An adaptive meta-imitation learning-based recommendation environment simulator: A case study on ship-cargo matching

Guangyao Pang, Jiehang Xie, Fei Hao

Abstract: Highlights•We propose a shipowner state generation model based on the guided diffusion model, which utilizes contextual states and responds to environmental changes to generate diverse shipowner states.•We present a shipowner reward model based on adaptive meta-imitative learning, which emulates human learning behavior and allows multi-class shipowner reward from the environment of small sample size and unbalanced categories.•We provide an recommendation environmental simulator that considers multitasking and diversity in small samples.