Research On Survival Prediction Of Breast Cancer Based On SVM

With the integration of computer technology and medicine,this paper conducts SVM-based breast cancer survival prediction research in order to contribute to the improvement of the prognosis of breast cancer patients.Since the survival rate of breast cancer patients is much higher than the mortality rate,resulting in an imbalance of data sample categories,which is an inherent characteristic of breast cancer survival prediction studies and affects the improvement of the accuracy of survival prediction by machine learning algorithms.To address this problem,this paper uses Support Vector Machine(SVM)as the basic research algorithm to optimize and improve the breast cancer survival prediction method using the processed cancer database SEER(Surveillance,Epidemiology,and End Results)as the dataset.Firstly,to address the problem of breast cancer survival prediction and kernel function selection for SVM,the m-arcsinh kernel function is introduced based on the experimental results of breast cancer data,and the Q-arcsinh-based SVM model(QSVM)is constructed for breast cancer by reassigning different weights to the hyperbolic inverse function and square root function to address the problem that marcsinh cannot face the complex application environment.the study of survival prediction.Secondly,to solve the problems of poor initial population quality,limited information exchange between populations,and low convergence accuracy in the Sparrow Search Algorithm(SSA),a new ERBT-SSA algorithm incorporating three strategies of elite lens reversal learning,butterfly optimization strategy,and T perturbation is proposed for the optimization of key parameters of QSVM.Finally,an ERBT-SSA-QSVM method for breast cancer survival prediction including data preprocessing is proposed.For the imbalanced breast cancer data,a method is proposed that incorporates Random Under-Sampling(RUS),K-Nearest Neighbor(KNN)algorithm,and Synthetic Minority Oversampling Technology(SMOTE).SMOTE)of RKSMOTE algorithm,and then use ERBT-SSA algorithm to tune the key parameters of QSVM model,so as to construct a breast cancer survival prediction model with high prediction accuracy and realize the survival prediction of breast cancer.The experimental results showed that the ERBT-SSA-QSVM method combined with the RKSMOTE algorithm obtained more accurate results and better robustness in breast cancer survival prediction on the breast cancer data extracted from the SEER library.