International Journal of Communication Networks and

Information Security
ISSN: 2073-607X, 2076-0930
Volume 14 Issue 03 Year 2022 Page 313:329

_____________________________________________________________________________________________________________________________________________________

5G with Fog Computing based Privacy System in Data Analytics for

Healthcare System by AI Techniques
Encarnación Maria Torres Solís
Department of Post Grade, Universidad Cèsar Vallejo, Perú, South America
Juan Carlos Cotrina-Aliaga
Faculty of Human Medicine, Universidad Privada San Juan Bautista, Chincha, Perú, South America
[email protected]
Fidel Castro-Cayllahua
Associate Professor, Universidad Peruana los Andes, Perú, South America
Dr. Severo Simeón Calderón Samaniego
Principal Professor, Universidad Peruana los Andes, Perú, South America
Betsy Nordie Pardo Alarcon
Department of Post Grade, Universidad Cèsar Vallejo, Perù, South America
Yoni Magali Maita Cruz
Professional School of Human Medicine, Universidad Privada San Juan Bautista, Perú, South
America

Article History Abstract

Received: 13 July 2022 Fog computing architecture is an extended version of the cloud
Revised: 20 September 2022 computing architecture to reduce the load of the data transmission and
Accepted: 26 October 2022 storage in the cloud platform. The architecture of the fog increases the
performance with improved efficiency compared with the cloud
environment. The fog computing architecture uses the 5G based
Artificial Intelligence (AI) technology for performance enhancement.
However, due to vast range of data availability privacy is challenging in
the fog environment. This paper proposed a Medical Fog Computing
Load Scheduling (MFCLS) model for data privacy enhancement. The
developed architecture model of optimization-based delay scheduling
for task assignment in the fog architecture. The healthcare data were
collected and processed with the 5G technology. The developed
MFCLS model uses the entropy-based feature selection for the
healthcare data. The proposed MFCLS considers the total attributes of
13 for the evaluation of features. With the provision of service level
violation, the fog computing network architecture will be provided with
reduced energy consumption. The developed load balancing reduced the
service violation count with the provision of desired data privacy in the
fog model. The estimation of the time frame is minimal for the proposed
MFCLS model compared with the existing DAG model. The
performance analysis expressed that SLRVM and ECRVM achieved by
the proposed MFCLS are 28 and 43 respectively. The comparative
examination of the proposed MFCLS model with the existing DAG
model expressed that the proposed model exhibits ~6% performance
enhancement in the data privacy for the healthcare data.
Keywords: Fog computing, Load balancing, Optimization, data
CC License privacy, energy consumption
CC-BY-NC-SA 4.

Available online at: https://ijcnis.org

 5G with Fog Computing based Privacy System in Data Analytics for Healthcare System by AI Techniques

1. Introduction

The term healthcare refers to a system that entails the enhancement of health-related services in order
to fulfill the clinical requests of the individuals. In healthcare services, patients, doctors, clinicians,
researchers and medical industries are all making an effort to maintain and restore health records [1].
In recent years, with the remarkable development of technologies, data is continuously increasing day
by day in every sector including healthcare which in turn demands more and more data mining
applications. However, due to digitization of healthcare system, the medical organizations are
generating large amount of healthcare data [2]. In general, healthcare data consists of all records
related to health stored in a digital form. It may contain detailed information about patients’ medical
history, doctors prescribed notes, clinical reports etc. All these data are voluminous, high dimensional
and diversified in nature. Healthy decision making is a challenge in the modern era due to this rising
complexity of healthcare data [3]. Machine learning, data mining, and statistical techniques are the
key fields of study that enhance the ability of individuals to make the right decisions in order to
maximize the outcome of any working domain [3]. Human data analytical capability rate is much
smaller when compared to the amount of data that is stored [4]. This becomes even more critical when
it comes to healthcare domain as the number of available experts for healthcare data analysis is
comparatively less.
With the extension of the Cloud computing (CC) Cisco presented a Fog computing technology to
meet the requirement of the present era requirement such as 5G, Internet of Things, Artificial
Intelligence (AI) and so on [5]. The provision of services by the Fog computing (FC) comprises of
IoT users to perform data processing and storage. The technology of Fog computing is extended
towards the cloud computing technology for the stored data in the local node which increases the
burden towards the data transmission and reception in the cloud [6]. In this manner, Fog computing
increases the efficiency of the cloud network performance. Additionally, with the fog computing the
data processing rate in the cloud is reduced with the transfer of information in the cloud storage and
analysis of the data. The provision of services to the fog computing architecture decreases the latency
and network traffic in the environment. The position of the fog computing relies on the fog node over
the complete network [7]. The fog computing architecture comprises of the controller devices which
act as a switch to the deployed fog nodes for the targeted area in the environment. The data is
generated with the use of IoT devices involved in the analysis of information in the nodes. The fog
node analyses the data those are transmitted in the cloud with the minimized pressure in the cloud [8].
The reduction of burden in the cloud minimizes the latency and overcrowding in the cloud jobs. The
architecture of the cloud computing comprises of the fog computing environment to provide
information in the centralized resource access. In other hand, with the decentralized manner fog
computing is estimated for the provision of access.
Fog computing architecture exhibits the privileged cloud computing environment to increases the
effectiveness and efficiency of the network with the increased number of devices in the network [9].
The fog computing technology comprises of the innovation concept to evaluate the IoT devices
storage space and computing ability. The conventional fog computing technology for the medical
application comprises of the segments in three tier manner such as IoT Devices (Lower Tier) -
responsible for patient information collection, Fog- nodes (Middle Tier), and Cloud (Top Tier). The
IoT layer of the fog comprises of the objects those are wearable devices for the data collection such as
cameras, vehicles, sensors, home appliances those leads to massive collection of data [10]. The cloud
environment provides the assistance in the foundation level for the information upholds to evaluate
the framework. However, the vast range of applications demands for the reduced latency to improve
the performance of the fog [11]. In reduction of the latency in the fog computing environment the
services are reduced with the overcrowding in the network. To evaluate the massive volume of the
data enormous data need to be processed and handled with the effective storage information handling
in the dynamic environment for the broadband information frames. In IoT application the distributed
cloud computing environment demands for the data privacy for the cloud storage those are capable to
achieve the effective Quality of Service (QoS) based on the demand [12]. The provision data privacy
leads to higher computational capability with the lot of service operation to evaluate large data

