Though there are a handful of dimensionality reduction based machine learning algorithms used in breast cancer research, they are either poor with prediction accuracy or require a large numbers of features for malignancy detection. To overcome this cliche, we have considered Uniform Manifold Approximation and Projection (UMAP), a faster dimensionality reduction algorithm, capable of projecting any N dimensional data into a 2-D representation, this UMAP is used in combination with Machine Learning algorithms. To compare the performance of UMAP with respect to the other dimensionality reduction based ML algorithms, we have developed a new evaluation criterion based on parameters obtained from the literature review such as Projected Dimensions, Accuracy, Specificity, Sensitivity and the Classification time. The observations suggest that UMAP with SVM (Linear) performs faster and better than other reduction based ML algorithms in Breast Cancer malignancy detection.
Volume 11 | 02-Special Issue
Pages: 351-362