Feature construction in genetic programming hyper-heuristic for dynamic flexible job shop scheduling

Genetic Programming Hyper-Heuristic (GPHH) has shown success in evolving dispatching rules for dynamic Flexible Job Shop Scheduling (FJSS). In this paper, we focus on feature construction to improve the effectiveness and efficiency of GPHH, and propose a GPHH with Cooperative Co-evolution with Feature Construction (CCGP-FC). The experimental results showed that the proposed CCGP-FC could improve the smoothness of the convergence curve, and thus improve the stability of the evolutionary process. There is a great potential to improve the FC methods, such as filtering the meaningless building blocks, and incorporating with feature selection to improve the terminal set.