Stratified filtered sampling in stochastic optimization

[摘要] We develop a methodology for evaluating a decision strategygenerated by a stochastic optimization model. The methodology isbased on a pilot study in which we estimate the distribution ofperformance associated with the strategy, and define an appropriatestratified sampling plan. An algorithm we call filtered searchallows us to implement this plan efficiently. We demonstrate theapproach's advantages with a problem in asset / liability managementfor an insurance company.

[学科分类] 自然科学（综合）