Pith. sign in

REVIEW 2 major objections 4 minor 88 references

Dual Interpretation of Machine Learning Forecasts

T0 review · 2 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Every machine-learning forecast can be rewritten as a weighted combination of past economic episodes, and the paper shows how to compute those weights for ridge regression, forests, boosted trees, and neural networks.

desk verdict Exact for ridge/KRR/RF, approximate for NN — a useful interpretability toolkit with one unsupported 99% claim that needs evidence, not rejection. read the letter →

arxiv 2412.13076 v1 pith:UDYIH352 submitted 2024-12-17 econ.EM cs.LGstat.ML

Philippe Goulet Coulombe , Maximilian Goebel , Karin Klieber This is my paper · ORCID
classification econ.EMcs.LGstat.ML MSC 62J0762H3062M2068T0791B84
keywords machinelearninginterpretabilitydualsolutionkernelridgeregressiondataportfolioweightsmacroeconomicforecastinghistoricalanalogiesforecastdecompositionrandomforest
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 argues that in macroeconomic forecasting, any out-of-sample machine learning prediction can be read as a weighted combination of past values of the target variable, with weights measuring how similar current conditions are to each historical episode. Working through the dual form of ridge regression, the paper derives the exact weights for linear and kernel ridge models, and gives analogous constructions for random forests, boosted trees, and neural networks. These weights turn a dense predictor-based explanation into a sparse time series of historical contributions, which can be summarized with portfolio-inspired diagnostics such as forecast concentration, short position, and leverage. The empirical applications show how such decompositions expose why models failed or succeeded during the post-pandemic inflation surge, the Great Recession, and recent recession-risk assessments.

What carries the argument

The load-bearing object is the data-portfolio weight vector $w_j$, obtained from the dual solution of the regularized least-squares problem rather than from primal coefficients. For ridge-type estimators, $w_j = K_j(K + \lambda I)^{-1}$; the representer theorem guarantees the dual form for kernel methods, and the matrix inversion lemma makes the primal and dual solutions numerically equivalent. For neural networks the paper treats the final linear layer as a ridge regression on the penultimate-layer features $\Psi(X)$, tuning $\lambda$ so that the auxiliary ridge model's out-of-sample predictions replicate the network's. For random forests the weights are the averaged leaf-membership counts, and for boosted trees a recursive accounting scheme allocates each prediction back to the training observations.

What would settle it

Take a trained feed-forward network from the paper's setup (three hidden layers, 400 neurons, ReLU, dropout 0.2, early stopping) and compare its actual out-of-sample predictions to those of the auxiliary ridge regression on the penultimate-layer features with $\lambda$ tuned as described. If the discrepancy exceeds a few percent, or if the ridge predictions cease to track the network when dropout strength or early-stopping tolerance changes, the claimed identity between network forecasts and ridge-representable forecasts is falsified.

Watch

Extended reading notes

Core claim

The paper's central claim is an identity: each out-of-sample forecast can be written as a linear combination of in-sample target values, namely $\hat{y}_j = w_j y$. For (kernel) ridge regression the weight vector is $w_j = K_j(K + \lambda I)^{-1}$, where $K_j$ records pairwise proximity between the test observation's features and each training observation; for neural networks the same formula applies after replacing inputs with penultimate-layer activations, and for forests and boosted trees the weights are assembled by accounting over leaf memberships and tree recursions. The paper shows that this dual representation is available for essentially all major ML families used in macro forecasting and that it often provides a sparser, narrative-friendly interpretation than the usual sum of predictor contributions. Viewing $w_j$ as portfolio weights yields diagnostics that quantify how concentrated, short, leveraged, or unstable a forecast is.

Load-bearing premise

The fragile step is the claim that a neural network with a linear output layer, trained with early stopping and dropout, can be reproduced almost exactly by an auxiliary ridge regression on its penultimate-layer features; the paper states this as an approximation and asserts replication accuracy above 99% without showing the supporting evidence.

Editorial extensions

If this is right

  • For any covered model class, a single forecast can be decomposed into exact contributions from each training quarter, letting an analyst trace a prediction back to specific historical episodes.
  • Linear and kernel ridge forecasts can carry short positions (negative weights), while random forests cannot; the new diagnostics make that structural difference visible and connect it to forecast risk.
  • In the post-pandemic inflation application, models that placed heavy weight on the Great Recession or negative weight on the 1970s produced the largest errors, so concentration and short-position metrics can flag questionable forecasts before the outcome is known.
  • The dual representation extends to classification: recession probabilities are decomposed into contributions from past recessions and expansions, with contributions to log-odds being additive and order-invariant.
  • Because the weights vary with the test observation, the method also produces a time-varying measure of how a model's historical references shift across the business cycle.

