2.2.2. BP Neural Network Model

BP neural networks are multi-layer feedforward neural networks, with characteristics of input forward transmission and error backpropagation. These models are widely used for nonlinear dynamic problems such as regression prediction [43]. BP neural networks are generally composed of an input layer, a hidden layer, and an output layer. After hidden layer, layer-by-layer weighted summation, and the transformation of the transfer function, the data reaches the output layer, and the output value is obtained. The value and the actual value are then compared to calculate the error value, and this error information is back-propagated. These two processes are repeated until the error meets the expectation and the final result is output.