Improved Approach to Feature Dimension Reduction for Efficient Diagnostic Classification of Breast Cancer
Babatunde S. Emmanuel · 2021
Machine learning deals with the approaches that involved the scientific computing methodologies of understanding patterns in complex datasets. Scientific computing is pervasive in the physical, biological and engineering sciences. This paper focuses on the application of machine learning technique for diagnostic analysis of cancer datasets. The problems encountered in the proposed method include: irrelevant and redundant data features resulting in high dimension of extracted data features. Feature selection method was employed to reduce the problem of extreme data dimensionality defined by irrelevant and redundant features that characterized the training dataset in order to improve the performance of the classification model. The goal is to apply machine learning approach to diagnose patients with breast cancer by analyzing the data of patients and classifying them into reliable clinical diagnosis results using reduced feature dimension and at the same time reducing errors to the barest minimum. Standard performance metric such as classification accuracy was used to evaluate the resulting model. The proposed classification algorithm produced an accuracy of 98.24% when the optimal number of 17 significant features were selected out of 31 extracted features.