Machine learning and optimization models for supplier selection and order allocation planning

Supplier selection and order allocation have significant roles in supply chain management. These processes become major challenges when the demand is uncertain. This research presents a new two-stage solution approach for supplier selection and order allocation planning where a forecasting procedure is integrated with an optimization model. In the first stage, the demand is forecasted to handle the demand vagueness. A novel Relational Regressor Chain method is introduced to determine the future demand, which is compared with the Holt's Linear Trend and the Auto-Regressive Integrated Moving Average methods to ensure the forecasting accuracy. The forecasted demand is then fed to the second stage where a multi-objective programming model is developed to identify suitable suppliers and order quantities from each supplier. Weighted-sum and epsilon-constraint methods are utilized to obtain the efficient solutions. To our knowledge, this paper is the first study that has integrated demand forecasting with the supplier selection and order allocation planning. A real dataset from a Canadian food supply network is used to examine the results of the forecasting methods and to determine the orders allocated to each supplier. The results of the forecasting methods show that the proposed Relational Regressor Chain method can forecast demand with a higher precision than the other forecasting methods considered in this paper. It is also evident from the results that the selection of the forecasting methods may have impact on both the selection of suppliers and the orders allocated to them.