EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Designing a Hybrid Financial Forecasting Model for Optimizing Cash Flow, Budgetary Control, and Resource Allocation in Multinational Enterprises

Carol Kpobie, Ifeoma E. Okoli, Olawunmi Odufuye, Rita Korkor Agyei

Indian journal of Environment & Development · 14.06.2023 · 14 (01) · 200-235

DOI: 10.21590/ijed.14.1.5 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Indian journal of Environment & Development, online erschienen am 14.06.2023. Verfasst von Carol Kpobie und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Multinational enterprises (MNEs) plan liquidity, set budgets and allocate capital across subsidiaries whose cash flows are driven by different currencies, commodity exposures and demand cycles. Classical statistical forecasts capture trend and seasonality well but miss nonlinear shocks such as sharp currency devaluations, whereas machine-learning forecasts capture nonlinearity but are data-hungry and unstable in short corporate series. This paper designs and evaluates a Hybrid Financial Forecasting Model (HFFM) that integrates the two traditions and links forecasting directly to the three decisions it is meant to serve. The forecasting engine combines a seasonal ARIMA model with lagged exogenous drivers (SARIMAX), a gradient-boosted residual corrector and a direct gradient-boosting learner through adaptive inverse-error weights, then translates subsidiary forecasts into the reporting currency, aggregates them bottom-up and attaches split-conformal prediction intervals. Three decision modules sit on top: a rolling-forecast budgetary control layer with operating and translation variance decomposition and an early-warning alert rule; and a chance-constrained linear program that allocates capital expenditure subject to probabilistic liquidity buffers and intragroup transfer limits. The design is illustrated on a calibrated simulation of a three-subsidiary MNE (North America, Europe and West Africa) over 2015 to 2023, with a 24-month rolling-origin hold-out that includes a stylised 2023 devaluation in the West African currency. The HFFM achieved the lowest mean absolute percentage error at group level (2.78% against 2.99% for SARIMA, 3.78% for gradient boosting and 4.35% for a seasonal naive benchmark) and significantly outperformed gradient boosting and the naive benchmark under Diebold-Mariano tests, while remaining statistically indistinguishable from SARIMA on squared error. Replacing the static annual budget with the rolling hybrid forecast cut the 2023 group budget error from 4.18% to 2.79% per month and from 3.5% to 0.7% on the annual total. In allocation, the chance-constrained design reduced the probability of breaching the West African liquidity buffer from 50% to 10% at a cost of 1.7% of portfolio net present value. The paper contributes an integrated architecture, explicit model equations, an evaluation protocol and governance guidance for practitioners, and it reports honestly where the hybrid did not dominate.

Zitieren

APA

Kpobie, C., Okoli, I. E., Odufuye, O., & Agyei, R. K. (2023). Designing a Hybrid Financial Forecasting Model for Optimizing Cash Flow, Budgetary Control, and Resource Allocation in Multinational Enterprises. Indian journal of Environment & Development, 14(01), 200-235. https://doi.org/10.21590/ijed.14.1.5

Vancouver

Kpobie C, Okoli IE, Odufuye O, Agyei RK. Designing a Hybrid Financial Forecasting Model for Optimizing Cash Flow, Budgetary Control, and Resource Allocation in Multinational Enterprises. Indian journal of Environment & Development. 2023;14(01):200-235. doi:10.21590/ijed.14.1.5

BibTeX
@article{Kpobie2023,
  author = {Kpobie, Carol and Okoli, Ifeoma E. and Odufuye, Olawunmi and Agyei, Rita Korkor},
  title = {Designing a Hybrid Financial Forecasting Model for Optimizing Cash Flow, Budgetary Control, and Resource Allocation in Multinational Enterprises},
  journal = {Indian journal of Environment & Development},
  year = {2023},
  volume = {14},
  number = {01},
  pages = {200--235},
  doi = {10.21590/ijed.14.1.5}
}
Kostenlos starten →