The learning ability and the memory ability of the human brain have the excellent processing performance that present computers can't achieve. The neural network is a technology that reproduces the human brain artificially.
The hierarchical neural network used in this research is composed the input layer, the middle layer and the output layer. The time series has directionality. Therefore, it is used the hierarchical neural network that has directionality in the form of connection. The neural network with the layer which is called an internal memory layer is used in this research.
I make it learn the change in the temperature, and the temperature in the future is forecast by it. The temperature every one hour is input, and the temperature after an hour is forecast.
The input layer is given precipitation, a quantity of solar radiation, and former temperatures. Because it is considered that they influence the change in the temperature. As a result, the error became about 0.71℃. However, according to the result of the detailed forecast, the forecast was made well in the part that changed gradual, but the forecast was not made well in the part that changed greatly. It is considered that one of the causes is the complicated change in the temperature.
Our future task is how to forecast the big change in the temperature.