Study On Classification Model Of Common Infectious Diseases In Zhejiang Province Based On Bavesian Classification Algorithm

ObjectiveTo establish a classification model of common infectious diseases in Zhejiang province, by which diagnosis can be made according to the information such as symptoms, signs, laboratory examination results, epidemiological characteristics and morbidity of these diseases. To help site technician identify diseases rapidly and scientifically, strive for opportunities of immediate disease control in order to take measures to control diseases timely and effectively, and provide clues for lab pathogen detection.MethodsThis study comprehensive analysed outbreaks of infectious disease and public health emergencies in Zhejiang province in recent years, obtained common infectious diseases category as the range of disease in this study. Data were collected from articles of outbreaks and notifiable communicable disease network on symptoms, signs, laboratory examination results, morbidity and epidemiological characteristics of these diseases. Bayesian classification algorithm was applied for model construction and SAS software was used for programming. The model was applied to three sets of data of historical epidemic situation for model verification. Results25common infectious diseases in Zhejiang province were included in this study. The results of model verification showed that the coincidence rate of actual category and first diagnosis and the suggestion rate of first three diagnosis were93.55%and100%separately in the data of hand-foot-mouth disease, which were60.61%and100%separately in the data of epidemic hemorrhagic fever and88.73%and100%in the data of mumps. The average running time of programme were0.51,0.50and0.53seconds separately in those three sets of data.ConclusionsThe classification model of common infectious diseases has been constructed in this study. After the enter of symptoms, signs, laboratory examination results and epidemiological characteristics, possible diseases are descripted and sorted by there probabilistic with the output of a discrimination list. This model has high prediction accuracy, and shows great speed, interpretability, retractility and robustness, which is complied with the actual application requirements.