Approximate optimization of resource allocation strategy: the case of bank underwriting system

Consideration is given to the problem of risk assessment process optimization in commercial underwriting. The process is modeled by a dispatching queueing system with multiple single-server queues, several input regenerative flows of jobs with deadlines, complex job structure, two-stage job service process and partial availability of resources. One is interested in maximization of percent of jobs served in time (i. e., within deadline). The paper proposes the optimization method which requires development of the simulation model and further application of adaptive optimization algorithms on simulated paths. Based on numerical examples, it is shown that the approach allows one to find the most efficient strategy in a given set of heuristic strategies. The methodology is also applicable in general cases of resource allocation problems, which are not analytically tractable and which arise frequently when one needs to take into consideration multistage job service process, unpredictability of load and service times, and human factors.