Application of Machine Learning Methods to Forecast Financial Flows in Logistics Systems

Abstract

This article examines the theoretical and practical aspects of applying machine learning methods to forecasting financial flows in logistics systems. It presents the historical evolution of approaches to analyzing and forecasting economic processes, from classical statistical methods to modern machine learning algorithms. Particular attention is paid to the specifics of financial flows in logistics, their relationship with material flows, and the influence of external factors. The main machine learning methods, including linear models, ensemble algorithms, and neural networks, are analyzed for their applicability to forecasting problems. The stages of data preparation, feature selection, and model quality assessment are discussed. The advantages and limitations of using these methods in the face of uncertainty and the complex structure of logistics systems are identified. The feasibility of implementing intelligent methods in financial flow management processes is substantiated, and prospects for their further development and integration with modern digital technologies are demonstrated. The research results can be used to improve the efficiency of resource planning and management in logistics systems. Scientific Novelty. The scientific novelty of this study lies in its integrated approach to the application of machine learning methods to forecasting financial flows in logistics systems, taking into account their systemic interconnectedness and dynamic nature. The paper proposes combining various classes of machine learning algorithms to improve forecasting accuracy under uncertainty and nonlinear dependencies. The need to consider specific logistics factors, including material flow parameters and operating costs, when developing the feature space of models is substantiated. The role of integrating streaming data and digital technologies in improving the adaptability of forecast models is also explored.