Abstract: This paper used the data on the number of patients with respiratory disease、meteorological factors and pollutant concentration from 2015 to 2016 to construct a prediction model for the admission visits from respiratory disease by using BP artificial neural network and LSTM network.The results showed that the cold stress effect of winter monsoon would increase the incidence of respiratory diseases in related populations fluctuating from September to the peak of March before the transition from winter wind to summer wind in the following year.During the summer wind control period,the incidence of respiratory diseases among local residents fluctuated and decreased,which was 35% lower than that during the peak period.In addition,the main control factors of respiratory diseases were different. Compared with the two prediction models,on the whole,the LSTM network forecasting model has a higher accuracy rate in predicting the risk of respiratory diseases in Shenzhen,which can meet the business needs of health weather forecasting services.