Available online at: https://ijcnis.org 314

 International Journal of Communication Networks and Information Security

volumes. The conventional IoT application subjected to different challenges due to lack of network
capacity, latency, location awareness and support. To withstand the issues related to fog computing
cloud solution need to be leverages to provide large scale connectivity in the IoT devices to ensure
significant solution constraints [13]. With the virtualized platform in fog computing comprises of the
resource pool for the business model storage, end-user networking and cloud data centres.
This paper developed a medical data fog architecture model for the task scheduling in the network
with the task assignment. The developed model involved in the estimation of the entropy features with
the job scheduling and assignment of tasks in the network. This paper is organized as: Section 2
provides the existing literature and proposed methodology presented in section 3. The results obtained
for the proposed MFCLS model is presented in section 4 and overall conclusion is presented in
section 5.

2. Related Works

Data mining can be defined as a process of collecting, analysing and storing data in order to extract
useful and interesting information from it. A few of the key tasks in data mining involve
classification, prediction, association analysis, clustering, and hybrid approaches [14]. Notably, data
mining is now considered as a multidisciplinary area. In recent years, since the amount of digitally
recorded healthcare data has grown, so more and more data mining-based applications are developed
by researchers and scientists in order to improve healthcare services. This huge amount of healthcare
data can be mined in order to discover new patterns and value-added insights to enhance the quality of
healthcare. In the medical field, clinical diagnostic systems are developed to facilitate medical
practitioners in making healthcare decisions based on health records. According to [15], electronic
health records (EHRs) empower the clinicians by integrating patient health history for planning safe
and proper treatment. With a comprehensive evaluation of healthcare medicine based on consideration
of six tasks such as prognosis, management, monitoring, screening, and treatment.
In [16] designed a preterm birth model based on the pregnant women dataset those are considered as
diabetes mellitus or gestational diabetes mellitus. The logistic regression and Support Vector Machine
are used for the prediction purpose. With the developed a multi-parametric magnetic resonance
imaging (MRI) machine learning model to predict the risk of breast cancer. In this regard, eight
classifiers were used on MRI data of the breast in order to rank the features prediction with the
pathologies responsible for treatment of breast cancer. In [17] have introduced a multilevel perceptron
model for survival prediction. The introduced model predicts the survival of non-small cell lung
cancer patients with multi-layer neural network. Relief and Recursive Feature Elimination algorithms
were implemented to extract the features from clinical data in order to increase the accuracy of
prediction model. The suggested an intelligent-based expert system for identifying Type2 Diabetes
Mellitus (T2DM) patients with the help of electronic health record (EHR) data. For this purpose,
different machine learning algorithms such as Random Forest (RF), logistic regression (LR), Decision
Tree (DT), naïve bayes, K-Nearest Neighbour (KNN) and Support Vector Machine (SVM) were
applied.
In [18] presented a Computer aided diagnosis (CAD) system integrated with the fuzzy k-nearest
neighbour (FKNN) classifier to detect and diagnosis thyroid diseases. A thyroid prediction system by
employing various machine learning techniques including Decision Trees, Random Forest, Support
Vector Machine, ANN and logistic regression. In [19] suggested a heart disease prediction system to
detect the presence of cardiovascular disease based on data mining approaches. The three learning
classifiers namely, Naïve Bayes, J48 Decision Tree and Bagging algorithms are used in this
diagnostic model. According to big data analytics are applied to healthcare data acquired from several
sources to gain valuable insight from it. The authors evaluate historical health data and assess medical
quality of service from multiple US states by applying new healthcare-specific analytical software to
predict future healthcare activities. Despite this, no solutions for storing big data have been developed.
In [20] described the application of data mining techniques in healthcare and biomedicine. The study
first examines three different learning methods namely, classification, clustering and association, and
then turns its attention to their application in healthcare industry. This is accomplished by a brief
discussion given on each task with their advantages and disadvantages. Some of the most important
medical aspects that discussed in this paper are: prediction of healthcare costs, disease diagnosis and

