PURPOSE: To learn a time sequential pattern by a neural network by using a specific pattern as an input pattern from a context layer to an intermediate layer.
CONSTITUTION: In the time sequential pattern learning system for learning time sequential forecasting processing so that when a learning pattern P(t) at time (t) is an input layer, the output layer of a learning pattern P(t+1) at time (t+1) is outputted by using the neural network having an input layer, an intermediate layer, an output layer, and a context layer and connecting weighted connection lines whose weight is changed by learning between the input layer and the intermediate layer between the context layer and the intermediate layer and between the intermediate layer and the output layer, a pattern obtained by synthesizing the output pattern H(t-1) of the intermediate layer at time (t-1) and the output pattern C(t-1) of the context layer at time (t-1) is determined as an input pattern C(t) from the context layer to the intermediate layer at time (t). Consequently the loss of information relating to the input history of a learning pattern (a) from the output pattern of the context layer can be reduced.