Feature reduction from correlation matrix for classification of two basil species in common genus

Abstract

© 2019 IEEE. This research proposed ways with comparison results for feature selection and reduction for plant's leaf classification based on a key concept that features in a data set may include weakly relevant or redundant features. Six classifiers of support vector machine (SVM) model are demonstrated with ten features of about 320 leaves of two basil species sharing common genus. Plant species in a common genus typically have various aspects of similarity in their leaf features and this is our challenge in the way whether feature reduction should be done. Feature reduction provides the decrease in processing time in many cases, but it can easily reduce classification performance in terms of accuracy rate. According to our proposed techniques, an optimal feature reduction can still obtain while we still gain a perfect classification of 100 percent of accuracy.