ANALYSIS OF EXISTING METHODS OF FORECASTING DEMAND AND METHODS OF ASSESSING THEIR QUALITY

Abstract

This text explored various demand forecasting methods, including heuristic, extrapolation, regression, and machine learning methods. Heuristic methods are based on subjective assessments of experts and can be used for long- and medium-term forecasting, as well as in conditions of instability. However, they have their drawbacks, such as subjectivity and the possibility of biasing the forecast. Mathematical methods of extrapolation and regression are based on statistical trends and allow predicting future values based on past data. They have a simple implementation and a clear interpretation of the results, but are also limited by the unreliability of the data and the inability to predict unstable conditions in the future. Machine learning represents an alternative approach to forecasting using large amounts of data. It includes supervised learning, unsupervised learning and deep learning algorithms, each of which has its own characteristics and is used for different types of tasks. It is emphasized that regression models and machine learning methods have their advantages, such as the ability to take into account many variables and factors, but also have difficulty in implementing and interpreting forecast results. The text discusses the main methods of calculating the forecasting error and the scale for assessing the quality of forecasting depending on the value of the error. It is emphasized that the calculation of these indicators assumes a linear development of events, while non-linear events can affect the accuracy of the forecast. For example, disruptions in the work of a supplier or store can lead to an increase in the forecast error. Thus, it is necessary to take into account the non-linear nature of events when assessing the accuracy of the forecast Therefore, the choice of forecasting method depends on the specific situation and the availability of data. Each method has its advantages and limitations, and their use should be justified and carried out taking into account the specifics of a specific forecasting task.