Reading between the lines

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

  • If the ridge approximation for neural networks holds in practice, the same weights could be computed in real time to detect when a forecaster starts leaning heavily on one past episode, creating an early-warning signal for narrative-driven forecast instability.
  • The weight vectors $w_j$ define a time-varying similarity map between current and historical states; clustering these maps across models could reveal disagreement among forecasters about which historical regime is most relevant, a useful input to forecast combination.
  • Because $w_j$ obeys the algebra of portfolio weights, one could regularize forecasts directly on properties of $w_j$—concentration, turnover, or short positions—rather than on coefficients, giving an interpretability-driven objective that the paper mentions only as future work.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. The paper proposes a 'dual route' to interpreting ML forecasts in macroeconomics: rewrite each out-of-sample prediction as a weighted sum of in-sample target values, y_hat_j = w_j y, with weights w_j interpreted as pairwise proximity scores between current and past economic events. For ridge and kernel ridge regression the weights are derived as closed-form dual solutions (w_j = K_j(K + lambda I)^{-1}); for random forests the weights are leaf-averaging votes; for boosted trees the paper imports the AXIL recursion; for neural networks with a linear output layer it proposes approximating the network by an auxiliary ridge regression on penultimate-layer features Psi(X), with lambda chosen to replicate the network's out-of-sample predictions. Classification is handled through the dual of (kernel) logistic regression, with contributions reported either in log-odds or probability units. These tools are applied to post-pandemic inflation, Great Recession GDP and unemployment, and post-pandemic GDP and recession probabilities, with weights visualized as historical time series and summarized by concentration, short position, leverage, and turnover.

Significance. The central mathematical identities in Sections 2.1, 2.2, and 2.4 are correct: the dual representation is an exact algebraic identity for ridge and kernel ridge regression, and random forest weights are exact by construction. The proposed portfolio diagnostics (concentration, short position, leverage, turnover) are intuitive and potentially useful. If the neural-network approximation and classification probability treatments can be validated, the paper offers a genuinely useful interpretability tool for macroeconomic forecasting, where the sparsity of the time dimension is an advantage. The empirical applications show that the approach can generate interpretable narratives, e.g., identifying 1970s inflation episodes as drivers of 2022 forecasts. However, the neural-network surrogate step and the probability-level classification contributions are approximations whose accuracy is asserted but not demonstrated; this is the main gap between the manuscript's claims and what is actually shown.

major comments (2)
  1. [Section 2.3.1 (Eqs. 9-12)] The paper's headline claim—'each out-of-sample prediction can also be expressed as a linear combination of in-sample values'—is not established for neural networks. The passage from Eq. (9) to Eq. (10) is explicitly approximate ('~='), and the only justification in Section 2.3.1 is the assertion that 'the replication accuracy is always higher than 99%.' No figure, table, per-forecast error distribution, or code is supplied to support this claim, and the claim is load-bearing: the Section 3 narratives for NN and HNN (e.g., the discussion of NN's 2001Q4 and 2009Q1 weights in Section 3.2) interpret the weights of an auxiliary ridge regression fit to Psi(X), not necessarily the weights of the trained network. Matching aggregate predictions at 99% does not imply weight fidelity, because with a target y that has many near-mean entries, different weight vectors can produce nearly identical forecasts while implying different historical narratives. Moreover, lambda is chosen to match out-of-sample predictions, so the extracted w_j is one of many possible linear representations. The statement in Section 1 that 'for neural networks with a linear output layer... it is possible to obtain the proximity weights exactly' is contradicted by the ~= in Eq. (10). Please provide evidence on per-observation weight fidelity (e.g., sensitivity to lambda, agreement with influence-function or leave-one-out measures) and qualify the abstract and Section 3 claims accordingly.
  2. [Section 2.6 and Figure 8] The treatment of classification is not exact for probability-level contributions. For kernel logistic regression, the exact additive decomposition holds for the log-odds, where logit(P_j) = K_j alpha = sum_i K_ji alpha_i; the quantity c_proba_ji defined in Section 2.6 is a derivative-based local approximation (a first-order marginal effect), and cumulative sums of c_proba_ji do not in general equal the predicted probability. The caption of Figure 8 states that the cumulative sum of contributions 'collectively sum to the final predicted value,' which is not guaranteed for the probability contributions actually plotted. The statement that 'the aforementioned distortions were not overly obstructive in our applications' is unquantified. Please report log-odds contributions, or provide an error bound or empirical distribution for the probability-level approximation, and correct the Figure 8 caption.
minor comments (4)
  1. [Section 2.2] In the sentence introducing Phi(X_i), the notation 'Phi(Xi) in R^{P~ x N} as feature matrix' is inconsistent: Phi(X_i) is a single feature vector for observation i. The displayed polynomial expansion also contains doubled terms and unclear formatting; please fix.
  2. [Section 2.8] The concentration ratio FC is based on absolute weights, so the text '50% of y_j's forecast is driven by just 5% of the observations' should be phrased as '50% of the absolute weight mass'; signed contributions can cancel in the actual forecast.
  3. [Figures 2 and 5] The normalization 'scaled by the mean of absolute weights' used in the right panels of Figures 2 and 5 is not defined in Section 2.8; a sentence explaining the normalization would aid reproducibility.
  4. [Code availability] The paper states that R codes are available on Github but gives no repository URL; please provide a link or appendix reference for replication, especially for the neural-network surrogate construction and the reported 99% replication accuracy.

