Кафедри

Постійне посилання на розділhttps://repository.kpi.kharkov.ua/handle/KhPI-Press/35393

Переглянути

Результати пошуку

Зараз показуємо 1 - 1 з 1
  • Ескіз
    Документ
    Methodological approach to predicting producer prices for petroleum products
    (Чернігівський національний технологічний університет, 2018) Posokhov, Igor Mikhailovich; Horenko, Nadezhda Oleksiivna; Chelak, Viktor
    Urgency of the research. Every day, scientists solve problems in economics. To find, which action leads to the expected result with the smallest losses and risks, it’s necessary to predict the further development of events. Target setting. The most widespread problem is the allocation of resources. To make proper calculations and right decisions of distribution, the science of economic theory exists. Actual scientific researches and issues analysis. The studies of Khaikin S. and Callan R. are the most famous among the studies of foreign authors. Yakhyaeva G. E. investigated the theory of neural networks. Matviychuk A. V. suggested a methodical approach to forecasting financial time series with the use of neural networks. Uninvestigated parts of general matters defining. At the moment about 200 methods of estimation are being used, but in practice only a few of them are used. The research objective. The study of each criterion takes a lot of time on preparation of data for the study and careful verification of the original data. For this, it is necessary to choose the correct methodology for developing a forecast to identify the problems to be solved. The statement of basic materials. In this article, the stages of research and prediction are considered of wholesale prices for petroleum products, a methodological approach is proposed in order to evaluate the accuracy of forecasting using neural networks, based on an algorithm with linear partial descriptions of the method of group accounting of the argument. Conclusions. The proposed methodological approach to estimating the accuracy of forecasting using neural networks shows that neural networks allow us to obtain reliable predictions. However, the data on which the training took place had a high degree of similarity among itself, therefore the proposed methodological approach on the one hand does not pretend to be "universal" in forecasting for different sectors of the Ukrainian economy, since different industries have their own characteristics. On the other hand, it can become universal and will allow us to obtain reliable forecasts when taking into account modern features of the development of the Ukrainian economy.