A Data Mining Perspective of the Newsvendor Problem

This study examined a modified newsvendor problem. The modified newsvendor problem concerned with how wholesalers can achieve maximum profits. Instead of building a profit function for the wholesalers and finding the optimal solutions as previous studies would do, this study proposed a novel, data mining approach to address the problem. Specifically, the modified newsvendor problem were transformed into a classification problem. According to a set of relevant attributes, we used the regularized multiple criteria linear programming (RMCLP) model to classify dealers into two categories, namely Type A and Type B, according to the associated order quantity of dealers. Experiments showed that the RMCLP model gave high accuracy in predicting to which type a dealer belong to. One important implication of this study is to provide insights into the design and development of supply chain coordination policy.