A Hybrid Nature-Inspired and Deep Learning Framework for Feature Selection and Classification in Big Data

Abstract

The rapid expansion of Big Data has led to an increasing demand for effective, scalable methods for feature selection and classification. This paper introduces a hybrid framework combining Nature-Inspired Algorithms (NIAs) with Deep Learning techniques to address these challenges. The framework consists of four phases: Phase 1 - data preprocessing using an Enhanced Quality Rules Discovery Model to improve data quality; Phase - 2 feature selection through a Novel Nature-Inspired Algorithm to identify the most relevant features; Phase - 3 classification using a Hybrid Nature-Inspired Algorithm combined with data mining methods; and Phase - 4 advanced classification using Deep Learning Neural Networks integrated with the Hybrid Nature-Inspired Algorithm to boost accuracy. The proposed framework effectively reduces computational complexity while enhancing classification performance on large datasets. The outcomes show how accurate the framework is, which makes it appropriate for a variety of Big Data applications.