Estimation of adjacent substitution rate based on clustering algorithm and its application

摘要

Accurate demand forecasting could reduce the uncertainty of inventory and provide theoretical basis for strategic decisions. Without the accurate prediction of actual market demand, there will be a supply shortage or surplus, which influences the enterprise's inventory level and costs of operation. Product substitution is an important factor which affects the precision of demand forecasting. It could reduce the retailer's out-of-stock losses and raise the quality of services. However, it can also distort the genuine demand of the product by exaggerating the demand of substitute products. Product substitution presents a new challenge in demand forecasting. In this paper, an Estimation of Adjacent Substitution Rate based on Clustering Algorithm (EASR-CA) method is proposed according to a cornucopia of categories situation. First, all categories are divided into different clusters by weighted Clustering Algorithm. Then, in each cluster, the product adjacent substitution is calculated. According to the concept mentioned above, a Support Vector Machine (SVM) demand forecasting model, based on adjacent substitution rate estimation, is applied to PC product demand forecasting, which results in higher precision. The experiments identify that the precision is improved and the prediction obtains proper objectivity by considering clustering analysis method.