categories?id=3. note accessed: 2024-01-10

    2024

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.