Research On Enterprise Material Safety Stock Prediction Based On PSO-GA Optimization Neural Network

With the continuous development of China’s industrial informatization and the intensification of market competition in the printing industry,the pursuit of lean management of enterprises has become inevitable.The "zero inventory" management model is a part of lean management,and has developed well in enterprise inventory management,but in actual production,especially after three years of epidemic prevention and control,"safety stock" is more suitable for the needs of the printing industry than "zero inventory".Commonly used inventory control methods include ABC classification method,quantitative order method,etc.,and it is difficult to achieve accurate forecasting effect by relying on traditional methods.Therefore,according to the characteristics of printing enterprises,this thesis effectively improves the prediction accuracy of safety stock by improving the prediction algorithm model.Aiming at the problem of material safety inventory setting in printing enterprises,this thesis analyzes and selects seven main factors that affect enterprise safety inventory,including order price,storage cost,material qualification rate,etc.There is a fuzzy and complex nonlinear relationship among these factors,which makes it very difficult for enterprises to predict safety inventory.Because of its excellent characteristics,neural network has been widely used in inventory forecasting,but it still has some shortcomings,such as local minimum problem,which makes the prediction error larger.Therefore,this thesis proposes a hybrid algorithm for optimizing neural networks.The main content of the algorithm is to carry out adaptive cross design and adaptive variation design on the genetic algorithm,and introduce the method of memorizing the best information of the particle’s own motion in the particle swarm(PSO)algorithm,and replace the individual position update of the offspring of the genetic algorithm(GA)with particle position update.Allowing it to remember the sweet spot in its evolution.The optimization ability and convergence speed of genetic algorithm(GA)are improved,and the weight and threshold of BP neural network are optimized by this algorithm.Taking the safety inventory data of material-coated paper of a printing enterprise as sample data,Matlab software was used to construct the BP neural network prediction model,GA-BP neural network prediction model,PSO-BP neural network prediction model and PSO-GA-BP neural network prediction model respectively.Through the simulation experiment comparison of the prediction model,the results showed that,The PSO-GA-BP neural network prediction model has improved the convergence speed and prediction accuracy obviously,and the analysis of the prediction results proves the reliability of the prediction model.Finally,the human-computer interaction interface of enterprise material safety inventory forecasting system based on MATLAB GUI is designed,and the feasibility of the human-computer interaction interface is verified through the analysis of imported sample data,and the prediction work of enterprise material safety inventory is realized.