Analysis And Prediction Of Pathological Diagnosis Of Breast Cancer Based On BP Neural Network

In female malignant tumor patients,breast cancer patients have the highest fatality rate,seriously endangering the health of contemporary women.The early diagnosis of malignant tumor can effectively improve the survival rate of patients,so the early diagnosis of breast cancer is a very important subject.In recent years,decision tree,SVM,generalized regression and other methods have been applied to the diagnosis of breast cancer,and neural network as a new machine learning algorithm,the research content in this field is still relatively few.The training of neural network model does not need to make specific assumptions about input and output,has the advantages of self-learning and self-adaptation,and is a very suitable tool to deal with nonlinear problems.Some scholars have proposed the application of Probabilistic neural network,SOM Neural network algorithm and so on in the diagnosis of breast cancer.The BP algorithm used in this paper is a mature neural network algorithm,because of its special feedforward mechanism and error reverse propagation mechanism,compared with other neural network algorithms,BP algorithm classification effect is better.The empirical analysis of this paper was followed by the application of the BP network to specific cases,selected as a sample of breast cancer provided by Dr.William H.Wolberg of the Wisconsin School of Medicine.After the BP neural network model is established,it is compared with the traditional logistic regression method.The main evaluation indexes used in the evaluation process of the model are classification accuracy,obfuscation matrix,ROC curve and cross-verification method.The statistical software used in the modeling process is mainly the R language.At the end of this paper,the paper summarizes the whole article,draws the conclusion,the statement is insufficient,and puts forward the prospect.The output results of R language modeling show that the classification efficiency of BP Neural network for the pathological diagnosis of breast cancer is satisfactory.Logistic regression models tend to explain more intuitively the causal relationship between independent and variable variables.And when dealing with more complex data,the processing power of BP neural network model is stronger.