Volume: 19 Issue-01 (January-June) 2024

---

IMBALANCED DATA CLASSIFICATION USING IMPROVED GREY WOLF OPTIMIZATION AND ENHANCED ARTIFICIAL NEURAL NETWORK ALGORITHM

ABSTRACT

An imbalanced dataset refers to a situation where the distribution of classification classes is not roughly equal, with one class containing significantly more samples than the others. In such cases, because the bigger size of the majority class has a stronger effect, classifiers may perform poorly for the minority class but have excellent predicted accuracy for the majority class. To overcome the problem, in the existing system, Synthetic Minority Oversampling Technique (SMOTE) with Local Outlier Factor (LOF) is introduced. However, it encounters challenges related to misclassification error rates stemming from noise in the provided dataset, leading to lower accuracy. To address these issues, this study introduces the Improved Grey Wolf Optimization (IGWO) and Enhanced Artificial Neural Network (EANN) algorithm. The process begins with the collection of datasets, followed by pre-processing utilizing the K-Means Clustering (KMC) algorithm. The primary aim is to enhance classification accuracy by addressing missing values. Then the datasets are taken into class balance process via SMOTE-LOF technique. It performs oversampling and undersampling alongwith outlier detection process. After that, the balanced datasets are taken into feature selection process which is done by using IGWO algorithm. It produces superior fitness values characterized by increased classifier accuracy and reduced execution time. The classification process is ultimately carried out using the EANN algorithm, which yields improved accuracy. Experimental findings indicate that the suggested framework markedly enhances the performance of balanced datasets. With better accuracy, precision, recall, F-measure, Area Under Curve (AUC), and execution time than previous algorithms, the findings reveal that the IGWO-EANN method, as presented, performed better.

FULL TEXT

PDF