EasyChair Preprint

№ 9433

Cloud-Based Diabetic Prediction Framework:

Deep Learning Approach

Monther Tarawneh, Fiasal Alzyoud, Yousef Alsharrab and

Khalaf Khatatneh

EasyChair preprints are intended for rapid

dissemination of research results and are
integrated with the rest of EasyChair.

December 8, 2022
 Cloud-based Diabetic Prediction Framework: Deep
Learning Approach
Monther Tarawneh Fiasal AlZyoud Yousef AlSharrab Khalaf khatatneh
Computer Science Computer Science Computer Science Cyber security department,
department, Isra department, Isra University department Isra University Isra University
University. Amman. jordan Amman. jordan Amman. Jordan
Amman. jordan [email protected] [email protected] [email protected]
[email protected]

Abstract— Since Diabetic is one of most common growing will reduce the time and protect medical staff from
disease in the world. Which open the gate for another kind of infection. IoT is a collection of connected objects that
diseases such as blindness, kidney problems, heart disease and employ technology and algorithms on data to develop
more. Therefore, we need to develop a system the predict solutions or services for smart applications[6]. The
diabetic before it happens to people and advise them to avoid integration IoT with Artificial intelligent and cloud
it. The system is more than early detection as its prediction. computing gives and advantage to all environments[7],
We propose a cloud-based secure framework that integrates mainly the health sector. It will reduce the cost and protect
traditional machine learning methods with deep neural doctors and nurses from infectious diseases. It could be used
networks. The system collect patients readings using IoT
to give an initial prediction for their illness, and send an
devices and sensors, where it will be moved securely using
public key encryption to cloud storage. Then the prediction
alarm in case or critical situation[8]. While previous data
algorithm perform on time prediction on the data to see if the helps to diagnose patient and prescribe medicine and
patient expected to be diabetic in the future or not. The suitable treatment. All kind of reading can be collected such
prediction techniques tested on Pima Indian diabetic dataset as heart rate, EEG, blood pressure, temperature, glucose
from UCI. The result shows that it performs traditional ML level, x-rays, depression, mood and other readings. Usually
methods with accuracy of 98%. data can be collected using microcontroller such as
raspberry pie and Arduino, then stored in the cloud for
Keywords— E-Health, deep learning, Deep neural network, analysis[9]. Data processed on the cloud or fog computing
machine learning, diabetic prediction, diabetic prediction, IoT for real-time response. Security need more attention here
where data travel through the internet[10]. However, all
I. INTRODUCTION using smart health care system is moving to the cloud
The proliferation of new technologies and smart devices technology faster than before[11, 12]. For example; a
facilitate the human life, and decrease the human effort, personal health dashboard that use cloud-based system
which results in increasing the risk factors for developing developed to monitor patients, analyses their data and detect
diseases such as high blood pressure, and diabetic. The COVID-19[13, 14]. Such a system must be deployed with
number of patients in diabetic is expected to reach 625 high level of security to protect patient’s information.
million by 2045[1]. The focus of smart healthcare is to measure vital
Diabetes is a chronic common disease that affect the parameters of human body: temperature, Pulse rate,
health of the entre people in the earth. It destabilizes the Respiration rate, Oxygen level and Blood pressure. Vital
sugar level in the blood. The normal range of sugar in the signs are useful in detecting or monitoring patient’s health.
blood is from 70 – 180mg/dl. There are number of different Other useful parameters can be measured such as gesture,
diabetic and the most common are type 1 and 2. Type 1 facial expression, Consciousness level and body language.
diabetes appears in children and type 2 diabetes for the Remote monitoring systems developed to measure PPG,
middle aged and old people[2]. Diabetic open the body for EGC and temperature to check the patient status[15-17],
other disease such as kidney problems, heart disease, nerve most of them are using chair to measure these vitals such as
problem and disabilities[3]. Diabetic become a major health car seat, wheelchair or smart chairs. Wearable devices
problem. The traditional way of sugar level Self-monitoring developed over years to measures temperature and other
is finger stick samples[4]. Modern devices invented to vital sign[18-20]. Currently smart mobile devices has the
control the sugar level that record the patient state every ability to measure vital signs. Mobile health (m-health) uses
mints such as using electromagnetic radiation[5]. With the mobile devices to collect real time data and store it a
emerging of IoT and e-health framework, patient monitored server[21, 22], the collected data can be analyzed and
remotely and regularly. Health organizations concentrate processed using AI to give initial diagnose. M-health
their attention to use the new technologies and artificial reduces the costs and improve the quality of healthcare, but
techniques in diagnosing and predicting the probability of it comes with challenges in terms of interoperability and
diabetes occurrence, as it has undesired effects when it does security [4, 22]. More research done to develop a secure
not discover and treat in the first stages. Diabetes Mellitus framework for m-health the focus on authentication and
(DM) is a long-time term disease problem, which has an Encryption[23-25]. Any remote healthcare system has more
inverse impact on the human social life benefits and great transformation to healthcare, but raises a
security challenges. Security and privacy become more
Huge research carried out to apply the internet of things important in healthcare. Attackers can use any methods to
(IoT) in what smart healthcare. It helps doctors to diagnose find out sensitive data and release it to the public[26]. A
patients without direct contact. It can be done by analyzing secure framework must be used to prevent leak of
hug amount of patient data collected by IoT devices; this information and threats to patients. Data transmitted over

XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE

internet should be encrypted to insure integrity and privacy machine learning algorithms suggested predicting diabetes
of the patient. A combination of PKI, Smartcard and mellitus, these algorithms based on supervised learning,
Biometrics technologies developed to secure the clinical hybrid learning, or even ensemble learning[39].
data[27], it is a multi-layered healthcare framework depends
on public key and physical protection. Classification algorithms integrated with clustering
approach to enhance the accuracy of classification by
Doctors and health organization may depend on m- adapting the feature selection approach[40]. The proposed
health or smart house to monitor and track their patients. approach divided into three stages: firstly, feature selection
Smart house uses a network of sensors to record vital signs as it is noted that the clear feature selection led to
and monitor patient as it is in the hospitals[28, 29]. improvement in classification’s accuracy. Secondly,
Raspberry Pie device used to measure vital health signs for clustering by implementing K-means and hierarchical
home patient and send it to doctors at hospitals[26, 30], clustering of group datasets, and finally classification is
where family and patient himself can see his status and done by adapting naive Bayes classifiers. Different
prescription from doctors. The collected data can be used algorithms implemented to enhance the accuracy of diabetes
to predict the sickness level using data mining mellitus diagnosing[41]. It shows that soft voting classifier
techniques[31]. In addition, predict the future health achieved 79.08 % accuracy when it compared to the other
conditions of the patients[32]; such as diabetic existing machine-learning algorithms. ML algorithms
prediction[33] and heart disease prediction[33]. implemented in in healthcare using predictive analytics
Unpublished review on health monitoring using Smart through using six
Home[34] explored the use of IoT in healthcare and different algorithms with a dataset of patient's medical
conclude that a good collaboration and unified technology record[42], where KNN and SVM are superior to the other
must be used to have a complete reliable smart home for tested algorithms with accuracy reached to 77%.However,
healthcare monitoring. The health care transformation will the size of the dataset and number the missing values is a
make a cloud based secured framework where patients’ limitation. Artificial Neural Network (ANN) prediction
information can be securely transferred by having a network model proposed to recognize diabetes mellitus[43], dataset
between patient, hospital, doctors, and Labs. Where data is collected from Kurdistan region by gathering information
can be accessed securely from anywhere at any time. The from pregnant women with and without diabetes. The
main benefits of that is reduce the cost on patient, a unified proposed model enhances the prediction accuracy to reach
copy of patient record and deal with lack of specialist and 91% depending on the network design. Bansal, Diabetic
doctors. analyses model developed using K-means clustering with
support vector machine (SVM)[44]. It used Pima Indian
In this paper, we propose a secure cloud based diabetic diabetic dataset (PID). The model was effective but the
prediction framework. We test the prediction techniques on efficacy evaluated based on exploratory studies. Adaptive
existing diabetic dataset and compare it with traditional neuro-fuzzy inference system (ANFIS) for early diagnose
machine learning methods. of diabetic[45] used intelligent techniques. But developed to
II. RELATED WORK reduce the data’s characteristics. Diabetic analysis has been
studied using variety of machine learning methods such
The World Health Organization (WHO) states that the SVM, decision tree, naïve Bayes, and other methods, but
number of deaths in diabetes patients is increasing most of these studies used collected data from hospitals and
exponentially as it was increased by 70% during the interval the correctness depends on how clean is the data[46]. A
2000 to 2019 in all over the world, and diabetes results in comparison on six ML methods on the same data set shows
the death for more than 1.5 million individuals per year[35]. that SVM and KNN have better accuracy[47]. Another
The technology used more in diabetes treatment as it study shows that random forest is best [48]. The
includes set of hardware devices and software tools. These performance of the machine learning algorithms for the
tools enable the diabetes’ patients to manage monitor the prediction of diabetes evaluated[49]. The evaluation on
blood glucose levels, and the use of historical diabetes support vector machine, artificial neural network, logistic
technology which is divided into two main branches: insulin regression, classification tree, and K-nearest neighbour.
management by using syringe, pen, or pump, and blood Logistic regression got the highest accuracy 78%. The
glucose monitoring through using glucose monitor[36]. accuracy of artificial neural network was the best in another
Artificial Intelligence (AI) is used to enhance the medical evaluation of ML methods[50]. A comprehensive review of
diabetes treatment and to enhance the prediction of ML diabetics prediction models done lately [51] has
diagnosing diabetes’ patients[37]. reduced the conflict by previous studies on which method
There are several contribution on diabetic prediction and has the best accuracy by browsing the datasets conditions
detection techniques. Machine learning development and what affect the method performance.
increased with the new computer technology by improving
data analysis and train the computer applications to
diagnose diseases using supervised learning algorithm and
evaluate it using test data[38]. Detecting diabetic techniques
contains many processes starting from data collection, data
cleaning, feature extraction and selection, classifications
and predictions. Many ML methods applied on different
data structure.
The development of AI techniques contributes to
develop accurate models that used to recognize diabetes
mellitus patients and the probability of occurrence. Several
 Fig. 1. The architecture of deep neural network

