Application Of Machine Learning Model In Prediction Of Endometrial Cancer Survival

In recent years,machine learning methods have been explored and widely applied in clinical cancer research.In this paper,692 patients diagnosed with endometrial cancer were collected based on SEER database and divided into modeling group and model test group according to the ratio of 8:2.The factors influencing prognosis of patients with endometrial cancer were analyzed,and the prediction model of correlation graph was constructed.Univariate and multivariate Cox regression analysis of patients in the modeling group was performed to obtain the independent factors affecting the prognosis of endometrial cancer.Univariate and multivariate Cox analysis results showed that age,tumor grade,FIGO stage,mode of surgery,chemotherapy,lymph node metastasis,degree of tumor entry,site of metastasis,lymph node status,multiple primary,marital status,and tumor-related death would all affect the independent prognostic factors of patients with endometrial cancer.Completion of the construction of the prediction model of the line graph,and according to the verification of the model,the productivity of the modeling group and the mode test group on the calibration curve was obtained,which was generally consistent with the actual survival rate.Then,K-M was used for internal analysis of related factors,and the survival curve was drawn.The widespread use of machine algorithms in the medical field has also given rise to comparative analysis between models.Based on the generalization ability problems in the prediction of a single algorithm model,an integrated learning model for Stacking based on the fusion of multiple algorithms is proposed in this paper for forecasting.In this paper,decision tree and random forest models,which are widely used and mature in theory,are selected to focus on the prognosis prediction of endometrial cancer.By comparing the machine learning model with the traditional prediction model and the prediction ability between machine learning models,various combination methods are used to establish the relevant combination prediction model and compare with the single prediction model,so as to seek the optimal prediction.The results show that the accuracy of the forecasting model of Stacking is higher than that of a single stacking model.Experiments show that the data classification prediction accuracy of the integrated learning model in Stacking is 84.46%,which is higher than the decision tree,random forest 68.60%,and 74.09% of the single algorithm model,respectively.The model was applied to patient data collected from hospitals for empirical analysis,and the conclusions were consistent.Compared with a single algorithm model,the integrated learning model has strong forecasting ability and can better predict diseases,providing reference for the clinical diagnosis and early warning of endometrial cancer.