PURPOSE:To provide the device which makes the most of a fuzzy logic and a neural network to obtain a forecasted value of arbitrary data after a prescribed time and controls a control object based on this forecasted value at need. CONSTITUTION:An inferred value (y) of a fuzzy logic part 11 and the forecasted value of the error outputted from a neural network 13 are added to correct the inferred value (y) so that it corresponds to practice, thus obtaining the final forecasted value. The control object is controlled based on the final forecasted value at need. The neural network 13 learns relations between input data to the fuzzy logic part 13 and the residual term and learns self-correlations of a series of past errors to forecast the error.
展开▼
