Abstract: In order to solve the problems of difficult accurate evaluation and regulation of environmental quality of high-temperature and high humidity layer house in southern region in summer, the relational model between layer house environment and egg performance was constructed based on multiple neural networks. The collected environmental factor data of layer house were used as input, and the environmental quality grade divided according to egg production rate were used as output; three neural networks BP, RBF, and GRNN, were trained and validated. The preliminary judgment result of the environmental quality of the layer house was obtained. Then, the output data of the normalized neural network group is taken as the inputs of the basic probability distribution of the improved D-S evidence theory. Finally the comprehensive evaluation result of environmental quality of the layer house was obtained by the improved D-S evidence theory and the fusion judgment. The results showed that in the case of conflicting neural network primary fusion decisions, the improved D-S algorithm had support rates of 0.796 4 for poor environmental status in the layer house, while the classical D-S method, Murphy method, and Deng Yong method had support rates of 0.698 7, 0.687 8, and 0.694 0 respectively. The results indicated the comprehensive evaluation model constructed by the experiment had higher fusion judgment precision and accuracy for the laying house environment, and could effectively solve the fusion decision error problem, which provided a reference for the accurate regulation of high temperature and high humidity layer house environment in southern summer.