White shark beetle optimizer enabled deep fuzzy clustering for feature selection and big data clustering in MapReduce framework

Abstract

Big data analytics have gained substantial attention over traditional data-processing methods, as they excel in uncovering hidden patterns and correlations within massive datasets, commonly referred to as big data. Advancements in information technology and the rapid expansion of the web have significantly increased the volume of data generated and utilized in everyday life. Moreover, traditional methods often struggle with efficiency and accuracy. These challenges are crucial to address, as the era of big data is transforming various domains, from research to real-world applications, where accurate analysis is critical. This study introduces White Shark Beetle Optimizer + Deep Fuzzy Clustering (WSBO+DFC), an innovative method designed to efficiently process and analyze big data. At the beginning, the input big data is retrieved from the database and transmitted to the MapReduce framework for processing. The MapReduce architecture has two phases, namely the mapper phase and the reducer phase. In the mapper phase, key-value pairs are generated from the dataset, providing structure to the previously unstructured data. This phase consists of multiple mappers, where feature selection is performed using Support Vector Machine (SVM) and Recursive Feature Elimination (SVM-RFE). To optimize the weight parameters of SVM, the proposed White Shark Beetle Optimizer (WSBO) is employed. Alternatively, in a reduced phase, the entire selected features are merged. After that, the fused features are subjected to big data clustering, which is conducted by utilizing Deep Fuzzy Clustering (DFC). The weight update process within DFC is guided by the WSBO, which is developed through the integration of the White Shark Optimizer (WSO) and the Dung Beetle Optimizer (DBO). The developed method achieved a Maximum accuracy of 89.87% and a maximum DB Index of 0.995.