Data analytics for slab matching time problem

Producing open order slabs that are not associated with any customer orders is inevitable in the steel production process due to many technical reasons. To consume open order slabs, the most popular way is to allocate them to the existing suitable customer orders that have not been fulfilled. However, for some open order slabs, no suitable customer order is available at present. On this situation, there are two modes to handle these open order slabs. The first one is called waiting mode, that is, open order slabs are stored in the yard until suitable new customer order appears. The second one is called seeking mode, that is, artificial orders (called as self-designed-orders) are designed for open order slabs such that the sale department can seek potential customers according to the attribute of the self-designed-orders. The waiting mode increases inventory holding cost, while the seeking mode decreases income due to final products associated with self-designed-orders will be sold with lower price. To trade off inventory holding cost and income, we need to decide how long each type of open order slabs has to wait before suitable customer order appears. In this paper, a data analytics based method, least squares support vector machine (LSSVM), is presented to forecast the matching time that each type of open order slabs can wait based on the historical data. Moreover, the parameters used by the LSSVM are optimized by a tailored scatter search (SS) algorithm. According to the test results, the proposed method performs efficiency for solving the problem.