Research And System Construction Of Infectious Disease Detection And Prediction Mechanism Based On Deep Learning

Infectious diseases pose a major challenge to global public health,and their outbreaks and epidemics not only threaten human health but also have a significant impact on social development.Detection and prediction of infectious diseases,as important means of prevention and control,have become hot topics in the fields of public health and medicine.In current infectious disease research,researchers often use large-scale datasets provided by public health agencies or disease control centers for their studies.However,infectious disease detection and prediction not only require accuracy but also emphasize timeliness of results.If detection and prediction measures are not timely and effective,it may lead to the loss of control of the disease and casualties.This thesis aims to improve the timeliness of infectious disease detection and prediction while ensuring the accuracy of the results.Based on this,this thesis proposes an infectious disease detection method based on image data augmentation and an infectious disease prediction model based on deep learning and temporal features,and designs an infectious disease detection and prediction system.The specific research work of the thesis mainly includes the following three aspects:(1)This thesis proposes a new data enhancement method called MLMix,which improves the robustness and classification effect of images by combining single-sample data augmentation and multi-sample data augmentation.This thesis further improves the robustness of images by generating different images through multiple random function calls to various data augmentation methods.The experimental results show that the performance of the Resnet50 model has been significantly improved after using the MLMix method.Specifically,the Recall increased by 5.14%,Precision increased by 2.4%,F1-score increased by 6.3%,Accuracy increased by 6.8%,and AUC increased by5.19%.(2)This thesis proposes a new network model called Ls-NET.Combined with DLSTM and Transformer models,it conducts long-term learning of seasonal,monthly,yearly,and other time series changes.In addition,Ls-NET uses a custom Flag mechanism to score the prediction results and correct them according to different scores,which has good applicability.The experimental results showed that using the Ls-NET model can significantly improve the prediction accuracy of the number of infectious diseases in multiple countries.The relative error of the prediction in Italy was only0.17%.(3)This thesis implements the front-end and back-end combination through the Vue and Django frameworks,and designs a visual infectious disease detection and prediction system.The system mainly includes two modules,the infectious disease detection module and the infectious disease prediction module,which are based on the MLMix method and Ls-NET model,respectively,for classification and prediction of infectious disease lung CT images and infection numbers.Later,a module for infectious disease news was added,allowing users to read and submit infectious diseaserelated news for epidemic preparedness.The system ultimately presents the results in a visual way.