Integrated Ensemble Strategy for Breast Cancer Detection using Dimensionally Reduction Technique

Abstract

Abstract Breast cancer continues to be a prominent issue in global health, requiring the implementation of novel approaches for the timely identification and assessment of the disease. Machine learning has been extensively integrated into the field of breast cancer diagnostics to gain profound insights and enhance the precision and efficacy of recognizing potential instances of breast cancer. Given the global nature of this disease, the early detection of cancer continues to pose a considerable problem. Our study introduces an ensemble strategy that integrates the results of Dimensionality Reduction (DR) approaches, namely Principal Component Analysis (PCA), Non-negative matrix factorization (NMF), and Value Decomposition (SVD), and subsequently inputs them into a resilient classification algorithm. In this study, we examine many algorithms, namely Logistic Regression (LR), Support Vector Machines (SVM), Random Forests (RF), Decision Tree (DT), and Multi-Layer Perceptron (MLP), to evaluate their diagnostic accuracy. Our findings show that MLP, LR, and SVM have a maximum accuracy of 97.9%, but MLP performance varies when used with NMF & PCA, which is 97.20%. LR also produced good accuracy with NMF and PCA, which is 97.9%, but again, performance is reduced when used with SVD. The SVM gives a consistent result with PCA, SVD, and NMF, which is 97.9%.