Circularity Check

1 steps flagged · score 2.0 of 10

Ridge/KRR/RF/boosting decompositions are exact identities or imported external algorithms; the only mild circularity is the neural-network surrogate, whose penalty is tuned to match the network's own predictions, so the extracted weights describe the fitted ridge rather than a unique NN proximity measure.

  1. fitted input called prediction [Section 2.3.1 (Neural Networks), equations (9)-(12)]
    "The natural strategy is to select λ such that out-of-sample predictions from the original network in (9) are closest to those of the auxiliary RR in (10) used for interpretation purposes. In practice, the replication accuracy is always higher than 99%."

    For any λ, equation (10) equals Ψ(X_j)(Ψ'Ψ+λI)^{-1}Ψ'y = w_j y with w_j as in (12). Tuning λ to minimize the distance between (9) and (10) guarantees by construction that the ridge surrogate, and hence the extracted w_j, reproduces the network's out-of-sample predictions. The NN's 'dual representation' is therefore not derived from the network's optimization path; it is the representation of an auxiliary ridge model fitted to the network's outputs. Since many λ values may achieve near-identical predictive accuracy while yielding different w_j narratives, the specific weights are not identified by the network.

full rationale

The core derivations for ridge regression and kernel ridge regression are exact algebraic identities: the primal and dual solutions are linked by the matrix inversion lemma, so y_hat_j = K_j(K+λI)^{-1}y holds without any fitting of the interpretation to the target. Random forest weights are obtained by exact accounting across tree leaves, and boosting weights are taken from the external AXIL algorithm of Geertsema and Lu (2023), not from a self-citation. No uniqueness theorem is imported from the authors' prior work, and the HNN citation is simply a model choice used in the empirical section rather than a load-bearing justification for the dual-representation claim. The only step with a constructed element is the neural-network surrogate in Section 2.3.1, where λ is explicitly chosen to reproduce the network's out-of-sample predictions; this makes the auxiliary ridge's decomposition match the network by construction and means the NN weights describe the fitted surrogate. Because the paper itself acknowledges the approximation with '~=' and the central ridge/KRR/RF/boosting results remain independent, the overall circularity is minor rather than structural, supporting a score of 2.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The derivation of the dual weights for ridge and kernel ridge regression is self-contained and does not introduce free parameters beyond the model's own penalty. The neural network surrogate requires one fitted penalty lambda_NN. The main auxiliary assumptions are the representer theorem, the matrix inversion lemma, the ridge approximation of neural network training, and the correctness of the AXIL boosting algorithm.