Available online at: https://ijcnis.org 315

 5G with Fog Computing based Privacy System in Data Analytics for Healthcare System by AI Techniques

prognosis, and discover hidden patterns from healthcare data. However, in most cases, an accurate
data mining model is not suitable for clinical environment.
In [21] developed a energy-efficient resource scheduling scheme for the reduction of energy
consumption in the IoT based heterogeneous environment. The developed model uses the optimized
approach for the data transmission scheduling to reduce latency with the increased network traffic in
the network. The simulation analysis expressed that proposed model increases the energy utilization
rate by 18% and reduces the overall execution cost of network by 15%. Also, the developed model
increases the overall lifetime of the network by 1.17% and 5% increase in the sensor lifetime through
the scheduling scheme. In [22] proposed a heuristics method for the scheduling of tasks in the cloud
platform. Consequently, the developed model comprises of the effective task scheduling for the
complicated procedure based on the consideration of different assumptions. The evaluation is based
on the effective task scheduling with the FCFS (First Come First Serve), Minimum Completion Time
(MCT), Minimum Execution Time (MET). In [23] developed a task scheduling scheme for the
resource allocation in the fog network for the developed application to perform task scheduling and
resource redeployment. The analysis expressed that developed model decreases the delay 39% 11%
11% 21% 18% with the increased tasks scheduling in fog nodes.

3. Medical Fog Optimization for Load Balancing

Let 𝑃𝑐 be the energy consumed for processing instruction in a task and 𝑃𝑡 be the energy used for
transmission of data. Then choose to perform the task on fog, then the energy consumption is given by
the following formula.
𝑆
𝐸𝑓𝑜𝑔 = 𝑃𝑡 + (1)
𝑀
Where 𝑆 denoted as the byte of data those are processed with the network bandwidth 𝑀. Assume the
generates output is denoted as 𝑆 ′ those are processed effectivey where 𝑆’ is less than 𝑆. Now the
smart devices are performed based on the estimation of energy consumption task defined as in
equation (2)
𝑆
𝐸𝑚𝑜𝑏𝑖𝑙𝑒 = 𝑃𝑐 ∗ 𝐼 + 𝑃𝑡 + (2)
𝑀
Where, the task execution in the numerals are denoted as 𝐼, the saved energy is represented as (𝐸𝑠 )
presented in equation (3) and (4)
𝑆 𝑆′
𝐸𝑠 = 𝑃𝑡 ∗ − 𝑃𝑐 ∗ 𝐼 − 𝑃𝑡 ∗ (3)
𝑀 𝑀
𝑆
𝐸𝑠 = [𝑃𝑡 ∗ ]− 𝑃𝑐 ∗ 𝐼 − [𝑃𝑡 ∗ (𝑆 ∗ 𝑀)] (4)
𝑀
𝑠,
Where 𝑀 = 𝑠
and𝑠 are called the compression ratio as in equation (5)
𝐸𝑠 𝑀 = 𝑃𝑡 𝑆(1 − 𝑀) − 𝑃𝑐 ∗ 𝐼 (5)
The saved energy in the Medical Fog Computing Load Scheduling (MFCLS) comprises of the
positive evaluation in the task offloading. At first, based on the requrest the CPU is bounded based on
the output memory. The Proposed MFCLS algorithm compriseso f the different instances for the
classification those are logrithimc algorithm, polynomial algorithm and time. With the fog computing
environment the fog computation is based on the reduction of power consumption with the reduced
frequency as presented in equation (6)
𝑃 = 𝑐𝑓𝑉 2 + 𝑃𝑠𝑡𝑎𝑡𝑖𝑐 (6)
In the above equation (6), the 𝑐 represented as the gate capacitance transistor, frequency of the medica
data is denoted as 𝑓 and the operation voltage and static performance is represented as 𝑉. The
required voltage for the normal and static operation is based on the frequency dependent with the
clocked circuit. With the privacy of the data the energy consumption of the data is denoted as 𝐸 with
the computation of the Power*Time in which E is proportional to 𝑉 2 as given in equation (7)
𝐸 ∝ 𝑉 2 (𝑉 ∝ 𝑓) (7)
The MFCLS algorithm comprises of the scheduling process for job processing based on the
consideration of the VM position. The VM is evaluated based on the consideration of the arbitrary
process those are applied with the different VM1, VM2 and VM3. The estimation of the arbitrary
value comprises of the cost 1 to 3 it reaches till 15. In figure 1 the architecture of the fog computing
model is presented.

Available online at: https://ijcnis.org 316

 International Journal of Communication Networks and Information Security

Figure 1. Architecture of Fog Network

If a particular arbitrary cost is taken say from 1 to 3, which are 15.

The structure of job 2 is presented below. Firstly, contemplate the second job:
1. To execute the Job 2 there are 3 options available in the VM.
2. The allocation of Job 2 is based on the obliged VM in the network environment
a. The slot those are free are implemented with the VM for the job request
b. The comparison of the performance in the available VMs the cost should be minimal in the
VM
c. The reduction of the time caused by the reduced energy consumption.
Through the consideration of the job the VM is assigned in the network.

