Adapted skill optimization algorithm to solve constrained discrete stochastic optimization problems

The constrained discrete stochastic optimization problems (CDSOP) have a stochastic objective function and deterministic inequality constraints. The CDSOP is NP-hard due to the large and exponentially growing solution space. The ordinal optimization (OO) is treated as a recognized framework for resolving NP-hard problems. Although the OO framework has been successfully used in many areas, the constraints have significant influence on efficiency and performance. This research develops a metaheuristic algorithm that uses ordinal optimization (OO) in conjunction with skill optimization algorithm (SOA), abbreviated as OSOA, to resolve the CDSOP. The OSOA algorithm has three modules: approximation model, global search, and local search. The regularized minimal-energy tensor-product B-splines is adopted as a performance measure of an alternative in the approximation model. In global search, an adapted skill optimization algorithm is presented to determine N admirable alternatives from the solution space. In local search, an enhanced optimal computing budget allocation is developed to seek a prominent alternative among the N admirable alternatives. The OSOA algorithm is employed to determine the optimal stock levels of a multiple-item inventory system to minimize the expected cost. To validate the performance of the OSOA algorithm, it was compared with five heuristic approaches. The results verify that the OSOA algorithm outperforms the five approaches in both computational efficiency and solution quality.