Analysis Performance Classification of Wisconsin Diagnostic Breast Cancer WDBC Via Supervised Machine Learning Algorithms

Abstract

Breast cancer is one of the leading causes of cancer death in women. It frequently results in deadly outcomes due to delayed identification in advanced stages. Early detection and treatment significantly improve a breast cancer patient's chance of survival. Recent advancements in machine learning have opened up possibilities for early detection. Objective: The objective of this study was to apply and compare several machine learning (ML) algorithms to see which performed better for breast cancer prediction. Methods: To achieve the study's objectives, several steps were taken. The Wisconsin Diagnostic Breast Cancer (WDBC) has been used. Data preprocessing includes normalizing, removing outliers, and fixing missing values. The data was separated into two parts: training (80%) and a testing set (20%). Several techniques have been utilized, including Multi-Layer Perceptron (MLP), Ada Boost (AB), Gradient Boosting (GB), Support Vector Machines (SVM), K-Nearest Neighbors algorithm, Decision Trees (DT), and Gaussian Naïve Bayes (NB). Accuracy, precision, recall, and F scores were used to evaluate performance. Result: The Multi-layer Perceptron (MLP) achieved the highest accuracy of 0.998, compared with other techniques. Evaluated: To evaluate the proposed model via comparison with available studies. This study achieved superior performance to existing works for the classification prediction of breast cancer.