Proceedings of the Sixth International Conference on Trends in Electronics and Informatics (ICOEI 2022)

IEEE Xplore Part Number: CFP22J32-ART; ISBN: 978-1-6654-8328-5

Cardiovascular Disease Prediction using Deep

Learning
Paranthaman M Yaathash B Santhosh S Sanjairam M
Assistant Professor UG Student UG Student UG Student
Department of Electronics and Department of Electronics and Department of Electronics and Department of Electronics and
Communication Engineering Communication Engineering Communication Engineering Communication Engineering
2022 6th International Conference on Trends in Electronics and Informatics (ICOEI) | 978-1-6654-8328-5/22/$31.00 ©2022 IEEE | DOI: 10.1109/ICOEI53556.2022.9777135

M.Kumarasamy College of M.Kumarasamy College of M.Kumarasamy College of M.Kumarasamy College of

Engineering Engineering Engineering Engineering
Karur, India Karur, India Karur, India Karur, India
[email protected] [email protected] [email protected] [email protected]

Abstract—One of the most important structures in the human cardiovascular disease prediction system. As a result, new
body is the heart. It is at the focal point of the circulatory medical data min ing approaches and machine learning
framework. Heart disease is a potentially fatal condition that can techniques are being researched and developed. The major goal
result in death or severe long-term impairment. Therefore, of this research is to use classification algorith ms to find
efficient techniques for uncovering hidden linkages and there are essential patterns and features in medical data, and next select
no clear patterns in e-health data. Medical diagnosis is a difficult the most important characteristics for detecting a silent heart
task that is crucial to preserving lives, so it must be done attack. The results will be even more accurate thanks to the
correctly and quickly. To lower the cost of performing clinical
implementation of an Artificial Neural Network. While the
tests, a suitable and precise computer-based automated decision
support system is necessary. The use of machine learning in
development of such a system is not unique, current systems
health analytics has been presented as a way to anticipate reliable
have flaws and are not designed to detect the possibility of
patient data analysis. The information generated by the silent heart attacks. The goal of this study is to solve these
healthcare industry is not mined. In the medical industry, data issues and propose the installation of unique features to create a
mining techniques can be utilized to create an intelligent model more comprehensive system. The current technology for
employing data sets that include patient risk factors. The forecasting heart attacks does not produce results that are
emergence of ideas and techniques for making use of data has accurate enough. The machine learning approaches deployed,
surprised knowledge discovery in databases (KDD). This study as observed in the literature survey, are pushing the accuracy to
delves into the usage of deep learning and machine learning a certain limit. Furthermore, the utilizat ion of attributes is a
approaches in disease diagnosis. In recent years, many data problem with the current heart attack prediction algorithm. The
mining classifiers have been created to aid in the accurate and characteristics that should be chosen for heart attack prediction
timely identification of illnesses. This study offers a heart attack are conventional, and as a result, the findings frequently
forecasting model based on deep learning methodologies, notably produce incorrect output. The suggested approach seeks to
Multi-Layer Perceptron (MLP), to anticipate a patient's extract the appropriate properties from datasets to improve
likelihood of acquiring heart disease. MLP is an advanced type prediction precision. It will also provide a correct diagnosis to
approach that employs the Artificial Neural Network's Deep consumers so that they can grasp the problem without trouble.
Learning approach. Deep learning and data mining are used in The project distinguishes itself by merging deep learning and
the suggested approach to get reliable and error-free outcomes. data mining capabilities. The study provides a system that
Keywords— Artificial intelligence, Data Science and
includes a powerful classification module and a comp lete
Engineering, Deep learning, Heart disease, Multi-layer perceptron
report generating module, as well as a strong prediction
algorithm. This research proposes to build a self-learning
I. INT RODUCT ION procedure in which past inputs of disease results influence
future heart disease possibilit ies for a specific user. The
Cardiovascular illness is one of the most common diseases
proposed model makes extensive use of preprocessing methods
in today's society. According to a report, more than 17.7 to ensure that the classification and prediction of the dataset are
million people die each year due to heart disease over the
error-free. To make the prediction more accurate, a large
world. 7.4 million people died as a result of coronary heart number of training sets will be used.
disease, whereas 6.7 million people died as a result of a stroke.
Heart attacks are one of the most lethal diseases that can strike II. RELAT ED W ORK
at any time and without warning, and silent heart attacks are
something that most doctors are unable to predict. The paucity Manmohan Singh and his colleagues , [1] Unsupervised
of specialists and the rising number of incorrectly diagnosed input clustering is a data analysis approach. Cluster analysis is
patients have forced the development of a reliable a technique for breaking down a big set of data into smaller

