Mathematical modeling of price forecast in a competitive environment
Abstract
In this article, the research in the areas of competitive environment detection, pricing formation, and pricing methods has been conducted. A competitive environment is emerging in different markets and areas. To identify a competitive environment, we can use the following methods: analytical, expert and simulation. So, the application of basic mathematical models of price forecasting in a competitive environment has been investigated and described. Stage pricing in market conditions is an important element in the development of economic mechanism of production, provides supply and demand balance, influences the interests and needs of society. The basic methods of detection of the competitive environment, the most common methods of pricing and the time series method and the method of neural network forecasting for conducting the price forecast in the competitive environment are considered. It is updated that subjective forecasting of the same number of variables on a regular basis can be very time-consuming. It was determined that there was a way to achieve maximum effect in the field of forecasting with the help of «artificial intelligence», when the computer itself can learn, because by increasing the amount of information resources used in the model, the accuracy of the prediction increases, and the damage associated with uncertainty in decision making, they decrease, and it is possible through the use of neural networks. Finance is extremely non-linear, and sometimes stock price data can even seem completely random. Traditional time series methods, such as the ARIMA and GARCH models, are only effective if the series is stationary, which is a limiting assumption that requires pre-processing the series by receiving journal returns (or other transformations). However, the main problem arises when implementing these models in a real trading system, as there is no guarantee of stationarity when adding new data. They are struggling with it through neural networks that do not require any stationarity. In addition, neural networks are inherently effective in linking data to using it to predict (or classify) new data.
Keywords