The use of deep learning has grown faster because it Fig. 2. Proposed framework
simulate the human mind. They are used in different forms
in the health field[52, 53] and proven to be the better by The cloud layer will employ deep learning techniques to
reducing error rate and robust against data noise[54, 55]. A analysis the data predict diabetic for current patient in the
deep neural network has hidden layers between the input future. Healthcare specialist and organization to view
and output layers. Its developed to predict by exploring the patient records in a secure channel based on access
patterns in data set. The accuracy is much better when the privileges. The main focus of this paper is on the accuracy
techniques trained well. The architecture of deep neural of diabetic prediction and the security layer will be
network contains many hidden layers and several neurons discussed in future work.
in every hidden layer as shown in Figure 1. II. DATASET
I. METHODOLOGY The framework depends on real time data collected by
IoT in time. Sense we have not implement this in reality, we
The proposed framework contains four phase distributed
will use a common used data set for prediction of diabetes
over three layers as shown in Figure 2. The first phase is
is the Pima Indian Diabetes dataset is retrieved from the UCI
Data collection where data can be collected using smart
machine learning repository database[56]. The data set
devices and with the use of IoT technology. The second
contains of 768 rows and 9 columns. The attributes are
phase is data pre-processing where we clean up the data and
glucose, pregnancies, skin thickness, blood pressure, BMI,
remove unnecessary features. The third phase is the
insulin, and age. The last column is test result: positive or
prediction phase where the implemented hybrid algorithm
negative for diabetic. Table 1 below the features and there
is used to analyses the collected data and draw a diabetic
properties.
prediction using medical facts and historical records. The
phase is the evaluation criteria. These phases distributed Table 1. Dataset Features
over three layers: sensor layer where IoT devices and
sensors used to collect data, security layer where data will Feature Description Type
be encrypted using public key encryption, the last layer is
the cloud layer where the computation will take place. Preg Number of times pregnant Numeric

