Machine Learning Approaches for Early Diagnosis of Breast Cancer: A Comparative Study of Performance Evaluation

Abstract

Breast cancer is a leading cause of death among women worldwide. Early detection and diagnosis are crucial to improving the chances of survival. This paper presents a study on the diagnosis of breast cancer using various machine-learning approaches. The study includes the performance evaluation of nine different techniques using confusion matrix accuracy for Sensitivity, Specificity, Precision, PME, PPV, NPV, and Model Accuracy. AdaBoost is found to have the highest Sensitivity and PME, while Random Forest and MLP gave the best Specificity and Precision. Logistic Regression is found to be the best model for accuracy with 97.8%, followed by SVM with 96.49%, Random Forest with 95.61%, and KNN & Decision Forest with 94.73%. The proposed approach is found to have the highest accuracy of 97.80% compared to other approaches studied. Our results indicate that the proposed approach using discretization can significantly improve the signal-to-noise ratio in the diagnosis of breast cancer. This approach can accurately predict and diagnose breast cancer using a subset of features.