As we know, due to many parameters

As we know, due to many parameters in the BP neural network, a large number of thresholds and weights need to be updated each time, which will cause the convergence speed to be too slow. At the same time, from a mathematical point of view, the BP algorithm is a faster gradient descent algorithm, which is easy to fall into the problem of local optimal. Therefore there are two ideas to solve this problem. Firstly, the weights and thresholds of the BP neural network can be initialized by the BA algorithm, thereby improving the convergence speed and prediction accuracy of the BP neural network and reducing the possibility that the BP neural network easily falls into a local optimal. Secondly, this paper uses theories and tools such as House of Quality to rank the importance of the factors that affect inventory demand, and selects several major factors thatrank first in importance. After analyzing the House of Quality model based on customer needs, the principal component analysis method is used to select the principal component factors to reduce the input variables of the input layer of the BP neural network, which reduces the training difficulty of BP neural network and improves the training efficiency. This is all the innovations I think.