Classification Performance Evaluation on Diagnosis of Breast Cancer

Cancer, which has many different types such as breast, pleural, and leukemia, is one of the common health problems of today. Most of them cause pain and treatment processes are so challenging. Medical authorities report that the diagnosis of cancer at early stages has a positive effect on medical treatments' success. On the way to design a computer-aided cancer diagnosis system about breast cancer to support the decisions of doctors about medical treatments, classification performances of six classifiers are investigated in this study. For this purpose, classifier models have been created with machine learning algorithms such as Support Vector Machine (SVM), Naïve Bayes (NB) and Random Forest (RF); and deep learning algorithms such as Recurrent Neural Network (RNN), Gated Recurrent Unit (GRU), and Long Short-Term Memory Network (LSTM). Two different open-source breast cancer datasets were used namely Wisconsin and Coimbra on experiments. Accuracy (Acc), Sensitivity (Sens), Precision (Pre), F-measure (F) were used as performance criteria. As a result of the tests, Acc values between 85% and 98% were obtained in the Coimbra breast cancer dataset; while Acc values were obtained between 92% and 97% in Wisconsin. According to the results obtained, it is seen that deep learning algorithms (RNN, GRU, and LSTM) are more successful than machine learning algorithms (SVM NB and RF). Among the deep learning algorithms, LSTM is more successful. © 2021, The Author(s), under exclusive license to Springer Nature Switzerland AG.2-s2.0-8510983488