WITHDRAWN: Correlation-based high distinction feature selection in digital mammogram

Abstract

For any classification algorithm to be flourishing, attribute choice from the given dataset is the most significant work that determines the whole process and classes of classified data. Feature assortment strategies to extract the most related features, and to drive out the inappropriate attributes with respect to given decision attribute. In this proposed work, the textural features of digital mammographic images are extracted from the Regions of Interest (ROIs) using Gray Level Co-occurrence Matrix. An optimal feature selection technique, namely Correlation-based High Distinction Feature Selection (CHDFS) is attempted to select the most significant features alone to classify the ROIs as Normal, Benign and Malignant. These selected features employed for categorization. With Support Vector Machine (SVM) and Artificial Neural Network (ANN) classifications algorithms, the validation was carried out. The investigational results illustrate that the proposed method be capable of producing better classification performance than the existing feature selection techniques like Correlation-dependent Feature Selection (CFS) and Principle Component Analysis (PCA).