The collection layer will employ IoT devices to collect Plasma glucose concentration at 2 Hours
patient’s data with. Sensor devices can be wearable, Gluc
in an oral glucose tolerance test (GTIT)
Numeric
embedded in the body, on Apps on smart devices. The
communication technology on the sensor device will enable BP Diastolic Blood Pressure (mm Hg) Numeric
it to transfer the data to a health gateway, which in turn
forward these accumulated data to cloud storage. The data Skin Triceps skin fold thickness (mm) Numeric
will be encrypted using patient private key before
transmitting it to cloud. There are two operations for the Insulin 2-Hour Serum insulin (µh/ml) Numeric
security process:
Body mass index [weight in kg/(Height in
Key management: RSA algorithm will be used to BMI Numeric
m)]
generate pair of keys for each patients and specialist. The
public key will be stored in a directory on the cloud and used DPF Diabetes pedigree function Numeric
to encrypt and decrypt data to/from patient
gateway/specialist. In addition, give access to authorized Age Age (years) Numeric
users to access patient records.
Binary value indicating non-diabetic
Data encryption/decryption: patient data will be Outcome Factor
/diabetic
encrypted using his own private key, and decrypted using
his public keys. Request to access data encrypted using III. PROPOSED FRAMEWORK
requester private keys and replies encrypted using requester
public key. The framework is design for real time analysis and data
collected on time remotely over years to predict diabetic.
Due some network disconnections or device faults (down),
some readings could not reach the cloud storage. Therefore,
pre-processing is an essential step to avoid in accurate Machine learning classifiers used on the dataset. Data
prediction. The algorithm will consider the reading as a split into two sets: training (90%) and test (%10). KNN,
DNA for the person over time and any missing values will SVM and Random forest are supervised ML where .KNN is
be replaced based on previous and current readings. used to classify dataset’s columns, SVM is suitable for small
datasets and RF classifier is a several decision tree. The ML
In this research, Pima data set to check the accuracy methods are used more like a feature selection techniques.
prediction. Therefore, we will focus on cleaning data and
feature selection at this stage. We use python to check the The prediction system implemented using python
data set for missing or none values. The test shows that the library. Deep neural network with 4 hidden layers is
data has not missing or none values but some features values implemented and the number of neurons are 12, 16, 16 and
are inconsistent: glucose, blood pressure, skin fold 14 respectively for layer 1, 2, 3 and 4. The best performance
thickness, insulin and BMI. This may affect the accuracy of for diabetic prediction of the algorithm on four layers
the algorithm. If we look at the statistic of the data set as architecture. The input layer layers are eight and the output
shown on table 2. We can see many zeros as all null values is one. The output is one layer because the output can be
replaced by Zero. The real time dataset will not contain inferred as positive or negative. To evaluate the supervised
many zeros’, because a genetic algorithm will replace
machine learning algorithms, the confusion matrix is used.
missing values with a values base on previous and future
Where it makes the evaluation of the accuracy is easy.
value for this feature. However, the data set we used can be
replace because we have no history and not on time data. Table 3: Correlation between features
Also, dropping one of eight features is not a good idea. It
may reduce the noise but reduce the accuracy as well.