free parameters (1)
  • Auxiliary ridge penalty lambda_NN for neural network surrogate = Not reported exactly; described as typically small, selected to maximize replication of the network's out-of-sample…
    Introduced in Section 2.3.1 to make the auxiliary ridge regression reproduce the neural network's predictions; the extracted proximity weights depend on this choice.
assumptions (4)
  • standard math Representer theorem: any minimizer of regularized empirical risk over an RKHS is a finite linear combination of kernel evaluations at training points.
    Invoked in Section 2 to justify expressing kernel ridge regression predictions as K alpha; textbook result.
  • standard math Matrix inversion lemma (Woodbury identity) equates primal and dual ridge solutions.
    Used in Section 2.1 to show beta = X'(XX' + lambda I)^{-1} y.
  • domain assumption For neural networks with a linear output layer, early-stopped gradient descent is well approximated by ridge regression on the penultimate layer features.
    Central to Section 2.3.1; the paper uses '~=' rather than '=' and cites prior theoretical results, but the approximation quality is asserted, not proven for the specific networks used.
  • domain assumption The AXIL algorithm for boosted trees correctly yields instance weights that sum to the prediction.
    Section 2.5 relies on Geertsema and Lu (2023) without reproducing the proof.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Dual Interpretation of Machine Learning Forecasts." pith.science (2026). https://pith.science/paper/UDYIH352

@misc{pith2026241213076,
  author       = {Pith},
  title        = {Pith review of: Dual Interpretation of Machine Learning Forecasts},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UDYIH352}},
  note         = {Machine review of arXiv:2412.13076}
}
read the original abstract

Machine learning predictions are typically interpreted as the sum of contributions of predictors. Yet, each out-of-sample prediction can also be expressed as a linear combination of in-sample values of the predicted variable, with weights corresponding to pairwise proximity scores between current and past economic events. While this dual route leads nowhere in some contexts (e.g., large cross-sectional datasets), it provides sparser interpretations in settings with many regressors and little training data-like macroeconomic forecasting. In this case, the sequence of contributions can be visualized as a time series, allowing analysts to explain predictions as quantifiable combinations of historical analogies. Moreover, the weights can be viewed as those of a data portfolio, inspiring new diagnostic measures such as forecast concentration, short position, and turnover. We show how weights can be retrieved seamlessly for (kernel) ridge regression, random forest, boosted trees, and neural networks. Then, we apply these tools to analyze post-pandemic forecasts of inflation, GDP growth, and recession probabilities. In all cases, the approach opens the black box from a new angle and demonstrates how machine learning models leverage history partly repeating itself.

Figures

Figures reproduced from arXiv: 2412.13076 by the authors.

Figure 9
Figure 9. Moving Average Forecast Contributions for Inflation ( [PITH_FULL_IMAGE:figures/full_fig_p051_9.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

88 extracted references · 65 canonical work pages

  1. [1]

    Adrian, T., Boyarchenko, N., and Giannone, D. (2019). Vulnerable growth. American Economic Review , 109(4):1263--89

    2019

  2. [2]

    Ali, A., Tibshirani, R., and Wainwright, M. J. (2019). A continuous-time view of early stopping for least squares. Journal of Machine Learning Research , PLMR 89:1370--1378

    2019

  3. [3]

    Amburgey, A. J. and McCracken, M. W. (2023). On the real-time predictive content of financial condition indices for growth. Journal of Applied Econometrics , 38(2):137--163

    2023

  4. [4]

    D., Jord \`a , \`O ., and Kuersteiner, G

    Angrist, J. D., Jord \`a , \`O ., and Kuersteiner, G. M. (2018). Semiparametric estimates of monetary policy effects: string theory revisited. Journal of Business & Economic Statistics , 36(3):371--387

    2018

  5. [5]

    Angrist, J. D. and Kuersteiner, G. M. (2011). Causal effects of monetary shocks: Semiparametric conditional independence tests with a multinomial propensity score. Review of Economics and Statistics , 93(3):725--747

    2011

  6. [6]

    Babii, A., Ghysels, E., and Striaukas, J. (2022). Machine learning time series regressions with an application to nowcasting. Journal of Business & Economic Statistics , 40(3):1094--1106

    2022

  7. [7]

    Ball, L., Leigh, D., and Mishra, P. (2022). Understanding us inflation during the covid-19 era. Brookings Papers on Economic Activity , 2022(2):1--80

    2022

  8. [8]

    and Pereverzev, S

    Bauer, F. and Pereverzev, S. V. (2007). Regularization in hilbert spaces: A finite sample analysis with early stopping. SIAM Journal on Numerical Analysis , 43(6):2110--2131