Figure 2. VM assignment

In figure 3 presented about the assigned jobs for the VM in the proposed MFCLS. The
estimation is based on the computation of the tasks and jobs in the network.

Available online at: https://ijcnis.org 317

 5G with Fog Computing based Privacy System in Data Analytics for Healthcare System by AI Techniques

Figure 3. Assignment of Job in VMs

The medical data is more sensitive and stable with the selection of features in the created medical
datasets. In order to make this approach stable, we need to pay attention on the probability measures
of the attributes over the small subsets of data to understand the contribution of the attributes. Based
on the consideration of the scenario the entropy based technique need to be implemented for the
processing of large amount of data. In figure 4 data analytics model implemented for the proposed
MFCLS model is presented.

Figure 4. Process in MFCLS data analytics model

In figure 5 the communication interface for the proposed MFCLS model is presented. The developed
architecture model comprises of the information exchange between patient and doctor towards the
decision making process.

Available online at: https://ijcnis.org 318

 International Journal of Communication Networks and Information Security

Figure 5. Flow of fog architecture in the MFCLS

In the table 1 the assigned number of job cost for the VM is presented. The assigned number of job is
considered as 10 for the VM count of 3.

Table 1. Job cost for different VM

Job No VM job cost
VM1 VM2 VM3
1 1 3 7
2 3 4 1
3 2 3 2
4 1 2 3
5 4 2 5
6 3 1 4
7 11 10 11
8 4 7 8
9 10 3 7
10 7 2 5
The calculation is explained using the following architecture shown in Figure 4.1 and Table 1 shows
the costing of different VMs
The total cost of execution is:
𝑇𝑜𝑡𝑎𝑙 𝑐𝑜𝑠𝑡 = 𝐸𝑐 + 𝑇𝑐 (8)
Where:
𝐸𝑐 = 𝐸𝑥𝑒𝑐𝑢𝑡𝑖𝑜𝑛 𝑐𝑜𝑠𝑡
𝑇𝑐 = 𝑇𝑟𝑎𝑛𝑠𝑓𝑒𝑟 𝑐𝑜𝑠𝑡

3.1 Entropy based feature selection

The data cleaning operations like handling missing values in the attribute of the datasets, removing
redundant features/irrelevant features from the datasets, handling data consistency, features dimension
reduction etc. are performed with the proposed MFCLS evaluate the healthcare dataset (D). With the
proposed MFCLS scheme feature selection is performed with the discretise of the by Minimum-
information loss (MIL) for the 3 different subsets D1, D2 and D3 with data partitioning in the equip
class data distribution. The feature selection in the entropy based model for the estimation of the
features is presented in figure 6.

Available online at: https://ijcnis.org 319

 5G with Fog Computing based Privacy System in Data Analytics for Healthcare System by AI Techniques

Figure 6. Feature Selection with MFCLS

Equi-class Distribution
The proposed MFCLS model integrates the equi-class distribution for the healthcare medical dataset
evaluation with the training set of data.
In the feature selection process the proposed MFCLS model perform the classification presented as
follows:
1. D1 (1st sub-set): In this dataset, the 30% of data were randomly derived from the complete
healthcare data D with the elimination of the selected instances D.
2. D2 (2nd sub-set): In the random manner 30% of data were utilized for the selected data in D
with the distinct instances D2 in the dataset D.
3. D3 (3rd sub-set): The remaining data in the dataset D aprat from D1 and D2 are placed in the
D3. This dataset is completely independent form for analysis.
In the proposed MFCLS model the dataset (D) comprises of the different attributes 𝐴𝑖 , 𝑖 = 1, … . , 𝑛.
Consider the feature set denoted as 𝐹𝑠 = {𝐴1 , 𝐴2 , … … , 𝐴𝑛 } for the dataset 𝐷 those are classified in the
3 subsets 𝐷1 , 𝐷2 𝑎𝑛𝑑 𝐷3 for the three different features in the distinct data subsets. Consider the
feature set represented as 𝐹1 , 𝐹2 𝑎𝑛𝑑 𝐹3 with the value 𝐹𝑘 (𝑘 = 1,2,3) = 𝐹𝑠 . In the obtained set the
resulted features 𝐹 denoted as 𝐹1 , 𝐹2 𝑎𝑛𝑑 𝐹3 .
Consider the classification problem 𝑃 with the total attributes count of 𝑛 stated as 𝐴𝑖(𝑖=1,…..,𝑛) . The
feature set is represented as F with the value of 𝐹 = 𝐹𝑠 = {𝐴1 , 𝐴2 , … … , 𝐴𝑛 }.
1. for every subset data (𝐷𝑘 , 𝑘 = 1,2,3) do
2. begin
3. for each attributes in the data 𝐴𝑖(𝑖=1,…..,𝑛) do
4. begin
5. Compute the information gain the in the healthcare data 𝐴𝑖 (ie 𝐺𝑎𝑖𝑛(𝑆, 𝐴𝑖 ) as presented in
equation (9)
|𝑆𝑣𝑗 |
𝐺𝑎𝑖𝑛 (𝑆, 𝐴𝑖 ) = 𝐸𝑛𝑡𝑟𝑜𝑝ℎ𝑦 (𝑆) − ∑𝑣𝑗𝜖𝐴𝑖 |𝑆|
𝐸𝑛𝑡𝑟𝑜𝑝ℎ𝑦 (𝑆𝑣𝑖 ) (9)
where, 𝑣𝑗 (𝑗 = 1, … . 𝑘) represented as the attribute Ai with the computation of entrophy
𝐸𝑛𝑡𝑟𝑜𝑝ℎ𝑦 (𝑆) = ∑𝑐𝑚=1 𝑝𝑚 𝑙𝑜𝑔2 𝑃𝑚 where number of exapmles P is computed in the sample P. The
non-zero probability is denoted as 𝑆𝑚 with the class 𝑚, out of 𝑐.
end for
6. Compute 𝑟 = ((max(𝑆, 𝐴𝑖 ) − min _𝐺𝑎𝑖𝑛(𝑆, 𝐴𝑖 )𝑙𝑛, 𝑖 = 1, … . . , 𝑛
𝐺𝑎𝑖𝑛
7. For every healthcare data attributes 𝐴𝑖(𝑖=1,…..,𝑛) do
Begin