978-1-6654-8328-5/22/$31.00 ©2022 IEEE 1399

Authorized licensed use limited to: National Institute of Technology. Downloaded on July 27,2023 at 16:37:20 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Sixth International Conference on Trends in Electronics and Informatics (ICOEI 2022)
IEEE Xplore Part Number: CFP22J32-ART; ISBN: 978-1-6654-8328-5

groups known as clusters. Each cluster consists of a group of patterns. This is a knowledge discovery analyzer for databases
data objects that are comparable in some way. The K-Means that will be utilized in the decision-making process. It is
algorithm is a non-hierarchical clustering technique that typically used by large corporations to develop in novative
divides data into one or more clusters. The data is clustered methods to improve revenues and save costs. Data mining
using this method, with data of similar qualit ies grouped in one examines data and aids in the discovery of hidden aspects,
cluster and data of different qualities grouped in another. To allowing for the generation of meaningful patterns and
cluster the data, the Expectation-Maximization Algorith m information.
(EM-clustering) is applied.
Wisaeng and Kittipol, [8] Data mining is a technology that
Yannis Papaniko laou and others, [2] The proposed combines machine learn ing, statistics, and database systems to
Structure of Multi-label Classifiers (HOM ER) is a mult i-label uncover patterns in large data sets. The analytical stage of the
learning approach that splits early developmental activity into KDD (knowledge discovery in databases) process is data
mu ltiple categories, simp ler sub-tasks by first constructing a mining. Using data mining techniques, it is possible to extract
label structure from a provided information set and then knowledge and discover interesting and useful patterns.
applying the resulting sub-problems to a given base mult i-label
A. T. Saeb, et.al,… [9] We go through data mining
classifier (M LC). The main purpose is to successfully handle
the problems of scalability and class imbalance that typically techniques in-depth, with a particular focus on classification
techniques as an essential supervised learning tool. WEKA
occur in multi-label categorization jobs in the real world. A
software is also discussed as a viable option for performing
balanced version of the k means the technique used in the early
phase of labeled structure development has been added as a categorization analysis on various types of data. The
application has a well-defined methodology to make it easy to
bonus feature. Multi-label learn ing has gained a lot of scientific
attention in the last decade. use for a wide range of users. WEKA has 49 data preparation
tools, 76 classification/regression algorithms, 8 clustering
T. Sajana, et.al,… [3] Big Data principles were methods, 3 association rule algorithms, 15 attribute/subset
incorporated, which refers to the Data Mining environment that evaluators, and 10 feature selection search algorithms. WEKA
processes a large amount of data. To put it another way, It's a pulls important information fro m data and allows for the
set of massive, complicated databases that typical data identification of an appropriate method for producing an
processing tools can't handle. Big Data is the process of accurate prediction model from it.
transforming unstructured, valuable, imprecise, and complex
data into actionable information. Pon Periasamy, [10] Data min ing and Big Data analytics
are assisting in the diagnosis, treatment, assistance, and healing
A. Jenneth and others, [4] The Hubness requirement is used of all patients in need of medical care, with the ultimate
in a unique approach known as affine subspace clustering to objective of this area being enhanced health care output.
solve the local image relevant values and the Curse of
Vijayaku mar, T., [11] Emp loying a revolutionary deep
Dimensionality. The Hubness property is a clustering strategy
with effects related to cluster structures that min imize the convolutional neural network to solve inverse problems
(CNN). Unrolled iterative approaches appear to take on the
discrimination difficu lty in cluster formation. We harness large
dimensionality by using the downstream closure condition and CNN form when the forward model's normal operator is
understood to be a convolution. In light of this finding, we
outliers identification in the k closest neighbor list, rather than
trying to avoid the plague of dimensionality by observing a developed an approach that employs CNN after direct
inversion to solve the convolutional inverse issue.
lesser dimensional component subspace. The Feature
weighting technique is also used to reduce the average within - Pandian, A. Pasumpon, [12] Traditional approaches, such
cluster scatterings while increasing the average among cluster as the surface approach, would rely on a time -consuming
scatterings through all element spaces. manual feature extraction procedure, which is a key co mponent
Lamine M. Aouad, et.al,.. [5] suggested a dis tributed of feature-driven developments. These strategies serve as a
solid foundation for determining the predictability of features,
method that merges sub-clusters to produce global clusters
under a variance restriction. This improves overall clustering as well as the ideal platform for incorporating deep learning
techniques. The suggested study introduces a deep learning
quality and enables automated cluster count identification.
However, a suitable maximu m rising value must be chosen. approach that may be used in conjunction with feature
extraction.
This can be inferred fro m the issue domain or discovered
through a variety of approaches. Hamdan and others [13] To solve nonlinearly divisible
Lashari and his colleagues, [6] A comparative examination challenges, the recognition techniques operate using numerous
ways that artificial neural networks and statistical
of data mining techniques applications are offered. As a result,
current research suggests that we should not overlook data methodologies that we have selected, among others. This
article discusses numerous methods for comparing and
mining approaches' current and future potential in impacting
the thematic map's successful classification of med ical data. As recognizing handwritten characters in picture documents.
Statistical, template matching, structural pattern recognition,
a result, there is a lot of promise for using data mining
and graphical approaches are also compared to the statistical
techniques to classify medical data.
support vector machine (SVM) classifiers network technology.
Sharma and others, [7] Data mining are the practice of It has been demonstrated that Statistical SVM for OCR system
applying algorithms to vast amounts of data to uncover

