References

Koenker, R., and Bassett, G. (1978). Regression quantiles. Econometrica, 46(1), 33–50. doi:10.2307/1913643

Used for: the quantile loss behind the median, 10th and 90th percentile forecasts.

Friedman, J. H. (2001). Greedy function approximation: A gradient boosting machine. The Annals of Statistics, 29(5), 1189–1232. doi:10.1214/aos/1013203451

Used for: gradient boosting, the learning method of each forecasting model.

Ke, G., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., Ye, Q., and Liu, T.-Y. (2017). LightGBM: A highly efficient gradient boosting decision tree. Advances in Neural Information Processing Systems 30 (NeurIPS 2017).

Used for: histogram-based gradient boosting, the approach of the scikit-learn models used here.

Efron, B. (1979). Bootstrap methods: Another look at the jackknife. The Annals of Statistics, 7(1), 1–26. doi:10.1214/aos/1176344552

Used for: the bootstrap. Each ensemble member is trained on a resampled history.

Lakshminarayanan, B., Pritzel, A., and Blundell, C. (2017). Simple and scalable predictive uncertainty estimation using deep ensembles. Advances in Neural Information Processing Systems 30 (NeurIPS 2017).

Used for: ensembles as a practical estimate of predictive uncertainty.

Hüllermeier, E., and Waegeman, W. (2021). Aleatoric and epistemic uncertainty in machine learning: An introduction to concepts and methods. Machine Learning, 110, 457–506. doi:10.1007/s10994-021-05946-3

Used for: splitting total variance into the average variance within ensemble members (day-to-day variation) and the variance between members (limited history).

Der Kiureghian, A., and Ditlevsen, O. (2009). Aleatory or epistemic? Does it matter? Structural Safety, 31(2), 105–112. doi:10.1016/j.strusafe.2008.06.020

Used for: the distinction between uncertainty that more data can reduce (limited history) and uncertainty it cannot (day-to-day variation).

Gneiting, T., Balabdaoui, F., and Raftery, A. E. (2007). Probabilistic forecasts, calibration and sharpness. Journal of the Royal Statistical Society: Series B, 69(2), 243–268. doi:10.1111/j.1467-9868.2007.00587.x

Used for: calibration as the standard for the backtest. The 80% range should contain about 80% of arrivals on days held out from training.

Deb, K., Pratap, A., Agarwal, S., and Meyarivan, T. (2002). A fast and elitist multiobjective genetic algorithm: NSGA-II. IEEE Transactions on Evolutionary Computation, 6(2), 182–197. doi:10.1109/4235.996017

Used for: NSGA-II, the algorithm that searches for staffing plans that trade staff cost against waiting time.

Blank, J., and Deb, K. (2020). pymoo: Multi-objective optimization in Python. IEEE Access, 8, 89497–89509. doi:10.1109/ACCESS.2020.2990567

Used for: the NSGA-II implementation.

Halfin, S., and Whitt, W. (1981). Heavy-traffic limits for queues with many exponential servers. Operations Research, 29(3), 567–588. doi:10.1287/opre.29.3.567

Used for: square-root safety staffing, which seeds the search with good starting plans.

Sakasegawa, H. (1977). An approximation formula L_q ≃ α·ρ^β/(1−ρ). Annals of the Institute of Statistical Mathematics, 29, Part A, 67.

Used for: the approximation of the mean queue length within each hour.

Green, L. V., Kolesar, P. J., and Whitt, W. (2007). Coping with time-varying demand when setting staffing requirements for a service system. Production and Operations Management, 16(1), 13–39. doi:10.1111/j.1937-5956.2007.tb00164.x

Used for: staffing for demand that changes hour by hour, including fluid models for periods of overload. The waiting model carries the backlog from one hour to the next.

Bassamboo, A., Randhawa, R. S., and Zeevi, A. (2010). Capacity sizing under parameter uncertainty: Safety staffing principles revisited. Management Science, 56(10), 1668–1686. doi:10.1287/mnsc.1100.1203

Used for: scoring plans on days sampled from the forecast. When the arrival rate itself is uncertain, square-root staffing alone is not sufficient.

