Optimization Of Survival Analysis Models And Its Application In Disease Prognosis

Medical industry is closely related to human life and health.In particular,the spread of COVID-19 and increased public awareness of life safety have boosted the development of medical research.In the field of medical and health care,a single evaluation index can not carry out comprehensive assessment of patients with chronic diseases.It is the current situation of medical industry to track the survival state of patients,collect vital signs of patients in a period of time,and determine specific diagnosis programs.Prognosis is the focus of medical research.By analyzing observational data,clinicians can develop treatment plans and patients can understand their disease status.Prognosis has a profound influence on clinical medicine.Survival analysis is commonly used in medical field.This method fully considers the time factor,analyzes the relationship between the time distribution of the event and the existence of the observed object,and explores the key information affecting the occurrence of the event.Therefore,the study of survival analysis method has been widely concerned.Survival analysis can be divided into two stages:traditional survival analysis and deep survival analysis.On the one hand,traditional survival analysis such as Cox proportional risk model needs to consider model assumptions,however,the distribution of a large number of real data is unknown.On the other hand,Cox model only considers the linear characteristic information of data,which loses part of the value of data,so it has limitations.At present,the application of deep survival analysis in the medical field is more common,and the model has strong expression ability.In this paper,the information transmission of survival data in time dimension is fully considered,and GA-LSTM is combined with deep survival analysis algorithm to conduct prediction analysis on medical data sets.The research content of this paper is as follows:1)This paper introduces in detail the relevant knowledge of survival analysis,the current development status and the corresponding theoretical basis.2)The GLDS model is proposed in this paper.GLDS model is an optimization algorithm based on the combination of genetics and LSTM.On the one hand,it can better screen the superparameters of the model.On the other hand,LSTM can fully mine the characteristics of time information in medical data,and give a certain interpretation.3)In this paper,GLDS method is applied to the clinical survival data of breast cancer,and the prediction model of breast cancer patients is constructed.By comparing with traditional survival analysis,the feasibility of the samples was tested and the main factors affecting the onset of patients were examined.In order to verify the reliability of the GLDS model,we compared it with the existing survival analysis model and verified on multiple public data sets,which show that the prediction effect of the model is significant.In the prediction analysis of breast cancer data set,the C-index index of the model used in this paper reached0.7742,while the prediction result of the traditional Cox model was 0.7389,and the prediction effect of GLDS method was improved by 3.53%.The improvement of prediction results not only indicates that GLDS is an effective survival analysis model,but also can be used by clinicians to predict the prognosis of severe patients,which has certain significance in the medical field.