Pregnancies

Glucose
Pressure
Blood
Thickness
Skin

Insulin

BMI
Function
Pedigree

Age

Outcome
Table 2: Data set features statistics.
Pregnancies

Thickness

Function
Pressure

Pedigree
Diabetes
Glucose

Insulin
Blood

result
BMI
Skin

Age

ies
Pregnanc

1.000000

0.129459

0.141282

-0.081672

-0.073535

0.017683

-0.033523

0.544341

0.221898
120.89453

69.10547

20.53646

79.79948

31.99258
3.84505

0.47188

Glucose

0.129459

1.000000

0.152590

0.057328

0.331357

0.221071

0.137337

0.263514

0.466581
mean

33.2

0.35
115.24400
31.97262

19.35581

15.95222

Pressure
Blood

0.141282

0.152590

1.000000

0.207371

0.088933

0.281805

0.041265

0.239528

0.065068
3.36958

7.88416

0.33133

11.8

0.48
std

Thickness
Skin

-0.081672

0.057328

0.207371

1.000000

0.436783

0.392573

0.183928

-0.113970

0.074752
0.078

21.0
min

0.0

0.0

0.0

0.0

0.0

0.0

0.0
0.2437
99.00

62.00

27.30
25%

1.00

0.00

0.00

24.0

0.00
5

Insulin
0.073535
-

0.331357

0.088933

0.436783

1.000000

0.197859

0.185071

-0.042163

0.130548
117.00

0.3725
72.00

23.00

30.50

32.00
50%

3.00

29.0

0.00

BMI

0.017683

0.221071

0.281805

0.392573

0.197859

1.000000

0.140647

0.036242

0.292695
0.62625
140.25

127.25
80.00

32.00

36.60
75%

6.00

41.0

1.00

Function
Pedigree

-0.033523

0.137337

0.041265

0.183928

0.185071

0.140647

1.000000

0.033561

0.173844
2.42000
199.00

122.00

846.00
17.00

99.00

67.10
max

81.0

1.00

Age

0.544341

0.263514

0.239528

-0.113970

-0.042163

0.036242

0.033561

1.0

0.238356

Table 3 shows the correlation between all variables. As

we can see that some features have significant correlation
with the outcome which are glucose levels, age, BMI and
Outcome