Available online at: https://ijcnis.org 320

 International Journal of Communication Networks and Information Security

8. If 𝐺𝑎𝑖𝑛 (𝑆, 𝐴𝑖 )(𝑖=1,….,𝑛) < 𝑟, then feature set need to be updated (𝐹𝑘 )as: 𝐹𝑘 = 𝐹𝑘− {𝐴𝑖 } //
discard the attributes 𝐴𝑖 from 𝐹
end for
end for
9. 𝐹 = 𝐹1 𝑈𝐹2 𝑈𝐹3
/*𝐹 comprises of the non-common and common attributes of the healthcare data 𝐹1 , 𝐹2 𝑎𝑛𝑑 𝐹3 . The
balanced dataset comprises of the subsets to 𝐹1 , 𝐹2 𝑎𝑛𝑑 𝐹3 are those are balances with the common
attributes 𝐹1 , 𝐹2 𝑎𝑛𝑑 𝐹3 increases. */. The assigned jobs for the propised MFCLS is presented.

Algorithm 1: Job Assignment with MFCLS

Input: List of Host (HL), List of Job (JL) and Deadline
JL.Decreasing with sorting_Optimization ()
for
Every job in JL, update list do
Minimum Power_Utilization ← Maximal power allocated
Assigned to host <--null
Sufficient job resources available
Estimate Power ← Power as (H,J)
if
Power < utilized power_then
Compute fitness (HL, Job, Deadline);
if fitness function is satisfied as fitness_Fxn
Optimize Host value
end if
end if
Return;
Optimize the HL & JL values
end
end for
With the proposed MFCLS laod scheduling with the optimization model for the job assignemtn is
presented.
Algorithm 2: Optimization for load scheduling
Input: Host List Optimization (HL) and Job list Optimization (JL)
Configure the ANN parameters such as Neurons (N): Epochs (E)
Performance Estimation with validation; Gradient; MSE; Mutation
Training of sequences with Levenberg Marquardt as (Trainlm)
Random division of data
for
each job compute host
Based on host ability the Group (G) compute based on categories
end
Initialize the data group and training using ANN
Set the parameters for training
Net = train (Groups, Jobs)
Return;
Train ANN for the allocation of job
End

4. Results and Discussion

The proposed MFCLS model comprises of the medical data processing based on the assigned job in
the resource scheduling. The parameters concentrated on the privacy improvement in the healthcare
data.

Available online at: https://ijcnis.org 321

 5G with Fog Computing based Privacy System in Data Analytics for Healthcare System by AI Techniques

Security Lifecycle Review (SLR): In the every VM system the task completion time is accomplished
based on the job model. The SLR in the calculated using the equation (10)
𝑆𝐿𝑅 = ∑𝑛𝑘=0 𝐽𝑜𝑏 𝑒𝑥𝑒𝑐𝑢𝑡𝑖𝑜𝑛 𝑡𝑖𝑚𝑒 (10)
Where,
n = number of jobs in the mode
SLVMR: The job is executed by the one VM at the time it provides the ration of the SLR to the
computed time as it is given in equation (11) as follows:
𝑆𝐿𝑅
𝑆𝐿𝑉𝑀𝑅 = 𝐽𝑜𝑏 𝑡𝑖𝑚𝑒_𝑜𝑛𝑒 𝑉𝑀 (11)
EC: The total energy consumed to execute the all tasks in the sequences.
ECR: The tasks are executed by the one VM at a time and its is computed using the formula (12)
𝐸𝐶
𝐸𝐶𝑅 = (12)
𝐸𝑛𝑒𝑟𝑔𝑦 𝐶𝑜𝑛𝑠𝑢𝑚𝑒𝑑

Table 2. Health care data for scheduling

Medical Data Attributes Class Number Sample Number
Number
Breast Cancer 9 2 698
Dermatology 32 6 367
Ecoli 7 7 334
Heart 12 5 293
Heart (Swiss) 12 4 121
Heart (Cleveland) 12 4 301
Hepatitis 18 2 153
Liver Disorder 6 1 343
Lung Cancer 55 3 31
Lymphography 17 4 147
Thyroid 5 2 213
Diabetes 8 1 754
Primary Tumor 16 21 337
Sick 27 2 3772

In the table 2 different datasets such as dermatology, breast cancer, sick and primary tumor, heart,
lung cancer and Hepatitis based on consideration of different datasets dimensions. The feature
selection is implemented and evaluated with the 14 datasets. The presented datset feature selected are
presented in table 3. The dataset for analysis is presented in the dataset defined as DN, number of
instances (NI), feature number (NF), selected features number (NSF) and selected features (SF).

