Supporting complex real-time decision making through machine learning

摘要

This paper presents FMS-DSS, a system for supporting complex, real-time decision making in the FMS scheduling and control domain. FMS-DSS differs from traditional DSSs in that it can acquire scheduling and control knowledge from historical data comprising prior decisions. This knowledge is applied to support subsequent decision making. It manages complexity through hierarchically structuring the user's objectives, and can deal with noise in the form of missing, inaccurate, or erroneous data. Results indicate that a machine learning based approach can provide effective support for repetitive real-time decision making and outperform static scheduling rules.