MapReduce-based big data classification model using feature subset selection and hyperparameter tuned deep belief network

Surendran Rajendran,Saleh Alghamdi,Youseef Alotaibi,Osamah Ibrahim Khalaf

doi:10.1038/s41598-021-03019-y

Abstract

In recent times, big data classification has become a hot research topic in various domains, such as healthcare, e-commerce, finance, etc. The inclusion of the feature selection process helps to improve the big data classification process and can be done by the use of metaheuristic optimization algorithms. This study focuses on the design of a big data classification model using chaotic pigeon inspired optimization (CPIO)-based feature selection with an optimal deep belief network (DBN) model. The proposed model is executed in the Hadoop MapReduce environment to manage big data. Initially, the CPIO algorithm is applied to select a useful subset of features. In addition, the Harris hawks optimization (HHO)-based DBN model is derived as a classifier to allocate appropriate class labels. The design of the HHO algorithm to tune the hyperparameters of the DBN model assists in boosting the classification performance. To examine the superiority of the presented technique, a series of simulations were performed, and the results were inspected under various dimensions. The resultant values highlighted the supremacy of the presented technique over the recent techniques.

Highlights

In recent times, big data classification has become a hot research topic in various domains, such as healthcare, e-commerce, finance, etc
This study focuses on the design of a big data classification model using chaotic pigeon inspired optimization (CPIO)-based feature selection with an optimal deep belief network (DBN) model[14]
The Harris hawks optimization (HHO)-based deep belief network (DBN) model is derived as a classifier to allocate appropriate class labels

Summary

Introduction

Big data classification has become a hot research topic in various domains, such as healthcare, e-commerce, finance, etc. This algorithm includes optimization technologies (such as GDA and PSO algorithms), scientific modeling or other predictive technologies (such as NN, NB, roughset, DT), decision analysis technologies (such as multicriteria decision, gray decision, and so on), performance assessment technologies (for example, fuzzy comprehensive evaluation and data envelopment analyses). The drawback of higher dimension datasets includes redundant data, higher module construct time, and degraded quality, which makes information analyses highly complicated To resolve these problems, the selection of features is employed as a major preprocessing phase for choosing subsets of features from a large dataset and increases the accuracy of clustering and classification models, which triggers foreign, ambiguous, and noisy data e limination[5]. This study focuses on the design of a big data classification model using chaotic pigeon inspired optimization (CPIO)-based feature selection with an optimal deep belief network (DBN) model[14]. This selected k learner will take part in the phase in which this presented MANOFS technique is incorporated

Methods

Results

Discussion

Conclusion