An Enhanced Outlier Detection Approach for Multidimensional Datasets Using a Synergistic Firefly and Grey Wolf Optimization-Based Method

Abstract

Outlier identification and elimination are essential preprocessing steps for data analysis tasks such as clustering, classification, and regression. The accuracy of data analysis outcomes may be compromised if outliers are not adequately addressed. Detecting outliers is particularly challenging when they are characterized by unusual combinations of multiple attributes. Furthermore, the presence of outliers can impact various data processing activities, necessitating either the reduction of outlier influence or their complete removal. Outlier detection in multivariate data presents a complex process that becomes increasingly difficult when dealing with high-dimensional datasets. Consequently, this study focuses on the identification of such outliers in multivariate datasets using intelligent techniques. In the proposed approach, outliers are detected using an Improved Neural Network (INN), where the hidden neurons are tuned by a novel Synergistic Firefly-Grey Wolf Optimization (SF-GWO) algorithm. This algorithm combines the strengths of the Firefly Optimization (SFO) and Grey Wolf Optimization (GWO) techniques to maximize accuracy. The unique method results in enhanced classification model performance, reduced computation time, and increased classification accuracy. The proposed model has been evaluated and compared with well-established traditional techniques, demonstrating its effectiveness in addressing the challenges of outlier detection in multidimensional datasets.