Machine learning based decision making for time varying systems: Parameter estimation and performance optimization

Abstract

The class of decision making problems focuses on the optimization of single or multiple design objectives, and the classical decision making procedures require the full scope of the system information. However, the system dynamics consist of unknown time varying parameters within a specific range of dynamic decision making problems, which cannot be handled by the classical procedures. To solve these problems, this paper proposes a machine learning based decision making algorithm. It uses the technique of machine learning to estimate the real-time unknown parameters using the recorded system data, and makes appropriate decisions using model predictive control (MPC) method to optimize some desired key performance indicators (KPIs). The effective performance of the proposed algorithm is further evaluated using a simulation based case study.