Desktop version

Home arrow Computer Science

  • Increase font
  • Decrease font


<<   CONTENTS   >>

Fuzzy Support Vector Machine with SMOTE for Handling Class Imbalanced Data in IoT-Based Cloud Environment

INTRODUCTION

Internet of Things (IoT) is assumed as an interlinked network of modern sensors that has minimum storage as well as computing ability. IoT in conjunction with cloud computing (CC) has enormous benefits like memory and adequate computing energy, which facilitates the required services such as medical sector [1, 2] and smart cities. Hence, observation as well as interacting with patients remotely is highly essential in this approach. In addition, the requirement to offer minimum cost, maximum superiority, and patient-based smart healthcare for the individuals is developed. The evolution of IoT [3] and CC models [4] are required in real-time, modern, and remote medical services for smart cities. Moreover,

~99

the combination of IoT and CC methods offers unique and tremendous benefits in smart healthcare monitoring. Recently, humans in smart cities have permission to apply modern sensor devices as well as latest mobile techniques. In a smart city, identifying a medical expert, hospitals, and pharmacies are highly a dark room. Also, a patient suffering from severe disease cannot move quickly to hospitals.

In order to resolve these issues, a smart healthcare-monitoring approach has been developed by combining the available resources which enhance the superiority and accessibility of medical services. Using the smart healthcare-monitoring model, the healthcare-oriented multimedia signals are transmitted from modern sensing devices and mobile tools to offer periodic guidance and qualified medical services to users. These medical data as well as signals are massive in size and complex to manage due to the complexity. The clinical sector has been introduced with massive requirements for industrial sector.

Even though IoT provides immediate and better treatments, it also meets maximum economic reviews for government as well as private sector. Recently, modern clinical centers have found a competition between diverse medical providers in offering better and comfortable facilities with good accuracy, dependability, and minimum expense [5]. Thus, the combination of CC and IoT in healthcare concentrates in better study and services. Various models of IoT tools have been developed in the field of healthcare such as portable tools like BP devices, portable insulin syringe, stress tracking machine, weight observing and fitness tools, hearing devices, and EEG and ECG monitors.

Though it has massive advantages, mobile approaches and ICT have resulted in tremendous benefits that are carried out in medical filed. Remote patient-monitoring function is performed using wireless and ubiquitous sensing model. A practical health-monitoring framework termed as Healthcare Industrial IoT has been projected in Ref. [6]. The newly deployed method holds vital capabilities for examining patient s medical information and lowers the mortality rates. Also, it gathers the patient details under the application of diverse healthcare tools and sensors. Followed by, privacy is considered as a major issue in the transmission of patient’s medical data to CC by doctors. Moreover, the theft action or clinical flaws can be identified by physicians using the strategies such as signal enhancement and watermarking methodologies. Gope and Hwang [7] applied the properties of IoT method to body sensor network (BSN). Here, patient will be observed with the help of tiny-powdered as well as light-weight sensor networks. Also, it serves an effective security and saves the patient details; a protective IoT-based medical network is described named as BSN-Care.

A practical health-monitoring approach is presented in Ref. [8] for distant heart patients under the variables. It is designed for offering the interface among a physician and patients and it facilitates two-way interactions. In addition, the application of BSN can be improved by providing warning alerts for serious cases, which are transmitted to remote users. An IoT-based, noninvasive monitoring system is presented in Ref. [9] for urban medical system. The structure of u-healthcare is composed of BSN; modern medical server and hospital system are considered to be vital units that describe the model. Hussain et al. [10] utilized a patient-based sensing method for aged and handicapped persons. The key purpose of this approach is to offer the service-centric response in serious condition of a user. Kim and Chung (2015) presented a serious case-monitoring technology with the help of context motion observation for patients who suffer from chronic diseases. It examines the recent condition of a patient on the basis of contextual data and offers essential data by investigating the patient’s lifestyle.

Catarinucci et al. [11] introduced latest techniques in IoT-based medical applications. It is surveyed that the existing platform, application, and trends in IoT-based medical solution are significant. Xu et al. [12] managed the heterogeneity issue of data type in IoT environment under the application of data model. Furthermore, resource-driven data-accessing technology has been developed for gaining and computing IoT data exclusively for the purpose of enhancing permission of IoT resources. Also, IoT-based system is applicable in handling critical situations and has been established to show the way of collection, combination, and interoperate IoT data elasticity. Box et al. [13] deployed and executed smart- home-based environment like iHome Health-IoT. It is embedded with open-platform relied intelligent clinical and improved connectivity for integration of tools and services. In addition, modern pharmaceutical packages as well as biomedical sensor-based tools are embedded in presented approach. Maia et al. [14] depicted an EcoHealth, a web middleware model for correlating persons with experts with the help of portable body sensors. It combines data attained from heterogeneous sensors. Then, the concatenated data is applied to send the alert message about patient’s state and significant symbols practically under the application of Internet service.

A viable solution is to apply the health conditions and find the name of a disease from health data gathered with the help of personal IoT devices [15]. A prototype has been developed for diagnosing disease by converting the disease diagnosing models into machine readable format. Hajihashemi et al. [16] introduced a novel model to process the affinity of two multiattribute time series on the basis of temporal model of Smith-Waterman, named as bioinformatics model. It mimics the complexities based on data irregularity and aggregation, which emerge while processing the sensed data. This model can be validated by using integrated data collected from electronic health records (EHR) and non-wearable’s fixed at apartments in the United States. The accomplished outcome implies that different health patterns have been examined to predict the anomalies in humans. Liu et al. [17] proposed a new motif discovery model for large-scale time series, named as MDLats. Motifs are identical sequences that are important in examining ECG patterns of cardiac patients. Hence, it reapplies the previous data to a greater extent and makes use of a relation among traditional data as well as current information.

This study devises an efficient fuzzy support vector machine (FSVM) with Synthetic Marginal Oversampling Technique (SMOTE) model called SMOTE-FSVM for class imbalance problem in IoT- and cloud-based disease diagnosis. The proposed SMOTE-FSVM model initially involves data collection, upsampling, and data classification. At the first stage, the data collection process takes place using IoT devices connected to patients and the data is transmitted to cloud server. Then, in the second stage, SMOTE-based upsampling of marginal data samples takes place by the generation of synthetic data. Finally, in the third stage, the medical data classification process is carried out using FSVM model. The proposed SMOTE-FSVM model has been assessed using PIMA Indians Diabetes dataset and the experimental validations are investigated under distinct performance measures.

 
<<   CONTENTS   >>

Related topics