0.221898

0.466581

0.065068

0.074752

0.130548

0.292695

0.173844

0.238356

1.000000

number of pregnancies. However, higher blood pressure is

correlated with a person not being diabetic and Logistic
Regression model confirm that blood pressure have
negative impact on the outcome variable. however,
combining blood pressure with other features will give
positive impact on the accuracy.
 IV. RESULT AND ANALYSIS time from continuous glucose monitoring sensor time-series," IEEE
Transactions on biomedical engineering, vol. 54, pp. 931-937, 2007.
The paper propose a diabetic prediction framework [4] L. Olansky and L. Kennedy, "Finger-stick glucose monitoring:
using Deep Neural network (DNN) integrated with issues of accuracy and specificity," vol. 33, ed: Am Diabetes Assoc,
traditional machine learning methods for feature selection. 2010, pp. 948-949.
We have used a python open source library for machine [5] C. Choleau, J.-C. Klein, G. Reach, B. Aussedat, V. Demaria-Pesce,
learning called Scikit. The best performance of deep neural G. Wilson, et al., "Calibration of a subcutaneous amperometric
network in five-fold cross validation with accuracy 98.04, glucose sensor: Part 1. Effect of measurement uncertainties on the
determination of sensor sensitivity and background current,"
sensitivity 98.7, and F1 score equal to 0.99. Table 4 show Biosensors and Bioelectronics, vol. 17, pp. 641-646, 2002.
the comparison between existed methods and the deep [6] A. Elhadad, F. Alanazi, A. I. Taloba, and A. Abozeid, "Fog
neural network approach. Computing Service in the Healthcare Monitoring System for
Managing the Real-Time Notification," Journal of Healthcare
Table 4. Comparison study of ML methods and DNN
Engineering, vol. 2022, 2022.
Method Accuracy [7] V. Bhardwaj, R. Joshi, and A. M. Gaur, "IoT-based smart health
SVM 82.8 monitoring system for COVID-19," SN Computer Science, vol. 3,
GA 80 pp. 1-11, 2022.
DNN 98.4 [8] M. Islam and A. Rahaman, "Development of smart healthcare
KNN 76 monitoring system in IoT environment," SN computer science, vol.
RF 83 1, pp. 1-11, 2020.
Decision tree 78 [9] H. Pandey and S. Prabha, "Smart health monitoring system using
LR 78 IOT and machine learning techniques," in 2020 sixth international
Naïve Bayes 80 conference on bio signals, images, and instrumentation (ICBSII),
2020, pp. 1-4.
[10] Y. Alghofaili, A. Albattah, N. Alrajeh, M. A. Rassam, and B. A. S.
The prediction techniques applied on existing data set to Al-rimy, "Secure Cloud Infrastructure: A Survey on Issues, Current
compare its accuracy with existing methods. The best Solutions, and Open Challenges," Applied Sciences, vol. 11, p. 9005,
accuracy happen to be with the proposed DNN. The 2021.
accuracy reached this limit because we get benefit from the [11] I. Dirgová Luptáková, M. Kubovčík, and J. Pospíchal, "Wearable
traditional machine learning methods for classifications and sensor-based human activity recognition with transformer model,"
Sensors, vol. 22, p. 1911, 2022.
feature selections. The correlation between features help us
[12] I. D. Luptáková, M. Kubovčík, and J. Pospíchal, "Wearable Sensor
to decide which features are related to each other’s and Based Human Activity Recognition with Transformer," 2022.
important for increasing the accuracy. However, [13] A. Bahmani, A. Alavi, T. Buergel, S. Upadhyayula, Q. Wang, S. K.
accumulative reading may help us with the use of integrated Ananthakrishnan, et al., "A scalable, secure, and interoperable
DNN to predict the diabetic before it happen based on the platform for deep data-driven health management," Nature
patient readings over years. This accuracy will help us to communications, vol. 12, pp. 1-11, 2021.
distinguish real diabetic expectation from on time data [14] A. A.-R. El-Douh, S. F. Lu, A. A. Elkouny, and A. Amein, "Hybrid
analysis. cryptography with a one-time stamp to secure contact tracing for
COVID-19 infection," International Journal of Applied
Mathematics and Computer Science, vol. 32, 2022.
V. CONCLUSION
[15] A. N. Soni, "Smart devices using internet of things for health
Diabetic is a chronic disease that force people to live in monitoring," International Journal of Innovative Research in
different way from normal people. Otherwise, it may cause Science, Engineering and Technology, vol. 7, pp. 6355-6361, 2018.
death or bring other diseases. Discovering diabetic in early [16] R. Lavanya, M. Nivetha, K. Revasree, and K. Sandhiya, "Smart
stage is good for medical treatment. However, predicting it chair-a telemedicine based health monitoring system," in 2018
years before it happen is a way to avoid getting it and live second international conference on electronics, communication and
aerospace technology (ICECA), 2018, pp. 459-463.
normal life. This paper, implement diabetic prediction
