Research On Method Of Forecast Vector-Borne Diseases Based On Deep Learning

The epidemic has comprehensive impact on human life causing discomfort or dysfunction in patients as well as panic and social unrest due to the prevalence of major infectious diseases.In addition,it may also retard economic development in one country.Vector-borne disease is one category of acute infectious diseases transmitted by animals.Mosquito-borne transmittion account for most cases of this diseases.In recent years,with the inpact of various natural and social factors such as global climate change,ecological change,population movement and urbanization,the scope of mosquito-borne infectious diseases and outbreak intensity has increased significantly.Timely and effective prediction regarding outbreak risk of infectious diseases can minimize the impact of infectious diseases.Most of the current infectious disease prediction models developed in China are based on the traditional infectious disease dynamics model with poor fitting effect on the non-linear time series data.Whatsmore,the error is large and the results are not satisfactory.Traditional mosquito-borne infectious diseases prediction relies too much on mosquito monitoring data and the data cannot be released in time,affecting the timeliness of prediction.In view of the current severe situation of infectious diseases,establishment of a more practical prediction model to improve the accuracy and timeliness of prediction is of great significance.This is also helpful for early warning,prevention as well as implementation of control strategies in the government and related organizations.Compared with machine learning models that requires manual features extract and weights determination,deep learning models have a feature of end-to-end model training making it one of the most widely used approach in time series data analysis.In addition,the deep learning models have been proved robust approaches in the field of time series analysis.In view of excessive precision deviation and prediction delay present in current infectious disease risk prediction models,deep learning technology was utilized to directly link environmental data with time series data of infectious diseases.This is helpful in improving practicality and timeliness of prediction models and thus avoiding impact result from delay release of monitoring data.Moreover,deep learning technology which predict disease with an end-to-end style can greatly improve the accuracy of infectious disease risk prediction.The main contents of current study are listed as follows:(1)Aiming at the epidemic of Japanese encephalitis in Chongqing,the long-and short-term memory neural network was applied to predict the trend.A risk prediction model of Japanese encephalitis was also established.The long short-term memory neural network is a recurrent neural network with a special structure.The gradient extinction and explosion problems of the traditional neural network was significanly alleviated by adding the forgetting gate,memory gate and output gate in the structure of the model.It has the ability to identify the long-term dependency information between sequence data variables,making it a suitable approch for time series prediction.The long-and short-term memory neural network with features of short-term predictions as well as obtainment of the long-term trend of time series data,was employed to make model of the Japanese encephalitis cases in Chongqing area.Experimental results showed that the model has good performance in terms of accuracy in predicting vector infectious disease risk.(2)In response to the increasingly severe dengue fever epidemic in mainland China,a long-and short-term memory network prediction model was established.In view of the periodic characteristics of dengue fever month data,the long short-term memory network model is utilized to predict the risk of dengue fever.The model can be used to predict the long-term development trends and seasonal periodic fluctuations of dengue fever time series data.This achieves end-to-end accurate prediction.In addition,by using Transfer Learning(TL)technology,the model can not only accurately predict the epidemic trend of dengue fever in high-risk areas of dengue fever,but also improve the predictive accuracy for early warning in areas with low dengue fever incidence.Experimental results indicated that the proposed method can effectively predict the risk of vector-borne infectious diseases.