Novel feature selection technique with deep learning for big data classification

Abstract

In recent years, big data become more popular among the public and business enterprises. Big data became a suitable and effectual form of services, resources and applications. Big data is a collection of data which could not be collected, handled, and processed by traditional software models in a particular duration. Big data analytics has become a hot research topic due to its applicability in various real time applications. The recent advent of machine learning and deep learning models can be applied to analyze big data with better performance. At recent times, various domains have started to deal with big dataset which involves numerous set of features. Feature selection models intend to remove noise, repetitive and unwanted features which reduce the performance of the classification process. The conventional FS models do not have adequate ability to handle big dataset and filter effective results in limited time duration. At the same time, Feature Selection (FS) is considered a major process used to enhance the efficiency of big data analytics techniques. Since big data involves numerous features and necessitates high computational time, feature selection methodologies using metaheuristic optimization algorithms can be adopted to choose optimum set of features and thereby improves the overall classification performance. Metaheuristic optimization algorithms can be used to design effective feature selection methodologies. From the state of art literature, and it has been found that Meta heuristic algorithms perform better compared to other wrapper based techniques for FS. However, popular techniques like Genetic Algorithm (GA), Particle Swarm Optimization (PSO) algorithm suffers from slow convergence and local optima problem. These problems have been seen to solve using later generation algorithms like Firefly heuristic and Fish Swarm Heuristic. newline