The aim of the article is to substantiate the feasibility of using adaptive forecasting as an effective management tool for enterprises in conditions of economic uncertainty. The essence of adaptive models is revealed as self-tuning recurrent systems capable of rapidly changing their parameters in response to new information, taking into account the dynamic nature of the processes under study. Adaptive approaches enable the model to quickly respond to market fluctuations, unpredictable changes in demand, seasonal shifts, and other external and internal factors. The scientific approaches to the classification and practical application of adaptive forecasting methods are analyzed, including: exponential smoothing methods (SES, Holt, Winters), adaptive filters (in particular, the Kalman filter), artificial neural networks (RNN, LSTM, GRU), as well as methods for changing model parameters (adaptive ARMA). The article systematizes the advantages, limitations, and risks of applying each of the approaches in the context of managerial decisions. Particular attention is paid to the flexibility of models, their ability to self-learn and adapt, which is critically important for businesses operating in a changing environment. It is substantiated that the combination of adaptive methods with expert evaluations and artificial intelligence tools can ensure increased accuracy of forecasts and quality of strategic planning. Prospects for further research include automating the model selection process, improving adaptation algorithms, and developing interpretable hybrid models for managerial practice.
Iryna Р. Porsiurova (Wed,) studied this question.