    2007

Show all 88 references
  1. [9]

    and Peersman, G

    Baumeister, C. and Peersman, G. (2013). Time-varying effects of oil supply shocks on the us economy. American Economic Journal: Macroeconomics , 5(4):1--28

    2013

  2. [10]

    Belkin, M., Hsu, D., Ma, S., and Mandal, S. (2019). Reconciling modern machine-learning practice and the classical bias--variance trade-off. Proceedings of the National Academy of Sciences , 116(32):15849--15854

    2019

  3. [11]

    Blanchard, O. (1993). Consumption and the recession of 1990-1991. The American Economic Review , 83(2):270--274

    1993

  4. [12]

    and Hartwig, B

    Bobeica, E. and Hartwig, B. (2023). The covid-19 shock and challenges for inflation modelling. International Journal of Forecasting , 39(1):519--539

    2023

  5. [13]

    and Ng, S

    Boivin, J. and Ng, S. (2006). Are more data always better for factor analysis? Journal of Econometrics , 132(1):169--194

    2006

  6. [14]

    J., M \"u hlbach, N

    Borup, D., Christensen, B. J., M \"u hlbach, N. N., and Nielsen, M. S. (2020). Targeting predictors in random forest regression. Technical report, Department of Economics and Business Economics, Aarhus University

    2020

  7. [15]

    and Vandenberghe, L

    Boyd, S. and Vandenberghe, L. (2004). Convex optimization . Cambridge university press

    2004

  8. [16]

    Brunnermeier, M. K. (2009). Deciphering the liquidity and credit crunch 2007--2008. Journal of Economic Perspectives , 23(1):77--100

    2009

  9. [17]

    and Joseph, A

    Buckmann, M. and Joseph, A. (2023). An interpretable machine learning workflow with an application to economic forecasting. International Journal of Central Banking , 19(4):449--552

    2023

  10. [18]

    Cho, S., Lee, G., Chang, W., and Choi, J. (2020). Interpretation of deep temporal representations by selective visualization of internally activated nodes. arXiv preprint arXiv:2004.12538

    2020 arXiv

  11. [19]

    E., Huber, F., Koop, G., and Marcellino, M

    Clark, T. E., Huber, F., Koop, G., and Marcellino, M. (2024). Forecasting us inflation using bayesian nonparametric models. The Annals of Applied Statistics , 18(2):1421--1444

    2024

  12. [20]

    Dendramis, Y., Kapetanios, G., and Marcellino, M. (2020). A similarity-based approach for macroeconomic forecasting. Journal of the Royal Statistical Society Series A: Statistics in Society , 183(3):801--827

    2020

  13. [21]

    and Doz, C

    Despois, T. and Doz, C. (2023). Identifying and interpreting the factors in factor models via sparsity: Different approaches. Journal of Applied Econometrics , 38(4):533--555

    2023

  14. [22]

    Didisheim, A., Kelly, B., and Malamud, S. (2022). Deep regression ensembles. arXiv preprint arXiv:2203.05417

    2022 arXiv

  15. [23]

    Dominguez, K. M. E. and Shapiro, M. D. (2013). Forecasting the recovery from the great recession: Is this time different? American Economic Review , 103(3):147--152

    2013

  16. [24]

    G., Hall, R

    Fernald, J. G., Hall, R. E., Stock, J. H., and Watson, M. W. (2018). The disappointing recovery in us output after 2009. FRBSF Economic Letter , (4)

    2018

  17. [25]

    Foroni, C., Marcellino, M., and Stevanovic, D. (2022). Forecasting the covid-19 recession and recovery: Lessons from the financial crisis. International Journal of Forecasting , 38(2):596--612

    2022

  18. [26]

    and Popescu, B

    Friedman, J. and Popescu, B. (2004). Gradient directed regularization. Technical Report, Stanford University

    2004

  19. [27]

    and Gertler, M

    Gagliardone, L. and Gertler, M. (2023). Oil prices, monetary policy and inflation surges. NBER Working Paper , (w31263)

    2023

  20. [28]

    and Lu, H

    Geertsema, P. and Lu, H. (2023). Instance-based Explanations for Gradient Boosting Machine Predictions with AXIL Weights . arXiv preprint arXiv:2301.01864

