Simulation analytics for virtual statistics via k nearest neighbors

Abstract

“Virtual statistics” are performance measures that are conditional on the occurrence of an event; virtual waiting time of a customer arriving to a queue at time t is one example. In this paper, we describe a k-nearest-neighbor method for estimating virtual statistics post-simulation from the retained sample paths, examining both its small-sample and asymptotic properties. We implement leave-one-replication-out cross validation for tuning the parameter k, and compare the prediction performance of the k-nearest-neighbor estimator with a time-bucket estimator.