Performance analysis of machine learning algorithms in breast cancer diagnosis

Breast cancer is recognized as one of the foremost causes of death among women in worldwide with more than one million of cases and nearly 600,000 deaths each year. It is extremely important to identify it at the early stage in considerably to increase the chances of survival. The breast cancer can be classified into two types of tumors which are benign and malignant. Benign tumors are undangerous tumors where they develop slowly in organ while malignant are dangerous tumors where they would spread to the other organ of body. In this paper, five machine learning algorithms are used to predict if the tumor is benign or malignant based on the Wisconsin Prognostic Breast Cancer dataset, while one of the algorithms is modified to achieve a better performance. The five algorithms used are Gaussian Naïve Bayes Classifier, Random Forest Classifier, Decision Tree Classifier, Kernel Support Vector Machine Classifier, and K-Nearest Neighbors Classifier while the modified algorithm is Kernel Support Vector Machine Classifier. The aim is to use Machine Learning algorithms to make prediction of breast cancer and improved the accuracy of the algorithm. 10-fold Cross Validation is implemented after compared it with Bootstrapping as a resampling method as it is more efficient. At the end, the comparison in results shown that the modified Kernel Support Vector Machine Classifier predicted the highest accuracy among these five machine learning models.