Breast Cancer Diagnosis using Quality Control Charts and Logistic Regression

In this work, we propose data mining techniques for the diagnostic of breast cancer. The contribution of this work is data pre-processing which includes outliers removal and dimensionality reduction to visualize the data. After data visualization, the most appropriate machine learning algorithm can be proposed to perform this diagnosis. For the evaluation of our proposed approach, we have used UCI machine learning breast cancer Wisconsin (original) dataset (WBCD). We applied principal component analysis for dimensionality reduction in the pre-processing and different control charts and box plots for data visualization. After data visualization, WBCD is considered to be linearly separable and the logistic regression model is proposed for classification in the diagnosis. The evaluation of our model is performed using 10 fold cross validation with accuracy, recall, F1 score, confusion matrix and receiver operating characteristic (ROC) curves. In our proposed work, we have achieved 99.48% as accuracy, which is the highest result found on this dataset for 70-30% training-test partition.