A Breast Cancer Patients' Survival Prediction Model Using mRMR Feature Selection with Artificial Neural Network

Wei Xing, Kangping Wang, Huiyan Sun, Yan Wang

Abstract: Breast cancer is one of the most common cancers among women in the world, and it poses a huge threat to women's health. Predicting the survival status of breast cancer patients is of great significance to the patients. At present, due to the low classification ability of traditional machine learning algorithms, it is not enough to assist clinical diagnosis. This study combined deep learning with more powerful classification performance with medical diagnosis to improve the accuracy of diagnosis. This study constructed a model (MA) combining min-redundancy and max-relevance algorithm(mRMR) and artificial neural network, which could not only predict whether breast cancer patients would survive for more than 5 years, but also selected the features (genes)set that made classification results optimal. In terms of accuracy, the MA model's classification effectiveness could achieve 72.38%. Through survival analysis to the optimal genes set, 11 genes highly correlated with cancer survival were obtained.