Fildes, R., Goodwin, P., Lawrence, M., and Nikolopoulos, K. (2009). Effective forecasting and judgmental adjustments: An empirical evaluation and strategies for improvement in supply-chain planning. International Journal of Forecasting, 25(1), 3–23. doi:10.1016/j.ijforecast.2008.11.010

Used for: evidence that judgmental adjustments to system forecasts can improve accuracy but also carry biases. Planners' notes therefore become explicit constraints that are reviewed before they take effect.

Liao, Q. V., Gruen, D., and Miller, S. (2020). Questioning the AI: Informing design practices for explainable AI user experiences. Proceedings of CHI 2020, 1–15. doi:10.1145/3313831.3376590

Used for: organizing explanations around the questions users ask. The explanation layer answers why, how sure and what would change it.

Miller, T. (2019). Explanation in artificial intelligence: Insights from the social sciences. Artificial Intelligence, 267, 1–38. doi:10.1016/j.artint.2018.07.007

Used for: explanations as contrastive answers, the basis of the "what would change it" step.

Wachter, S., Mittelstadt, B., and Russell, C. (2018). Counterfactual explanations without opening the black box: Automated decisions and the GDPR. Harvard Journal of Law and Technology, 31(2). arXiv:1711.00399

Used for: counterfactual explanations. Here they are the neighbouring plans on the trade-off and the narrower forecast range without limited history.

Springer, A., and Whittaker, S. (2020). Progressive disclosure: When, why, and how do users want algorithmic transparency information? ACM Transactions on Interactive Intelligent Systems, 10(4), 29:1–29:32. doi:10.1145/3374218

Used for: progressive disclosure. The explanation layer opens on request and shows one step at a time.

Shneiderman, B. (1996). The eyes have it: A task by data type taxonomy for information visualizations. Proceedings of the IEEE Symposium on Visual Languages, 336–343. doi:10.1109/VL.1996.545307

Used for: overview first, details on demand. Each page shows a summary, with explanations behind a button.

Szymanski, M., Millecamp, M., and Verbert, K. (2021). Visual, textual or hybrid: The effect of user expertise on different explanations. Proceedings of IUI 2021, 109–119. doi:10.1145/3397481.3450662

Used for: hybrid explanations. The "why" step shows a chart and sentences together.

Bhatt, U., Antorán, J., Zhang, Y., Liao, Q. V., et al. (2021). Uncertainty as a form of transparency: Measuring, communicating, and using uncertainty. Proceedings of AIES 2021. doi:10.1145/3461702.3462571

Used for: presenting forecast uncertainty as part of the system's transparency, alongside explanations.

van der Bles, A. M., van der Linden, S., Freeman, A., Mitchell, J., Galvao, A., Zaval, L., and Spiegelhalter, D. (2019). Communicating uncertainty about facts, numbers and science. Royal Society Open Science, 6(5), 181870. doi:10.1098/rsos.181870

Used for: guidance on communicating uncertainty in numbers. Both wordings keep the numeric ranges.

Amershi, S., Weld, D., Vorvoreanu, M., et al. (2019). Guidelines for human-AI interaction. Proceedings of CHI 2019.

Used for: design guidelines for human-AI interaction, applied in the confirmation step for changes proposed in Claude.

Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M., and Duchesnay, É. (2011). Scikit-learn: Machine learning in Python. Journal of Machine Learning Research, 12, 2825–2830. jmlr.org

Used for: the forecasting models.

Model Context Protocol. Specification. modelcontextprotocol.io

Used for: the connection between Claude and the server's tools.

MCP Apps: Interactive user interfaces for MCP (SEP-1865). modelcontextprotocol.io

Used for: the panel that shows the charts inside Claude.

Pratama, R., Wicaksono, M. W., and Ramadhan, F. I. (2026). AdviceIT: An instrument for studying explainable AI investment advice (version 2.15.0) [Software]. github.com/Raditya-P/AdviceIT

Used for: the explanation design adapted here, which separates what is explained (why, how sure, what would change it) from how it is shown (chart, text or both).