Instance space analysis for the car sequencing problem

Abstract We investigate an important research question for solving the car sequencing problem, that is, which characteristics make instance hard to solve? To do so, we carry out space analysis by extracting a vector of problem features characterize instance. In order visualize space, feature vectors are projected onto 2-D using dimensionality reduction techniques. The resulting visualizations provide new insights into instances used testing and how these influence behaviours optimization algorithm. This guides us in constructing set benchmark with range properties. demonstrate more diverse than previous benchmarks, including some significantly difficult solve. introduce two algorithms compare them four existing methods from literature. Our shown perform competitively this but no single algorithm can outperform all others over instances. observation motivates build selection model based on machine learning, identify niche is expected well on. helps understand hardness select appropriate given