Abstract: A fixed set of evaluation factors makes the multi-layer comprehensive evaluation (CE) model difficult to use when the set of evaluation factors changes with different objects to be evaluated． To solve this problem, in this paper an adaptive multi-layer comprehensive evaluation model is proposed to solve the limitation of constant set of evaluation factors． This new model enables dynamic comprehensive evaluation to facilitate comprehensive evaluation with changing evaluation factors． In the proposed model, the cross-layer weight transfer mechanism based on factor distance and similarity kernel is given, and the corresponding weight transfer scheme is designed． To illustrate the rationality, a weight transfer scheme is designed based on entropy regular optimization, and the equivalence of the two schemes under identical settings is proved． Comprehensive evaluation experiments on typical cases verify effectiveness of the proposed model for adaptive multi-layer comprehensive evaluation．