Expected gain based early warning for real-time process monitoring

Abstract

This paper proposes a novel approach to real-time process monitoring using expected gain based early warning. At every monitoring period, a probability to satisfy the targeted outcome is estimated by an extended decision tree algorithm. Then, the expected gains and losses are estimated for decision alternatives: executions are stopped or continue until completion. If the expected gain of stopping is larger, the early warning is generated to provide intuitions about unstable status of the ongoing process. We conducted experiments to show how the real-time process monitoring is implemented and the early warning is generated. The proposed approach can provide sophisticated indicators reflecting the real-time progress of the ongoing process based on the observed attributes until the monitoring instant and possible outcomes after that. ICIC International © 2011.