Machine Learning Models of Breast Cancer Risk Prediction

Abstract

Breast cancer is the most common cancer in women both in the developed and less developed world. Early detection based on clinical features can greatly increase the chances for successful treatment. Our goal was to construct a breast cancer prediction model based on machine learning algorithms. A total of 10 potential clinical features like age, BMI, glucose, insulin, HOMA, leptin, adiponectin, resistin, and MCP-1 were collected from 116 patients. In this report, most commonly used machine learning model such as decision tree (DT), random forest (RF), K-nearest neighbors (KNN), support vector machine (SVM), logistic regression (LR), and artificial neural network (ANN) models were tested for breast cancer prediction. A repeated 10-fold cross-validation model was used to rank variables on the randomly split dataset. The accuracy of DT, RF, SVM, LR, ANN, and KNN was 0.71, 0.71, 0.77, 0.80, 0.81, and 0.86 respectively. However, The KNN model showed most higher accuracy with area under receiver operating curve, sensitivity, and specificity of 0.95, 0.80, 0.91. Therefore, identification of breast cancer patients correctly would create care opportunities such as monitoring and adopting intervention plans may benefit the quality of care in long-term.