978-1-6654-8328-5/22/$31.00 ©2022 IEEE 1400

Authorized licensed use limited to: National Institute of Technology. Downloaded on July 27,2023 at 16:37:20 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Sixth International Conference on Trends in Electronics and Informatics (ICOEI 2022)
IEEE Xplore Part Number: CFP22J32-ART; ISBN: 978-1-6654-8328-5

performance provides a decent outcome when combined with a classifier that can reliably and efficiently classify newly created
machine learning technique. data is required. For data categorization, this system primarily
uses the supervised learning approach known as Random
Tripathi and others, [14] With the increase of COVID19
forests, which is achieved by changing the values of numerous
cases and the government's months -long lockdown. Citizens hyperparameters in the Random Forests Classifier.
fro m various corners of the globe were adversely affected.
Educational institutions, businesses, and government offices IV. PROPOSED M ET HODOLOGIES
must rely on contemporary technology to carry out everyday
tasks to keep the country's economy viable. The full potential Cardiovascular disease continues to take the lives of an
of technology has been realized. The usage of social media for alarming number of people all over the world. The most
communication had been on the increase in the past, and this serious affliction afflicting developed countries is
shutdown gave it a tremendous boost. Twitter is one of the cardiovascular disease (CVD). CVD not only kills a high
fastest-growing social med ia platforms for people to exp ress percentage of the population without warning, but it also leaves
themselves on many issues. As a result, this platform has a great number of people suffering and disabled for a long
amassed a vast amount of data. As a consequence, time. Although a major proportion of CVDs are avoidable,
understanding this data will aid businesses in making better their prevalence continues to rise, owing to insufficient
decisions. preventive interventions. In the realm of medicine, diagnosing
heart disease has become a d ifficult undertaking. Th is
Swain, et.al,… [15] Coronary artery disease affects the diagnosis is based on a comprehensive and accurate review of
arteries of the heart and is a kind of heart disease. As a result, the patient's clinical testing and medical history. Machine
effective deep learning technology is employed to improve learning research is advancing at a rapid pace, to develop
illness detection accuracy. A Dense Neural Network, a sort of intelligent automated systems to assist doctors in illness
deep learning network, is used to construct the suggested prediction and decision-making. Such an automated medical
system. The experiment is carried out utilizing the Cleveland diagnosis system would improve timely medical care and
Heart Disease data set from the UCI repository. During proper follow-up treatment, leading to considerable lifesaving.
training, the classification accuracy was 96.03 percent, while
during testing, the accuracy was 94.91 percent. These intelligent systems achieve accurate diagnosis by
incorporating classification algorithms. Neural networks have
Miranda E, et.al,… [16] The method was developed by first become a popular categorization approach. The training
determin ing the number of cardiovascular disease risk algorithm used in this study was a mu lti-layer perceptron
variables for people based on their medical records, and then neural network with back-propagation. This research suggests
creating a mining methodology model using a naive Bayes a more accurate diagnostic technique for predicting cardiac
classifier. The classifier performance of this reco mmended disease. The propagation process was repeated until the lowest
model (accuracy, sensitivity, and specificity) was examined, possible error rate was achieved. And the accuracy rate is
and it correctly predicted the class label of tuples (above 80 maximized, as seen by the results reported in the preceding
percent). section. A neural network is a computer approach that
represents how the brain solves issues by connecting vast
III. EXIST ING M ET HODOLOGIES clusters of biological neurons through axons. Each brain unit is
In today's world, disease diagnosis and treatment is one of linked to a large nu mber of others. The effect of links on the
the most prevalent uses of machine learn ing algorithms. activation state of connected neuronal units can be either
Machine learning algorithms are also utilized to discover enforcing or inhibiting. Each neural unit might have a
illness correlations and relationships. Many people are dying summing function that adds the values of its inputs together.
these days as a result of a sudden heart attack. Heart disease Each link and the unit itself have a threshold function or
prediction and diagnosis have become a difficult task for limit ing function that must be exceeded before transmission to
doctors and hospitals in India and worldwide. To minimize the neighboring neurons. These systems excel in situations where
number of fatalities caused by cardiac illnesses, we need to be the solution or feature recognition is difficu lt to explain in a
able to predict whether or not a person is at risk of developing standard computer program since rather than being directly
one. In this field, data mining techniques and machine learning programmed, they are taught and self-learn. When a neuron
algorithms are extremely significant. Many researchers are connects to other neurons via a connection link and the neurons
working in this field to build software that can assist doctors in are arranged in a single-layered or mu lti-layered network. A
making decisions about heart disease prediction and diagnosis. mu ltilayer ANN is made up of one or more h idden layers, an
We investigated how data mining techniques may be used to input layer, and an output layer. Before translating the input to
predict cardiac disease in advance, allowing patients to receive the output layer, the hidden layer is useful for performing
the best therapy possible. Any diagnostic system must perform intermed iate calculations. When creating a model for a given
the work of determining and/or identifying a suspected disease application, inputs and targets are used to train the model until
or ailment, as well as the choice made as a result of this it learns to correlate a certain input with a specific output. The
procedure. This is when machine learning techniques come in weight change during a training cycle is trained until it reaches
handy. To be useful in medical d iagnostic problems, these a minimal value. A model is validated after it has been trained
machine learning techniques have excellent performance, the by determin ing whether it generates accurate results. Because
flexibility to deal with missing and noisy data, diagnostic of the huge number of synaptic weights available in the
knowledge transparency, as well as the ability to justify network, mult ilayered networks can memorize data. The output
judgment. People are generating more data every day, thus a of a neuron in the output layer is occasionally qualified using a

