Performance analysis of different machine learning algorithms in breast cancer predictions

Abstract

INTRODUCTION: There is a great percentage of failures in clinical trials of early detection of breast cancer. To do this, machine learning (ML) algorithms are useful to do diagnosis and prediction of cancer tumors with better accuracy.

OBJECTIVE: In this study, we develop an ML model coupled with limited features to produce high classification accuracy in tumor classification.

METHODS: We considered a dataset of 569 females diagnosed as 212 malignant and 357 benign types. For model development, three supervised ML algorithms namely support vector machines (SVM), logistic regression (LR), and K-nearest neighbors (KNN) were employed. Each model was further validated by 10-fold cross-validation and performance measures were defined to evaluate the model outcomes.

RESULTS: Both SVM and LR models generated 97.66% accuracy with total feature evaluation. With selective features,the SVM accuracy was improved by 98.25%. Whereas the LR model including limited features produced 100% of true positive predictions.

CONCLUSION: The proposed models involved by selective features could improve the prediction accuracy of a breast cancer diagnosis.