Research on forecasting method of mechanical equipment spare parts demand based on LS-SVM

Abstract

In view of the wide variety and quantity of mechanical equipment spare parts, the increasing difficulty of spare parts management, and how to accurately predict the demand for spare parts, the least squares support vector machine (LS-SVM) regression algorithm is proposed to predict the demand for mechanical equipment spare parts. Based on the analysis of the basic principle of least squares support vector machine, a prediction model of mechanical equipment spare parts demand is established. RBF kernel function is selected. LS-SVM is used to study the training samples, train its grid structure parameters, determine the optimal parameters through cross validation and grid search, and use the trained LS-SVM to predict the mechanical equipment spare parts demand, and carry out numerical simulation, The prediction methods such as first-order exponential smoothing, ARMA method and BP neural network are used for comparison. The results show that LS-SVM performs well in the demand forecast of mechanical equipment spare parts.