Table 3. Feature Selection for medical data

Medical Data Number Number Number Selected Features
of of of
instances features Selected
in the features
original
datasets
Breast Cancer 698 9 6 1,3,4,5,6,9
Dermatology 367 33 21 2,3,4,5,6,7,9,13,14,15,16,19,20,21,22,26,27,28,29,30,33
Ecoli 334 8 6 1,2,3,5,6,7
Heart 293 12 6 2,3,6,9,10,11
Heart (Swiss) 121 12 3 9,10,11
Heart 301 12 7 2,3,8,9,10,11,12,13
(Cleveland)
Hepatitis 153 19 8 1,2,6,11,12,14,17,18,19

Available online at: https://ijcnis.org 322

 International Journal of Communication Networks and Information Security

Liver Disorder 343 6 2 3,5

Lung Cancer 31 57 18 1,3,5,9,13,14,15,20,21,25,26,38,41,45,48,50,56
Lymphography 147 17 10 1,2,7,8,9,11,13,15,16,18
Thyroid 213 5 5 1,2,3,4,5
Diabetes 754 7 4 2,6,7,8
Primary 337 16 11 1,2,3,4,5,7,9,10,13,15,16,17
Tumor
Sick 3772 28 8 1,8,10,14,15,20,24,29

For more clarity about the feature reduction work performed in this study, the important terminologies
that are used here are presented below.
• Original Features: - Features that are available in the original datasets (uploaded in the UCI
repository) are known as original Features. These are, indeed, decided by the experts (physicians) to
diagnose the diseases.
Selected Features: - Features selected from each original dataset by removing noisy
(irrelevant/redundant) features using suitable feature reduction approach, are known as selected
features.
Independent Features: - Features that are not dependent among themselves, are known as independent
features. Each such features are dependent with target feature, i.e., their contribution in decision
making is high enough.
Let us consider the disease dataset named Heart (Swiss) dataset for explanation. In UCI repository, 13
non-target attributes are decided as the processed features by the experts (out of 76 unprocessed
features). These 13 features are now treated in the present experiment as the original features. These
are renamed here as A1, A2, A3, …, A13 (in short, 1, 2, 3, …, 13) with description as follows.
1. Age (A1): Age of patient in years.
2. Sex (A2): 0 = male; 1 = female).
3. Tc (A3): Type of chest pain under 4 types
1 – Angina typical
2 – Angina atypical
3 – Non-anginal pain
4 - Asymptomatic
4. Bp (A4): Blood pressure measured with mmHg in the hospita;
5. Cholestral (A5): Serum cholesterol measured in mg/dl
6. Fbs(A6): Blood sugar in fasting if 120mg/dl
1 – True
0 - False
7. ECresults (A7): Electrocardiographic results in the resting states
0 – Normal
1 – ST – T abnormality in wave >0.05mV
2 – Estes Criteria for the left ventricular hypertrophy
8. Maxrate (A8): Maximal heart rate obtained
9. Angex (A9) – Angina induced by angina
0 – No
1 – Yes
10. Depeak (A10) – The induced depression for the rest relative exercises
11. Slope (A11) – The exercise peak in the segmented T-wave
1 – upslope
2 – Flat
3 – Downslope
12. Cofl(A12) – Fluoroscopys major vessels ranges from 0 – 3
13. Blodis (A13) – Thalassemia
1 – Normal
2 – defect fixed

Available online at: https://ijcnis.org 323

 5G with Fog Computing based Privacy System in Data Analytics for Healthcare System by AI Techniques

3 – Defect those are reversible

Thus, the selected features by the adopted hybrid feature selection approach are namely, 9 (A9), 10
(A10) and 11 (A11).
The following are used for the comparison of the effective hybrid feature selection approach
1. Three states of the learners used – C4.5, RIPPER, naïve Bayes
2. The cross-validation in 10-fold manner for the 10 runs with the feature selection with the
classification accuracy with the desire standard deviation.
3. Number reduced features by the adopted approach