[17] N. Surantha, O. K. Utomo, M. I. Sani, and B. Soewito, "Iot system
framework that has three layers where patients data for sleep quality monitoring using ballistocardiography sensor,"
collected and sent protected to a cloud storage, where it will International Journal of Advanced Computer Science and
be explored by integrated prediction system that combine Applications, vol. 11, 2020.
traditional machine learning methods with deep neural [18] T. Mizuno and Y. Kume, "Development of a Glasses-Like Wearable
network to predict diabetic. The proposed prediction Device to Measure Nasal Skin Temperature," Cham, 2015, pp. 727-
method tested on Pima Indian diabetic dataset and shows 732.
best performance with 98% accuracy. The proposed [19] T. Tamura, M. Huang, and T. Togawa, "Current developments in
wearable thermometers," Advanced Biomedical Engineering, vol. 7,
framework will benefit individual who has historical pp. 88-99, 2018.
diabetic in their family to track themselves and be aware of [20] H. Nogami, H. Okada, T. Miyamoto, R. Maeda, and T. Itoh,
their diabetic early. Also, provide hospitals with early alarm "Wearable and compact wireless sensor nodes for measuring the
to advice patients with early procedure to avoid temperature of the base of a Calf’s Tail," Sens. Mater, vol. 25, pp.
complications. 577-582, 2013.
[21] S. H. Almotiri, M. A. Khan, and M. A. Alghamdi, "Mobile health
REFERENCES (m-health) system in the context of IoT," in 2016 IEEE 4th
international conference on future internet of things and cloud
[1] G. Li, S. Peng, C. Wang, J. Niu, and Y. Yuan, "An energy-efficient
workshops (FiCloudW), 2016, pp. 39-42.
data collection scheme using denoising autoencoder in wireless
sensor networks," Tsinghua Science and Technology, vol. 24, pp. [22] K. Shirato, N. Nao, H. Katano, I. Takayama, S. Saito, F. Kato, et al.,
86-96, 2018. "Development of genetic diagnostic methods for novel coronavirus
2019 (nCoV-2019) in Japan," Japanese journal of infectious
[2] N. Chaudhary and N. Tyagi, "Diabetes mellitus: An Overview,"
diseases, p. JJID. 2020.061, 2020.
International Journal of Research and Development in Pharmacy &
Life Sciences, vol. 7, pp. 3030-3033, 2018. [23] S. A. Butt, T. Jamal, M. A. Azad, A. Ali, and N. S. Safa, "A
multivariant secure framework for smart mobile health application,"
[3] G. Sparacino, F. Zanderigo, S. Corazza, A. Maran, A. Facchinetti,
and C. Cobelli, "Glucose concentration can be predicted ahead in
 Transactions on Emerging Telecommunications Technologies, p. [42] M. A. Sarwar, N. Kamal, W. Hamid, and M. A. Shah, "Prediction of
e3684, 2019. diabetes using machine learning algorithms in healthcare," in 2018
[24] S. Nazir, Y. Ali, N. Ullah, and I. García-Magariño, "Internet of 24th international conference on automation and computing
things for healthcare using effects of mobile computing: a systematic (ICAC), 2018, pp. 1-6.
literature review," Wireless Communications and Mobile [43] R. Jader and S. Aminifar, "Fast and Accurate Artificial Neural
Computing, vol. 2019, 2019. Network Model for Diabetes Recogni-tion," NeuroQuantology, vol.
[25] I. Chiuchisan, H.-N. Costin, and O. Geman, "Adopting the internet 20, pp. 2187-2196, 2022.
of things technologies in health care systems," in 2014 International [44] M. Ahmed, M. Elghandour, A. Salem, H. Zeweil, A. Kholif, A.
Conference and Exposition on Electrical and Power Engineering Klieve, et al., "Influence of Trichoderma reesei or Saccharomyces
(EPE), 2014, pp. 532-535. cerevisiae on performance, ruminal fermentation, carcass
[26] S. L. Rohit and B. V. Tank, "Iot based health monitoring system characteristics and blood biochemistry of lambs fed Atriplex
using raspberry PI-review," in 2018 Second International nummularia and Acacia saligna mixture," Livestock Science, vol.
Conference on Inventive Communication and Computational 180, pp. 90-97, 2015.
Technologies (ICICCT), 2018, pp. 997-1002. [45] K. Polat and S. Güneş, "An expert system approach based on
[27] A. Dwivedi, R. K. Bali, M. A. Belsis, R. N. G. Naguib, P. Every, principal component analysis and adaptive neuro-fuzzy inference
and N. S. Nassar, "Towards a practical healthcare information system to diagnosis of diabetes disease," Digital signal processing,
security model for healthcare institutions," in 4th International IEEE vol. 17, pp. 702-710, 2007.
EMBS Special Topic Conference on Information Technology [46] M. Anouncia, C. Maddona, P. Jeevitha, and R. Nandhini, "Design of
Applications in Biomedicine, 2003., 2003, pp. 114-117. a diabetic diagnosis system using rough sets," Cybernetics and
[28] T. S. Barger, D. E. Brown, and M. Alwan, "Health-status monitoring Information Technologies, vol. 13, pp. 124-169, 2013.
through analysis of behavioral patterns," IEEE Transactions on [47] Q. Zou, K. Qu, Y. Luo, D. Yin, Y. Ju, and H. Tang, "Predicting
