An Agglomerative Clustering of Simulation Output Distributions Using Regularized Wasserstein Distance

TL;DR

This paper introduces a new agglomerative clustering method for simulation output distributions using regularized Wasserstein distance, aiding in pattern detection, anomaly identification, and decision-making in stochastic systems.

Contribution

The paper proposes a novel clustering algorithm based on regularized Wasserstein distance for multivariate simulation outputs, enhancing analysis of stochastic systems.

Findings

Successfully identified similar staffing plans in a call-center model.

Enabled detection of performance anomalies through clustering.

Provided insights for system intervention based on queue length signals.

Abstract

Using statistical learning methods to analyze stochastic simulation outputs can significantly enhance decision-making by uncovering relationships between different simulated systems and between a system's inputs and outputs. We focus on clustering multivariate empirical distributions of simulation outputs to identify patterns and trade-offs among performance measures. We present a novel agglomerative clustering algorithm that utilizes the regularized Wasserstein distance to cluster these multivariate empirical distributions. This framework has several important use cases, including anomaly detection, pre-optimization, and online monitoring. In numerical experiments involving a call-center model, we demonstrate how this methodology can identify staffing plans that yield similar performance outcomes and inform policies for intervening when queue lengths signal potentially worsening system performance.