Butterfly Optimization-Based Feature Selection with Gradient Boosting Tree for Big Data Analytics in Social Internet of Things


The fundamental objective of Fourth Industrial Revolution, deployment and extensive application of Internet of Things (IoT) system, resulted in the enhancement of connectivity among humans, human to object, and object to device, reinforcing an emergence of hyper-connected public community [1]. Based on the statement, it is desired that a fresh metric of fusion services appear after the interconnecting massive smart tools. Besides, in hyper-connected society, intellectualized objects will not apprehend the human situation, be aware of their needs (Context Aware), rather independently provide and recommend the better solution; however, it supports in taking control measures as the human requirements [2]. The domain that helps to develop creative and effective service at a rapid pace (new measure) perform the extension of previous service (Tapped Value), which is meant to be Social IoT (SIoT) model.

Based on Ref. [3], the SIoT is referred to as the IoT in which the objects are applicable to develop social combinations with each other independently. In addition, SIoT concept shows the environment that enables people as well as objects to communicate inside a social structure of associations [4]. SIoT platform supports the organization by accomplishing diverse and massive amount of data, which has been collected named as Big Data that is said to be a smart service by processing a better interference [5]. It refers to the dynamic format of mutual data along with the objective of required service that exceeds a distribution and reading level of a text, image, and so forth in previous social media. Furthermore, the range of Social Network Service has been upgraded from single targeted to corporation targeted, which is often with IoT, and finally activates the business functional collaboration [6].

SIoT is simulated to be one of the well-known concepts for massive state-of-the-art, rapidly developing models, and some of them are IP-enabled embedded devices as well as smart objects, short- and long-range communication models, data accumulation, examination, computation, and visualization devices from big market giants that have various benefits in network direction, reliability, estimation of objects, security, service composition, object identification, behavior categorization, as well as detection. According to the parameters of SIoT, different types of studies were developed about developing a method for SIoT service environment and significance of the domains applied. In addition, the related works are carried out [7] such as performing the interference, which depends upon a research in data gathering process inside the SIoT platform as well as the gathered big data, and research to maintain the efficiency of data and security.

Building the IoT-based models and corresponding solutions is considered to be the major challenging task. Flence, IoT deals with the persistent accumulation as well as data distribution to a general objective [8]. In IoT, data means the parameter values like variables or integer measures; and it depicts some specialized conditions that are accomplished [9, 10]. IoT networks enable little functionality by using a previous interface. Developers have highly concentrated in finding threats while identifying and combining the data inside IoT [11]. Thus, SIoT is said to be a massive social network that connects people to people, people to objects, and objects to objects [12]. Therefore, developing the opportunities provide major challenging issues to data computation for the purpose of enhancing data collection, noise elimination, storage, as well as to perform actual analytics [13]. Furthermore, the big data contains diverse standards and devices by relational database vendors, and it is also applied for data collection and data analysis [14]. Here, “big data and SIoT are the exact depiction of social systems and IoT to simplify human development” [15]. Different types of feature selection (FS) model have been projected and divided into two classes, namely, filter and wrapper models. Initially, filter-enabled method performs the filtration task that has to be computed in prior to classify the data because of the random application of classification models [16].

To resolve the big data issues existing in SIoT, this chapter devises a new big data analytics method in SIoT using butterfly optimization-based feature selection with a gradient boosting tree (GBT) technique called BOAFS-GBT. The proposed BOAFS-GBT model primarily performs feature selection using the BOAFS model, which selects a constructive collection of features from the big data. Thereafter, the GBT model is used for the classification of the feature-reduced data into several classes. In addition, big data Hadoop framework has been employed for big data processing. The outcome of the BOAFS-GBT model has been validated against three datasets under diverse aspects.

< Prev   CONTENTS   Source   Next >