Prediction Of Breast Cancer Based On SVM-MLP

Breast cancer has always been the primary hidden danger of women’s health,and has become the highest incidence of malignant tumor in women around the world.In earlier times,the frequent,in the middle age of the female breast cancer in recent years,due to the rapid economic development,many jobs has a tendency to machines replaced manual,the employment pressure in young women,breast disease prevalence in young women white-collar group increase year by year,the high incidence of phenomenon have aroused the concern of the global medical technology.Today,the use of machine learning and deep learning methods to diagnose cancer has become a branch of the rapid development of current,if can only each index of judging by observing the constitute factors of breast cancer,the comprehensive analysis of patients at risk,whether can reduce the cost of related bureaucracy,assist doctors for early diagnosis of diseases,effectively improve the efficiency of breast cancer in clinic.In this paper,a variety of machine learning and deep learning methods are used for the prediction classification diagnosis of breast cancer data sets by using Logistic Regression(LR),K-Nearest Neighbor(KNN),Naive Bayes(NB),multi-kernel function Support Vector Machine(SVM),Multi-layer Perceptron(MLP),and Deep Belief Nets(DBN).The experimental data was the Breast Cancer Wisconsin data set from UCI Machine Learning Library,which was divided into benign and malignant categories.After the correlation analysis of the distribution of data attributes and the relationship between attributes,the data were processed with missing eigenvalues and normalization,and the processed data were retained as the input and output variables of the multi-layer perceptron and the deep confidence neural network with self-designed network structure,and various evaluation indexes were obtained for comparison.In this experiment,the training set,verification set and test set were allocated according to the ratio of 7:2:1.The machine learning method mentioned above was used to construct the model and simulate the same data set respectively.Finally,it was concluded that the iteration time of SVM classification model was the shortest and the prediction accuracy of multi-layer perceptron was the highest.In order to reduce the iteration time,improve the SVM prediction accuracy,this paper proposes a multi-layer perceptron classification based on SVM optimization model of breast cancer,namely the SVM-MLP model,because the SVM-MLP model does not represent a more complex network structure,so the introduction of depth incredibly complex neural network to predict the same data set again,the experimental results show that: Compared with logistic regression model,K-nearest neighbor algorithm classification model,Bayesian classifier and Gaussian kernel SVM model,SVM-MLP model has higher classification accuracy,It is verified that according to the number and attributes of the data set,choosing a more appropriate neural network learning method for model building and prediction can get a better evaluation index.