978-1-6654-8328-5/22/$31.00 ©2022 IEEE 1401

Authorized licensed use limited to: National Institute of Technology. Downloaded on July 27,2023 at 16:37:20 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Sixth International Conference on Trends in Electronics and Informatics (ICOEI 2022)
IEEE Xplore Part Number: CFP22J32-ART; ISBN: 978-1-6654-8328-5

threshold function. Although we are discussing artificial

neurons, we will call them neurons. Synapses between neurons
are represented by edges of a directed network with nodes
representing artificial neurons, and connections are synapses
between neurons.

The process of creating a model is divided into five steps:

1. Choosing the input and output data for the supervised
learning process.
2. Data normalization (both input and output).
3. The normalized data is trained using neural network
learning.
4. Checking the model's goodness of fit.
5. Making a comparison between the projected and desired
Fig. 1. Proposed framework
outcomes.
Layers, or subgroups of processing modules, make up a V. EXPERIMENT AL RESULT S
layered feed-forward neural network. A layer of processing This system uses a deep learning algorithm in Python as the
components performs separate calculations on data and front end and MYSQL as the database to forecast heart disease.
transmits the results to another layer. The follo wing layer may We may upload data with many characteristics and save it as a
then perform its calculations and send the results to yet another CSV file in this study. The following is the basic attribute
layer. Finally, the network's output is determined by a description are shown in Table 1.
collection of one or more processing components in a
subgroup. Each processing element calculates based on a Finally, use a deep learning system to detect heart disorders
weighted average of its inputs. The first and last layers are the such as high blood pressure, cardiac arrest, coronary artery
input and output layers. The layers between these two are disease, and arrhythmia. The specifics as testing data may be
known as the hidden layers. Cells, neuro mas, and artificial provided by a new user, and classification can be performed to
neurons are all terms used to describe processing components anticipate and diagnose different heart illnesses with a lower
that behave similarly to neurons in the brain. mistake rate. We can upload 200 patient files and acquire
samples fro m the UCI repository. In addition, 11
The following are the steps in the neural network characteristics were used to categorize the disorders, including
algorithm: cardiac datasets. systolic blood pressure, age, height, gender,
Step 1: Randomly initialize the weights and biases. weight, diastolic blood pressure, cholesterol, alcohol, glucose,
smoking, and a state of being active are all included in the
Step 2: Into the machine, feed the training sample. dataset. The training status can be shown in Figure 2.
Step 3: Carry the inputs forward; in the hidden and output
TABLE I. AT T RIBUT E DET AILS
layers, compute the net input and output of each unit.
Attribute Description
Step 4: Return the fault to the concealed layer. User Id Name
Age In years
Adjust the weights and biases to through back the Sex 1 = Female 2 = Male
propagating errors. Weight Kilogram
Step 5: The weight and bias of a network are automatically Height Centimeter
Ap high Systolic Blood
adjusted using training and learning functions, which are pressure
mathematical techniques. Ap low Diastolic Blood
pressure
Step 6: terminating condition
Cholesterol 1: Average 2: Over
The neural network algorith m beats standard machine average 3:
Considerably above
learning techniques based on these phases. The suggested average
architecture is shown in Figure 1. Glucose 1: Average 2: Over
average 3:
Considerably above
average
Cardio T arget variable
Active Binary features
Alcohol Binary features
Smoke Whether the patient
smokes or not

