Learning-based optimization methodology

The project focuses on learning-based techniques for scheduling problems with uncertainty. Stochastic operational optimization methods are developed to handle uncertain scenarios. Models and algorithms including constructive and repair heuristics for dynamic personnel scheduling will be investigated firstly. Then the methods will be adapted to solve other combinatorial optimization problems. The learning-based mechanism will be introduced to make the algorithms more adaptive and robust. Such algorithms will make use of some data in the process of running and also learn from historical data on the between-instance knowledge. In this way, we could propose a synergetic approach combining optimization and machine learning to make the models and algorithms in operations research grow in intelligence over runtime.