Table 4. Comparison of Attributes

Medical Data Number of attributes J48 JRIP Proposed MFCLS
Breast Cancer 9 73.28± 6.03 75.67± 4.74 88.74± 1.86
Dermatology 32 90.13± 3.33 88.74± 2.97 90.35± 5.83
Ecoli 7 81.32± 5.69 83.76± 3.78 91.56± 2.78
Heart 12 77.23± 7.85 82.73± 5.74 88.93± 6.83
Heart (Swiss) 12 34.56± 12.75 69.56± 9.06 78.45± 7.34
Heart (Cleveland) 12 75.67± 6.46 81.46± 4.62 88.93± 3.12
Hepatitis 18 76.32± 6.27 77.62± 3.02 92.58± 0.91
Liver Disorder 6 63.49± 7.39 67.83± 5.73 89.32± 1.36
Lung Cancer 55 68.35± 22.63 71.56± 8.73 93.46± 2.78
Lymphography 17 74.83± 11.35 73.67± 4.67 86.39± 1.93
Thyroid 5 86.57± 5.62 89.74± 9.76 91.36± 4.67
Diabetes 8 74.57± 4.78 81.49± 8.19 88.64± 6.83
Primary Tumor 16 43.68± 6.84 73.78± 3.67 93.67± 1.04
Sick 27 88.94± 1.34 89.56± 4.03 93.16± 2.74

The number of job assigned for the VM in the fog architecture is presented in figure 7. Based on the
assigned VM in the 1, 2 and 3 the computation cost is estimated.

Figure 7. Computation Cost for different VM

The data privacy is estimated comparatively with the proposed MFCLS model are presented in figure
8.

Available online at: https://ijcnis.org 324

 International Journal of Communication Networks and Information Security

Figure 8. Comparison of SLR

The privacy estimation of the SLR expressed that the proposed MFCLS model achieves the higher
privacy compared with the DAG model. In figure 9 and figure 10 the comparison of the energy
consumption for the total number of jobs and each job is presented.

Figure 9. Energy Consumption for total jobs

Available online at: https://ijcnis.org 325

 5G with Fog Computing based Privacy System in Data Analytics for Healthcare System by AI Techniques

Figure 10. Energy Consumption for each individual jobs

As the MFCLS model involved in the resource scheduling for the assigned VM the jobs assigned for
the each individual is computed as presented in the figure 11 and figure 12.

Figure 11. SLA violation

Available online at: https://ijcnis.org 326

 International Journal of Communication Networks and Information Security

Figure 12. Comparison of Response Time

The SLA violation comparison expressed that the proposed MFCLS model exhibits the minimal SLA
violation compared with DAG. The accuracy for the selected dataset is examined with the
consideration of the learners rate. The computation accuracy increases with the elimination of the
features those are redundant. Particularly, with the implementation of the naïve bayes classifier the
cases performance is increases significantly with the feature filtration technique. The naïve bayes
classifier exhibits the significant performance for the datasets with consideration of independent
features with the selection of feature attributes. Actually, the estimation of the features significantly
eliminates the set of filteration approach with the reduced information gain. In other words, the
demand is based on the consideration of the dependent features with the elimination of the
information in the datasets. The defined dataset need to be more reliable with effective standard
deviation value for the yielded accuracy level in the classifier compared with the standard deviation
value considered for the original ones. With the feature reductio in the proposed MFCLS model
learning rate need to be minimized with the higher extent for the induced small rule. In some datasets,
improvement in the used metrics yield by some learners (not for all the chosen learners) are observed
unchanged or very less or acceptably down, but amount (%age) of dimension reduction (i.e., noise
reduction) is considerably good. This may be due to the adopted learning strategies by the learners
that usually desire more features while training.

5. Conclusion

Fog computing technology exhibits the advancement over the CC implemented with the effective 5G
communication in the healthcare applications. In the healthcare application fog computing technology
concentrated on the reduced energy consumption with the load optimization. The proposed MFCLS
model comprises of the optimization based model for the healthcare data processing. The proposed
MFCLS model uses the entropy based feature selection model for the optimization of the ABC
algorithm with the supervision and stabilization of the load in the fog computing network with the
reduced energy consumption and task runtime length. The developed MFCLS model stabilize the load
in the fog computing with the improved performance in the framework with the recurring framework
of 1000 jobs. The formulated fitness function exhibits the significant performance towards
minimization of the task load in the VM with the initiation and execution time. The algorithm used for
the ranking comprises of the connected components for the total energy consumed for the completion
of the tasks. The proposed MFCLS algorithm uses the schedule length estimation for the developed
MFCLS architecture model to increases the data privacy in the fog. The proposed MFCLS model

Available online at: https://ijcnis.org 327

 5G with Fog Computing based Privacy System in Data Analytics for Healthcare System by AI Techniques

exhibits the smallest time frame compared with the existing algorithm. Additionally, the analysis of
the proposed MFCLS model comparatively examined with the DAG model. The performance analysis
expressed that the proposed MFCLS model achieves the SLRVM for the data achieves the value of 28
in case of the conventional DAG model it is achieved as 88. Additionally, ECRVM model achieves
the task value of 43 wheareas the DAG model provides 7. The comparative examination confirmed
that proposed MFCLS model exhibits ~6% increased performance than the conventional DAG
technique with the adequate data privacy in the heathcare data implemented in fog architecture.

