Feature subset selection for cancer detection using various rank-based algorithms

摘要

Feature selection in data mining is the process of identifying the profitable features that are more significant in giving accurate results. Feature selection approaches like filter method and wrapper method are used here to get the more significant attributes. These methods generate the list of highly important attributes by using various ranker algorithms like correlation, relief-F, information gain, Gini index and classifiers like OneR, support vector machine, naive Bayes, random tree. In this, we are using ranker methods to perform feature selection on breast cancer analysis. Various experiments have been carried out on breast cancer Coimbra dataset using different classifiers to predict the accuracy. The crucial attributes are identified using feature selection methods, analysed for both balanced and unbalanced datasets and classified using OneR classifier.