Comparative study of machine learning approaches on diagnosing breast cancer for two different dataset

Abstract

Over 50 years in the world Breast Cancer (BC) has become a major health issue. According to the World Health Organization (WHO) in each year, approximately 1.38M new breast cancer cases arise. Among women, it is the leading cause of cancer death and also the most second in common. However, we can diagnose, manage breast cancer effectively, and also improve the surviving rate of breast cancer by correctly and early detection. Because if we take a long time for diagnosing breast cancer it may increase the chances of cancer spreading rate. Lots of different Machine Learning technique is used nowadays to diagnosis BC. In this paper, we apply different classification algorithm such as Naive Bayes (NB), Support Vector Machine (SVM), Logistic Regression (LR), K-Nearest Neighbor (KNN), on two different datasets of BC (Wisconsin Breast Cancer (WBC), Wisconsin Diagnosis Breast Cancer (WDBC). By using the confusion matrix on 10-fold cross-validation, classification accuracy, precision, sensitivity and f-score we compare the performance of different classifiers. Here ROC curve is also used to show the comparison of the different classifiers on different datasets after extracting key features. The prime objective of this paper is to use key features to detect breast cancer with high accuracy and a low error rate in a short time. To find objective and systematic prognostic it will highly help the pathologists.