    2023 arXiv

  21. [29]

    and Zou, J

    Ghorbani, A. and Zou, J. (2019). Data shapley: Equitable valuation of data for machine learning. In International Conference on Machine Learning , pages 2242--2251. PMLR

    2019

  22. [30]

    Giannone, D., Lenza, M., and Primiceri, G. E. (2021). Economic predictions with big data: The illusion of sparsity. Econometrica , 89(5):2409--2437

    2021

  23. [31]

    Goodfellow, I., Bengio, Y., Courville, A., and Bengio, Y. (2016). Deep learning , volume 1. MIT press Cambridge

    2016

  24. [32]

    Goulet Coulombe, P. (2024a). The macroeconomy as a random forest. Journal of Applied Econometrics , 39:401--421

    2024

  25. [33]

    Goulet Coulombe, P. (2024b). A neural phillips curve and a deep output gap. Journal of Business & Economic Satistics , (forthcoming)

    2024

  26. [34]

    Goulet Coulombe, P. (2024c). To bag is to prune. Studies in Nonlinear Dynamics & Econometrics , (forthcoming)

    2024

  27. [35]

    Goulet Coulombe, P., Frenette, M., and Klieber, K. (2023a). From reactive to proactive volatility with hemisphere neural networks. arXiv preprint arXiv:2311.16333

    2023 arXiv

  28. [36]

    Goulet Coulombe, P., Leroux, M., Stevanovic, D., and Surprenant, S. (2021). Macroeconomic data transformations matter. International Journal of Forecasting , 37(4):1338--1354

    2021

  29. [37]

    Goulet Coulombe, P., Leroux, M., Stevanovic, D., and Surprenant, S. (2022). How is machine learning useful for macroeconomic forecasting? Journal of Applied Econometrics , 37(5):920--964

    2022

  30. [38]

    Goulet Coulombe, P., Rapach, D., Sch \"u tte, E. C. M., and Schwenk-Nebbe, S. (2023b). The anatomy of machine learning-based portfolio performance. Available at SSRN 4628462

    2023

  31. [39]

    Grinsztajn, L., Oyallon, E., and Varoquaux, G. (2022). Why do tree-based models still outperform deep learning on typical tabular data? Advances in Neural Information Processing Systems , 35:507--520

    2022

  32. [40]

    and Zhong, M

    Guerr \'o on-Quintana, P. and Zhong, M. (2023). Macroeconomic forecasting in times of crises. Journal of Applied Econometrics , 38(3):295--320

    2023

  33. [41]

    Hamilton, J. D. (1983). Oil and the macroeconomy since world war ii. Journal of Political Economy , 91(2):228--248

    1983

  34. [42]

    Hamilton, J. D. (2011). Nonlinearities and the macroeconomic effects of oil prices. Macroeconomic Dynamics , 15(S3):364--378

    2011

  35. [43]

    Harvey, C. R. (1989). Forecasts of Economic Growth from the Bond and Stock Markets . Financial Analysts Journal , 45(5):38--45

    1989

  36. [44]

    Hauzenberger, N., Huber, F., Marcellino, M., and Petz, N. (2024a). Gaussian process vector autoregressions and macroeconomic uncertainty. Journal of Business & Economic Statistics , (forthcoming)

    2024

  37. [45]

    Hauzenberger, N., Marcellino, M., Pfarrhofer, M., and Stelzer, A. (2024b). Nowcasting with mixed frequency data using gaussian processes. arXiv preprint arXiv:2402.10574

    2024 arXiv

  38. [46]

    Hooker, G., Mentch, L., and Zhou, S. (2021). Unrestricted permutation forces extrapolation: variable importance requires at least one more model, or there is no free variable importance. Statistics and Computing , 31:1--16

    2021

  39. [47]

    Huber, F., Koop, G., and Onorante, L. (2021). Inducing sparsity and shrinkage in time-varying parameter models. Journal of Business & Economic Statistics , 39(3):669--683

    2021

  40. [48]

    Jeyapaulraj, J., Desai, D., Mehta, D., Chu, P., Pasquali, S., and Sommer, P. (2022). Supervised similarity learning for corporate bonds using random forest proximities. In Proceedings of the Third ACM International Conference on AI in Finance , pages 411--419

    2022

  41. [49]

    A., Hynes, N., G \"u rel, N

    Jia, R., Dao, D., Wang, B., Hubis, F. A., Hynes, N., G \"u rel, N. M., Li, B., Zhang, C., Song, D., and Spanos, C. J. (2019). Towards efficient data valuation based on the shapley value. In The 22nd International Conference on Artificial Intelligence and Statistics , pages 116...

