Metric Learning for Simulation Analytics

摘要: The sample path generated by a stochastic simulation often exhibits significant variability within each replication, revealing periods of good and poor performance alike. As such, traditional summaries aggregate measures overlook the more fine-grained insights into operational system behavior. In this paper, we take analytics view output analysis, turning to machine learning methods uncover key from dynamic path. We present k nearest neighbors model on state information facilitate real-time predictions measure. This is built premise system-specific measure similarity between observations state, which inform via metric learning. An evaluation our approach provided activity network wafer fabrication facility, both give us confidence in ability provide interpretation improved predictive performance.