References

[1] Sivasangari, A., Lakshmanan, L., Ajitha, P., Deepa, D., & Jabez, J. (2021). Big Data
Analytics for 5G-Enabled IoT Healthcare. In Blockchain for 5G-Enabled IoT (pp. 261-275).
Springer, Cham.
[2] Kamruzzaman, M. M., Yan, B., Sarker, M. N. I., Alruwaili, O., Wu, M., & Alrashdi, I.
(2022). Blockchain and Fog Computing in IoT-Driven Healthcare Services for Smart
Cities. Journal of Healthcare Engineering, 2022.
[3] Chu, K. M. Understanding Ioht and Edge/Fog Computing Solutions for Smart In-Home
Remote Healthcare. Fog Computing Solutions for Smart In-Home Remote Healthcare.
[4] Tselios, C., Politis, I., Amaxilatis, D., Akrivopoulos, O., Chatzigiannakis, I., Panagiotakis, S.,
& Markakis, E. K. (2022). Melding Fog Computing and IoT for Deploying Secure, Response-
Capable Healthcare Services in 5G and Beyond. Sensors, 22(9), 3375.
[5] Prasad, V. K., Tanwar, S., & Bhavsar, M. D. (2021). Advance cloud data analytics for 5G
enabled IoT. In Blockchain for 5G-Enabled IoT (pp. 159-180). Springer, Cham.
[6] Muneeb, M., Ko, K. M., & Park, Y. H. (2021). A fog computing architecture with multi-layer
for computing-intensive IoT applications. Applied Sciences, 11(24), 11585.
[7] Singh, J., Singh, P., & Gill, S. S. (2021). Fog computing: A taxonomy, systematic review,
current trends and research challenges. Journal of Parallel and Distributed Computing, 157,
56-85.
[8] Sicari, S., Rizzardi, A., & Coen-Porisini, A. (2022). Insights into security and privacy towards
fog computing evolution. Computers & Security, 102822.
[9] Sodhro, A. H., & Zahid, N. (2021). AI-enabled framework for fog computing driven e-
healthcare applications. Sensors, 21(23), 8039.
[10] Sarrab, M., & Alshohoumi, F. (2022). Assisted Fog Computing Approach for Data Privacy
Preservation in IoT-Based Healthcare. In Security and Privacy Preserving for IoT and 5G
Networks (pp. 191-201). Springer, Cham.
[11] Alzoubi, Y. I., Al-Ahmad, A., & Kahtan, H. (2022). Blockchain technology as a Fog
computing security and privacy solution: An overview. Computer Communications, 182, 129-
152.
[12] Ahanger, T. A., Tariq, U., Ibrahim, A., Ullah, I., Bouteraa, Y., & Gebali, F. (2022). Securing
IoT-Empowered Fog Computing Systems: Machine Learning
Perspective. Mathematics, 10(8), 1298.
[13] Alwakeel, A. M. (2021). An overview of fog computing and edge computing security and
privacy issues. Sensors, 21(24), 8226.
[14] Samann, F. E., Abdulazeez, A. M., & Askar, S. (2021). Fog Computing Based on Machine
Learning: A Review. International Journal of Interactive Mobile Technologies, 15(12).
[15] Patwary, A. A. N., Naha, R. K., Garg, S., Battula, S. K., Patwary, M. A. K., Aghasian, E., ...
& Gong, M. (2021). Towards secure fog computing: A survey on trust management, privacy,
authentication, threats and access control. Electronics, 10(10), 1171.
[16] Laghari, A. A., Jumani, A. K., & Laghari, R. A. (2021). Review and state of art of fog
computing. Archives of Computational Methods in Engineering, 28(5), 3631-3643.
[17] Abdulqadir, H. R., Zeebaree, S. R., Shukur, H. M., Sadeeq, M. M., Salim, B. W., Salih, A. A.,
& Kak, S. F. (2021). A study of moving from cloud computing to fog computing. Qubahan
Academic Journal, 1(2), 60-70.

Available online at: https://ijcnis.org 328

 International Journal of Communication Networks and Information Security

[18] Tewari, N., & Budhani, S. K. (2022). Analysis of Risk and Security Within Fog Computing-
Enabled e-Healthcare System in Uttarakhand. In Cyber Security and Digital Forensics (pp.
21-30). Springer, Singapore.
[19] Veni, T. (2022). Quantum-Based Resource Management Approaches in Fog Computing
Environments: A Comprehensive Review. Mobile Computing and Sustainable Informatics,
743-752.
[20] Ngabo, D., Wang, D., Iwendi, C., Anajemba, J. H., Ajao, L. A., & Biamba, C. (2021).
Blockchain-based security mechanism for the medical data at fog computing architecture of
internet of things. Electronics, 10(17), 2110.
[21] Hewa, T., Braeken, A., Liyanage, M., & Ylianttila, M. (2022). Fog computing and
blockchain-based security service architecture for 5G industrial IoT-enabled cloud
manufacturing. IEEE Transactions on Industrial Informatics, 18(10), 7174-7185.
[22] Hewa, T., Braeken, A., Liyanage, M., & Ylianttila, M. (2022). Fog computing and
blockchain-based security service architecture for 5G industrial IoT-enabled cloud
manufacturing. IEEE Transactions on Industrial Informatics, 18(10), 7174-7185.
[23] Karakaya, A., & Akleylek, S. (2021). A novel IoT-based health and tactical analysis model
with fog computing. PeerJ Computer Science, 7, e342.

Available online at: https://ijcnis.org 329