    2019

  42. [50]

    and Schumacher, C

    Kaufmann, S. and Schumacher, C. (2017). Identifying relevant and irrelevant variables in sparse factor models. Journal of Applied Econometrics , 32(6):1123--1144

    2017

  43. [51]

    Kilian, L. (2008). Exogenous oil supply shocks: how big are they and how much do they matter for the us economy? The Review of Economics and Statistics , 90(2):216--240

    2008

  44. [52]

    and Wahba, G

    Kimeldorf, G. and Wahba, G. (1971). Some results on tchebycheffian spline functions. Journal of mathematical analysis and applications , 33(1):82--95

    1971

  45. [53]

    Klieber, K. (2024). Non-linear dimension reduction in factor-augmented vector autoregressions. Journal of Economic Dynamics and Control , 159:104800

    2024

  46. [54]

    Kliesen, K. L. et al. (2003). The 2001 recession: How was it different and what developments may have caused it? Review-Federal Reserve Bank of Saint Louis , 85(5):23--38

    2003

  47. [55]

    Koh, P. W. and Liang, P. (2017). Understanding black-box predictions via influence functions. In International Conference on Machine Learning , pages 1885--1894. PMLR

    2017

  48. [56]

    and Kr \"u ger, F

    Koster, N. and Kr \"u ger, F. (2024). Simplifying random forests' probabilistic forecasts. arXiv preprint arXiv:2408.12332

    2024 arXiv

  49. [57]

    Kronenberg, P., Daniele, M., and Reinicke, T. (2024). Targeted transformations for macroeconomic forecasting. Available at SSRN 4848879

    2024

  50. [58]

    Kwan, S. H. and Liu, L. (2023). Financial market conditions during monetary tightening. FRBSF Economic Letter , (3)

    2023

  51. [59]

    and Zou, J

    Kwon, Y. and Zou, J. (2023). Data-oob: Out-of-bag estimate as a simple and efficient data value. In International Conference on Machine Learning , pages 18135--18152. PMLR

    2023

  52. [60]

    Li, M., Sarmah, B., Desai, D., Rosaler, J., Bhagat, S., Sommer, P., and Mehta, D. (2024). Quantile regression using random forest proximities. arXiv preprint arXiv:2408.02355

    2024 arXiv

  53. [61]

    and Eck, D

    Lin, J. and Eck, D. J. (2021). Minimizing post-shock forecasting error through aggregation of outside information. International Journal of Forecasting , 37(4):1710--1727

    2021

  54. [62]

    and Jeon, Y

    Lin, Y. and Jeon, Y. (2006). Random forests and adaptive nearest neighbors. Journal of the American Statistical Association , 101(474):578--590

    2006

  55. [63]

    Longo, L., Riccaboni, M., and Rungi, A. (2022). A neural network ensemble approach for gdp forecasting. Journal of Economic Dynamics and Control , 134:104278

    2022

  56. [64]

    Lundberg, S. M. and Lee, S.-I. (2017). A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems , 30

    2017

  57. [65]

    and Eck, D

    Lundquist, D. and Eck, D. (2024). Volatility forecasting using similarity-based parameter correction and aggregated shock information. arXiv preprint arXiv:2406.08738

    2024 arXiv

  58. [66]

    P., and Mart \' nez-Rodr \' guez, A

    Mart \' nez, F., Charte, F., Fr \' as, M. P., and Mart \' nez-Rodr \' guez, A. M. (2022). Strategies for time series forecasting with generalized regression neural networks. Neurocomputing , 491:509--521

    2022

  59. [67]

    and Ng, S

    McCracken, M. and Ng, S. (2020). Fred-qd: A quarterly database for macroeconomic research. NBER Working Paper , (w26872)

    2020

  60. [68]

    McCracken, M. W. and Ng, S. (2016). Fred-md: A monthly database for macroeconomic research. Journal of Business & Economic Statistics , 34(4):574--589

    2016

  61. [69]

    Medeiros, M. C. and Mendes, E. F. (2016). l1-regularization of high-dimensional time-series models with non-gaussian and heteroskedastic errors. Journal of Econometrics , 191(1):255--271

