A Novel Crow Search Optimization Based Feature Selection With Optimal DNN for Big Data Classification

Abstract

Big data analytics has become popular due to its applicability in various real-time applications. To attain better performance, big data can be analyzed using the machine learning and deep learning models at recent times, various domains have started to deal with big dataset which involves numerous sets of features. Repetitive and unwanted features, which reduces the classification performance can be removed using Feature Selection (FS) models. The conventional FS models do not have adequate ability to handle big dataset and filter effective results in limited time duration. FS is also regarded as a key procedure for boosting the effectiveness of big data analytics methods. Big data has many properties and requires extensive calculation. Hence, feature selection methodologies using metaheuristic optimization algorithms were adopted to choose optimum set of features and thereby improves the overall classification performance. To better categorize large datasets, a novel feature selection model is created in the MapReduce framework. Oppositional Crow Search (OCS) is an optimization-based feature selection approach that is used in the suggested model. Additionally, a Deep Neural Network (DNN) model based on Black Widow Optimization (BWO) is used to categorize the big data using the selected attributes. The creation of feature selection processes based on OCS and parameter tuning processes based on BWO considerably improves classification outcomes.