Applying least square support vector machines in efficient consumer response system

The logistics costs in the field of fast moving consumer goods in our country are much higher than those in developed countries, while implementing an efficient consumer response (ECR) system is one of the most effective approaches towards the final solution. To tackle the difficult problem of forecasting the rolling daily sales of one stock keeping unit that is posed in the ECR system, the least square support vector machines are firstly adopted. The least square support vector machine was used in a way that differ from their traditional usage in that corresponding background information and high order reasoning are integrated for the purpose of boosting the prediction performance. The approach presented in the paper is designed for the ECR system between a Shanghai based large dairy corporation and another supermarket chain, and the feasibility of the approach and the performance promotion by integrating background knowledge have been approved by both the experimental simulation and practical running of the ECR system.