    2016

  62. [70]

    C., Vasconcelos, G

    Medeiros, M. C., Vasconcelos, G. F., Veiga, \'A ., and Zilberman, E. (2021). Forecasting inflation in a data-rich environment: the benefits of machine learning methods. Journal of Business & Economic Statistics , 39(1):98--119

    2021

  63. [71]

    H., Prescott, B., and Sheng, X

    Meyer, B. H., Prescott, B., and Sheng, X. S. (2022). The impact of the covid-19 pandemic on business expectations. International Journal of Forecasting , 38(2):529--544

    2022

  64. [72]

    Pruthi, G., Liu, F., Kale, S., and Sundararajan, M. (2020). Estimating training data influence by tracing gradient descent. Advances in Neural Information Processing Systems , 33:19920--19930

    2020

  65. [73]

    Raihan, T. (2017). Predicting us recessions: A dynamic time warping exercise in economics. Available at SSRN 3047649

    2017

  66. [74]

    J., and Yu, B

    Raskutti, G., Wainwright, M. J., and Yu, B. (2014). Early stopping and non-parametric regression: an optimal data-dependent stopping rule. The Journal of Machine Learning Research , 15(1):335--366

    2014

  67. [75]

    S., Cutler, A., and Moon, K

    Rhodes, J. S., Cutler, A., and Moon, K. R. (2023). Geometry-and accuracy-preserving random forest proximities. IEEE Transactions on Pattern Analysis and Machine Intelligence , 45(9):10947--10959

    2023

  68. [76]

    why should i trust you?

    Ribeiro, M. T., Singh, S., and Guestrin, C. (2016). " why should i trust you?" explaining the predictions of any classifier. In Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining , pages 1135--1144

    2016

  69. [77]

    Rosaler, J., Desai, D., Sarmah, B., Vamvourellis, D., Onay, D., Mehta, D., and Pasquali, S. (2023). Towards enhanced local explainability of random forests: A proximity-based approach. arXiv preprint arXiv:2310.12428

    2023 arXiv

  70. [78]

    Rozemberczki, B., Watson, L., Bayer, P., Yang, H.-T., Kiss, O., Nilsson, S., and Sarkar, R. (2022). The shapley value in machine learning. arXiv preprint arXiv:2202.05594

    2022 arXiv

  71. [79]

    Rudebusch, G. D. and Williams, J. C. (2009). Forecasting recessions: The puzzle of the enduring power of the yield curve. Journal of Business & Economic Statistics , 27(4):492--503

    2009

  72. [80]

    Sahm, C. (2019). Direct stimulus payments to individuals. Recession Ready: Fiscal Policies to Stabilize the American Economy , pages 67--92

    2019

  73. [81]

    Saunders, C., Gammerman, A., and Vovk, V. (1998). Ridge regression learning algorithm in dual variables. In Proceedings of the 15th International Conference on Machine Learning . ICML

    1998

  74. [82]

    and Taskin, F

    Sayek, S. and Taskin, F. (2014). Financial crises: lessons from history for today. Economic Policy , 29(79):447--493

    2014

  75. [83]

    Sch \"o lkopf, B., Herbrich, R., and Smola, A. J. (2001). A Generalized Representer Theorem . In International conference on computational learning theory , pages 416--426. Springer

    2001

  76. [84]

    Sinai, A. (2010). The business cycle in a changing economy: Conceptualization, measurement, dating. American Economic Review , 100(2):25--29

    2010

  77. [85]

    Slack, D., Hilgard, S., Jia, E., Singh, S., and Lakkaraju, H. (2020). Fooling lime and shap: Adversarial attacks on post hoc explanation methods. In Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society , pages 180--186

    2020

  78. [86]

    and Wasserman, L

    Verdinelli, I. and Wasserman, L. (2023). Feature importance: A closer look at shapley values and loco. arXiv preprint arXiv:2303.05981

    2023 arXiv

  79. [87]

    J., Li, F., and Kang, Y

    Wang, X., Hyndman, R. J., Li, F., and Kang, Y. (2023). Forecast combinations: An over 50-year review. International Journal of Forecasting , 39(4):1518--1547

    2023

  80. [88]

    Yao, Y., Rosasco, L., and Caponnetto, A. (2007). Early stopped gradient descent is a biased regularization method. Annals of Statistics , 35(4):1810--1828

    2007

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.