FORECASTING BASED ON THE NEWSLETTER THROUGH ASSOCIATIVE RULES
И. А. Черенков · Actual problems of improving of current legislation of Ukraine · 2020
In a dynamically changing market environment, it is relevant to solve the problem of forecasting data presented in the form of time series, in particular, price forecasting. Among the existing methods for solving the problem of price forecasting, the most widespread methods are mathematical statistics, in particular exponential smoothing and autoregressive models. It is typical for these methods that forecasting is carried out on the basis of product price values, while factors affecting the formation of prices are included in the forecast directly through historical price values, which negatively affects the quality of the forecast, since different sets of external and internal factors can lead to the same value of the price. The accuracy of forecasts can be improved both by optimizing the forecast algorithm based on associative rules, and by optimizing the methods for identifying events in ovostnom stream. The superiority of the methods of price forecasting based on news flow through associative rules over regression methods has been experimentally confirmed. It is proved that its use is advisable in those cases when the maximum accuracy of forecasts is necessary, because the total forecast costs, including the formation of many associative rules, are much larger than for the regression methods. Short-term price forecasting methods are considered on the example of the polymer market, which can be applied to the electricity market.