A DATA-DRIVEN APPROACH TO BREAST CANCER DETECTION USING NON-NEGATIVE MATRIX FACTORIZATION
Zulfikar A. Ansari1, Manish M. Tripathi2, Rafeeq Ahmed3
1Integral University, CSE, Lucknow, 226022, India, 1Koneru Lakshmaiah Education Foundation, CSE, Vaddeswaram, 522302, India
2Integral University, CSE, Lucknow, 226022, India
3Government Engineering College, CSE West Champaran, 845438, India
Abstract: Breast cancer (BC) remains a prevalent form of cancer among women globally, with the potential to be fatal. Timely detection is crucial for effective therapy and improved survival rates. Machine learning methodologies have demonstrated their potential to assist physicians in the early detection of breast cancer. Non-negative Matrix Factorization (NMF) is an effective method for dimensionality reduction and feature extraction. It is utilized in various domains, including medical image analysis. We examined the efficacy of Support Vector Machines (SVM), Logistic Regression (LR), Decision Trees (DT), Random Forests (RF), Multi-Layer Perceptron’s (MLP), and Non-negative Matrix Factorization (NMF) in diagnosing various types of issues. The combination of MLP and NMF achieved the highest accuracy of 97.90%. The study’s findings indicate that machine learning techniques, particularly MLP in conjunction with NMF, may enhance the accuracy of breast cancer diagnosis. This is crucial for the early detection of breast cancer and obtaining effective treatment, ultimately resulting in improved patient outcomes. Employing Non-negative Matrix Factorization (NMF) to diminish the dimensionality of gene expression data, in conjunction with machine learning methodologies, may enhance the precision of breast cancer detection and treatment.
Index Terms: Breast Cancer, Machine Learning, Non-Negative Matrix Factorization, Support Vector Machine, Multilayer Perceptron.