Improved Coefficient Vector Based Grey Wolf Optimization Algorithm with Ensemble Learning for Imbalanced Data Classification

Abstract

Data mining is aprominent research field, where the greatest endeavor lies in the class imbalance which is recently concentrated by various researchers. The various real-world applications such as face recognition, cancer diagnosis, fraud identification etc, imbalanced data is occurs recurrently as one type of datasets. The accuracy enhancement is considered to be the most important factor for minority class recognition for imbalanced dataset. Researchers have proposed various approaches for mitigating the class imbalance problem. A Modified Step size based Glowworm Swarm Optimization algorithm (MSGSO) with Improved Regularization based Convolutional Neural Network (IRCNN) is utilized for feature selection and classification correspondingly in case of earlier research. Conversely, remarkable predictions are achieved through single classifier for specific problem only, but essential class data are discarded 0r sometime causes over-fitting. An Enhanced Coefficient vector   based Grey Wolf Optimization (ICGWO) Algorithm with ensemble classifier is greatly utilized for classification in this proposed work. Synthetic Minority Oversampling TEchnique (SMOTE) with Locally Linear Embedding (LLE) algorithm is used primarily for oversampling. Meanwhile, the selection of optimal features is achieved through Improved Coefficient vector   based Grey Wolf Optimization (ICGWO) Algorithm by means of  classifier accuracy rate improvement. The combination of Improved Regularization based Convolutional Neural Network (IRCNN), Enhanced Weighted Support Vector Machine (EWSVM) and k-Nearest Neighbour (k-NN) classifiers are added advantage for classification depending on the features designated. The notion of manifold individual learners and a combined stratagem is involved for achieving better outcomes than every distinct learner. The proposed research methodology outperforms well pertaining to various factors such as accuracy, precision and f-measure.