AN INTEGRATED PROCESS-MINING AND EXPLAINABLE MACHINE-LEARNING FRAMEWORK FOR BOTTLENECK AND INTERNAL-CONTROL RISK DETECTION IN THE MULTI-ENTITY FINANCIAL CLOSE
Social Science and Management · 2024
DOI: 10.61784/ssm4095 ↗Kurz gefasst
Fachartikel in der Zeitschrift Social Science and Management, 2024. Verfasst von YaNan Jiao und 1 weiteren Autor:innen. Die Arbeit wurde bisher 3-mal zitiert (Crossref). Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The financial close of a multi-entity group is a recurring, deadline-driven process in which operational congestion and internal-control weaknesses interact, yet the two are normally analysed by separate functions with separate tools. This paper proposes CLOSE-XPM, a framework integrating process mining and explainable machine learning to attribute close-cycle delay to individual record-to-report activities, to flag close instances carrying elevated internal-control risk, and to rank remediation actions against both objectives at once. Two contributions are introduced: a cause-aware Bottleneck Contribution Index (BCI-C) that decomposes every pre-activity delay into resource contention, calendar unavailability and residual, then propagates a contention-truncation counterfactual along a per-case critical path, yielding an estimate in calendar days saved; and a screening procedure that estimates the risk effect of capacity and control actions from the event log alone. Because no public financial-close log exists, we build MEFC-Sim, a resource-aware generator (30 entities, 24 periods, 14,187 events) in which weaknesses are injected as behaviour and the outcome depends on variables the log cannot reveal; paired counterfactual re-runs provide an oracle. BCI-C reaches 0.907 rank correlation with the oracle against 0.544 for the best observable baseline, and predicts the attainable reduction within 0.076 days. Risk detection reaches 0.828 AUC across unseen entities with a 2.86-fold lift in the top-20% inspected sample. Delay and control risk are strongly coupled (rho = 0.708): relieving the three dominant bottlenecks shortens the close by 1.70 days and cuts the deficiency rate by 22.7%. A ground-truth fidelity analysis shows that SHAP is faithful to the model but only moderately faithful to the data-generating process. Validation on a real public log of 1,434 cases confirms transferability and identifies a boundary condition: coupling requires waiting time to be contention-driven.
Zitieren
APA
Jiao, Y., & Liang, Y. (2024). AN INTEGRATED PROCESS-MINING AND EXPLAINABLE MACHINE-LEARNING FRAMEWORK FOR BOTTLENECK AND INTERNAL-CONTROL RISK DETECTION IN THE MULTI-ENTITY FINANCIAL CLOSE. Social Science and Management. https://doi.org/10.61784/ssm4095
Vancouver
Jiao Y, Liang Y. AN INTEGRATED PROCESS-MINING AND EXPLAINABLE MACHINE-LEARNING FRAMEWORK FOR BOTTLENECK AND INTERNAL-CONTROL RISK DETECTION IN THE MULTI-ENTITY FINANCIAL CLOSE. Social Science and Management. 2024. doi:10.61784/ssm4095
BibTeX
@article{Jiao2024,
author = {Jiao, YaNan and Liang, YiWen},
title = {AN INTEGRATED PROCESS-MINING AND EXPLAINABLE MACHINE-LEARNING FRAMEWORK FOR BOTTLENECK AND INTERNAL-CONTROL RISK DETECTION IN THE MULTI-ENTITY FINANCIAL CLOSE},
journal = {Social Science and Management},
year = {2024},
doi = {10.61784/ssm4095}
}
EUVIMED