systems, man, and cybernetics-part A: systems and humans, vol. 35, diabetes mellitus with machine learning techniques," Frontiers in
pp. 22-27, 2004. genetics, vol. 9, p. 515, 2018.
[29] R. Basatneh, B. Najafi, and D. G. Armstrong, "Health sensors, smart [48] M. K. Hasan, M. A. Alam, D. Das, E. Hossain, and M. Hasan,
home devices, and the internet of medical things: an opportunity for "Diabetes prediction using ensembling of different machine learning
dramatic improvement in care for the lower extremity complications classifiers," IEEE Access, vol. 8, pp. 76516-76531, 2020.
of diabetes," Journal of diabetes science and technology, vol. 12, pp. [49] A. K. Dwivedi, "Analysis of computational intelligence techniques
577-586, 2018. for diabetes mellitus prediction," Neural Computing and
[30] M. H. Kashani, M. Madanipour, M. Nikravan, P. Asghari, and E. Applications, vol. 30, pp. 3837-3845, 2018.
Mahdipour, "A systematic review of IoT in healthcare: Applications, [50] M. Heydari, M. Teimouri, Z. Heshmati, and S. M. Alavinia,
techniques, and trends," Journal of Network and Computer "Comparison of various classification algorithms in the diagnosis of
Applications, vol. 192, p. 103164, 2021. type 2 diabetes in Iran," International Journal of Diabetes in
[31] R. Nagavelli and C. G. Rao, "Degree of Disease possibility (DDP): Developing Countries, vol. 36, pp. 167-173, 2016.
A mining based statistical measuring approach for disease prediction [51] R. Saxena, S. K. Sharma, M. Gupta, and G. Sampada, "A
in health care data mining," in International Conference on Recent Comprehensive Review of Various Diabetic Prediction Models: A
Advances and Innovations in Engineering (ICRAIE-2014), 2014, pp. Literature Survey," Journal of Healthcare Engineering, vol. 2022,
1-6. 2022.
[32] P. K. Sahoo, S. K. Mohapatra, and S.-L. Wu, "Analyzing healthcare [52] A. Ashiquzzaman, A. K. Tushar, M. Islam, D. Shon, K. Im, J.-H.
big data with prediction for future health condition," IEEE Access, Park, et al., "Reduction of overfitting in diabetes prediction using
vol. 4, pp. 9786-9799, 2016. deep learning neural network," in IT convergence and security 2017,
[33] M. Tarawneh and O. Embarak, "Hybrid approach for heart disease ed: Springer, 2018, pp. 35-43.
prediction using data mining techniques," in International [53] J. Vijayashree and J. Jayashree, "An Expert System for the
Conference on Emerging Internetworking, Data & Web Diagnosis of Diabetic Patients using Deep Neural Networks and
Technologies, 2019, pp. 447-454. Recursive Feature Elimination," International Journal of Civil
[34] K. S. Sahu, A. Oetomo, and P. P. Morita, "Health Monitoring Using Engineering and Technology, vol. 8, pp. 633-641, 2017.
Smart Home and Active Assisted Living (AAL) Technologies: A [54] A. Thammano and A. Meengen, "A new evolutionary neural
Scoping." network classifier," in Pacific-Asia Conference on Knowledge
[35] M. Ati and R. Al-Bostami, "Multi Artificial Intelligence Approaches Discovery and Data Mining, 2005, pp. 249-255.
Comparisons for Chronic Disease Prediction," in 2021 IEEE 5th [55] S. I. Ayon and M. M. Islam, "Diabetes prediction: a deep learning
International Conference on Information Technology, Information approach," International Journal of Information Engineering and
Systems and Electrical Engineering (ICITISEE), 2021, pp. 315-319. Electronic Business, vol. 12, p. 21, 2019.
[36] M. Marigliano, A. J. Eckert, P. K. Guness, A. Herbst, C. E. Smart, [56] "Pima Indian Diabetes Data Set," ed. UCI machine learning
M. Witsch, et al., "Association of the use of diabetes technology repository database
with HbA1c and BMI‐SDS in an international cohort of children and
adolescents with type 1 diabetes: The SWEET project experience,"
Pediatric Diabetes, vol. 22, pp. 1120-1128, 2021.
[37] U. Ahmed, G. F. Issa, M. A. Khan, S. Aftab, M. F. Khan, R. A. Said,
et al., "Prediction of diabetes empowered with fused machine
learning," IEEE Access, vol. 10, pp. 8529-8538, 2022.
[38] M. M. Bukhari, B. F. Alkhamees, S. Hussain, A. Gumaei, A. Assiri,
and S. S. Ullah, "An improved artificial neural network model for
effective diabetes prediction," Complexity, vol. 2021, 2021.
[39] G. Pradhan, R. Pradhan, and B. Khandelwal, "A study on various
machine learning algorithms used for prediction of diabetes
mellitus," in Soft Computing Techniques and Applications, ed:
Springer, 2021, pp. 553-561.
[40] Y. K. Alapati and K. Sindhu, "Combining clustering with
classification: a technique to improve classification accuracy," Lung
Cancer, vol. 32, p. 3, 2016.
[41] S. Kumari, D. Kumar, and M. Mittal, "An ensemble approach for
classification and prediction of diabetes mellitus using soft voting
classifier," International Journal of Cognitive Computing in
Engineering, vol. 2, pp. 40-46, 2021.