Comparative Study On The Effect Of Neural Network Model On Early Warning Of Nosocomial Infection Cases

Background After SARS in 2003,the related concepts of nosocomial infection and its management were formally started in China.And through many laws and principles have gradually developed into the management department necessary for various medical institutions at all levels-the Hospital sense Section.However,the manage work of nosocomial infection is hard and overwhelming,all of which are taken care of by several or even only one full-time professional staff.It is a common phenomenon in our country.And because the whole sense of control work in our country started relatively later than in developed countries,the foundation is relatively weak,although in Under the rigid requirements of laws and regulations,the overall framework of the system is built relatively quickly,but the substantive ability and the overall quality of the practitioners,their professional skills and even their level of education are weaker than those of the health care system or the nursing system,that are already mature.However,the outbreak of nosocomial infection is of a bad nature,with serious consequences,extensive harm and can be prevented and controlled.Then,under such work content and working nature,the use of simple and clear auxiliary tools can not only make up for the lack of professional quality in hospital sense management,but also solve the tedious work in hospital infection management.Purpose In order to reduce the burden of hospital management,to reduce the working pressure of clinical medical staff,and to improve the quantity and quality of hospital report and for the follow-up work based on data precision,the neural network and the decision tree classifier are carried out separately.Method The method of interactive result,that is,the neural network is used for the early warning and prediction of nosocomial cases while the decision tree is used to grasp the high risk factors and critical control points,both of which work together to assist in the management of nosocomial infection.The two methods are used to analyze the patient information of the hospital during a given period of time.Get a rule generated by a trained neural network The algorithm is used to predict the patient information in another period of time,and the predicted results are compared with the actual results.At the same time,the above data are simulated and predicted by logistics regression,and theresults are also compared.In order to seek the best data analysis core algorithm for nosocomial infection information system.Results In the same 1047 test cases(28 nosocomial,1019 non-nosocomial),the predict accuracy of logistics regression analysis was 56.1%(14 false negative cases),accuracy of fine tree model was 94.5%(25 false negative cases),accuracy of medium tree model was 92.6%(22 false negative cases),accuracy of coarse tree model was 71.5%(12 false negative cases),accuracy of adjusted coarse tree model was 70.1%(10 false negative cases),and neural network prediction has an accuracy of 76.7%(10false negative cases),while it has the smallest quantity of false positive cases,which is the best model.Conclusion In this study,logistics regression model,classification tree model,coarse tree model and neural network model were calculated and tested using real data,and the results were compared,and the best model for hospital case prediction,neural network model,was obtained.And the best model for condition analysis-decision tree classifier model.The mutual application of the two plays a central role in the management of nosocomial infection and is the best solution to the problem of nosocomial infection reporting.