Winters' Model in Predicting the Yield of Particular Crops Using Neural Network Technologies

At the practical construction of economic efficiency forecasts of the enterprises' activities in the agro-industrial complex, it is often necessary to take into account the seasonality and cyclically of the initial data. This is especially true for the productivity indicators of different grain crops, which have a stable oscillatory component. In this situation, in order to obtain more accurate forecast estimates, it is necessary to correctly display not only the trend but also the oscillatory component. With a sufficient amount of initial data, the constant seasonal feature fluctuations are traditionally identified using additive seasonal models, and more dynamic changes that depend on the trend are researched on the basis of multiplicative models. In particular, autoregressive and Box-Jenkins models can be applied.