Study On Factors Of Breast Cancer Recurrence By Logistic Regression Model, ANN Model And SVM Model

To study the factors that may affect breast cancer recurrence,we took the patient’s age, menopausal age, tumor size, the number of involved lymph node, nodules risk, the degree of malignancy,the location of tumor and radiotherapy into consideration. In terms of the analysis methods, we use anova analysis first, and then took further analysis by logistic regression, artificial neural network and support vector machine respectively. Finally, we compare the predictive effects of these three methods, and got the best predictive model. This research may help guide clinical to take targeted prevention measures to reduce the rate of breast cancer recurrence and improve the prognosis of patients, prolong the patients’ life circle.The result showed that tumor size, nodules risk and the degree of malignancy are the main factors of breast cancer recurrence. The training samples’ overall accuracy of ANN prediction model is 77.4%, the test sample is 73.6%. The overall accuracy of Logistic Regression prediction model is 75.8%.The training samples’ overall accuracy of SVM prediction model is 79.5%, the test sample is 78.2%. The patients who get higher malignant degree, bigger tumor size and nodules may have higher probability of recurrence. What’s more, compared to the Logistic Regression Model, Support Vector Machine and Artificial Neural Network have better prediction performance.