978-1-6654-8328-5/22/$31.00 ©2022 IEEE 1402

Authorized licensed use limited to: National Institute of Technology. Downloaded on July 27,2023 at 16:37:20 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Sixth International Conference on Trends in Electronics and Informatics (ICOEI 2022)
IEEE Xplore Part Number: CFP22J32-ART; ISBN: 978-1-6654-8328-5

TABLE II. PERFORMANCE T ABLE

Algorithm/ Recall Precision F- me asure
Performance measures
SVM 82 44 57
BPNN 88 46 60
Naives Bayes 80 42 55

Precision

46
45
44

Fig. 2. T raining process 43

42
41
40
Naives Bayes SVM BPNN

Fig. 4(a). Precision

Recall

88
86
84
82
80
78
Fig. 3. Diagnosis information
76
In a testing phase, we can predict multip le heart diseases Naives Bayes SVM BPNN
with prescription details, and it is shown in Figure 3.
We can assess each algorithm's performance and compare it Fig. 4(b). Recall
based on the accuracy parameter.
F-measure parameter analysis can be used to evaluate the
system's performance. Recall, Precision, and F-measure are
used to assess the system's effectiveness.
Recall ={TP/(TP+FN)}
(1)
Precision ={TP/(TP+FP)}
(2)
F measure ={2* (Precision*Recall)/(Precision+Recall)}
(3)
The result of performance review is shown in following
Table 2 and shows in Figure 4(a),4(b),4(c).

978-1-6654-8328-5/22/$31.00 ©2022 IEEE 1403

Authorized licensed use limited to: National Institute of Technology. Downloaded on July 27,2023 at 16:37:20 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Sixth International Conference on Trends in Electronics and Informatics (ICOEI 2022)
IEEE Xplore Part Number: CFP22J32-ART; ISBN: 978-1-6654-8328-5

REFERENCES
F- measure [1] Singh, Manmohan, Harish Nagar, and Anjali Sant. "K-mean and EM
Clustering algorithm using attendance performance improvement
Primary school Student." (2016).
[2] Papanikolaou, Yannis, Ioannis Katakis, and Grigorios T soumakas.
60 "Hierarchical Partitioning of the Output Space in Multi-label Data."
59 arXiv preprint arXiv:1612.06083 (2016).
58 [3] Sajana, T ., CM Sheela Rani, and K. V. Narayana. "A survey on
clustering techniques for big data mining." Indian Journal of Science and
57 T echnology9.3 (2016).
56 [4] Jenneth, A., and K. Thangavel. "Dimensionality Reduction based on
55 Hubness Property using Feature Weighting Method for Clustering."
Indian Journal of Science and T echnology 9.19 (2016).
54 [5] Aouad, Lamine M., Nhien-An Le-Khac, and Tahar Kechadi. "variance-
53 based clustering technique for distributed data mining applications."
arXiv preprint arXiv:1703.09823 (2017).
52
[6] Lashari, Saima Anwar, and Rosziati Ibrahim. "Comparative analysis of
Naives Bayes SVM BPNN
data mining techniques for medical data classification." 4th International
Conference on Computing and Informations. 2013.
Fig. 4(c). F-Measure [7] Sharma, Anil, and Balrajpreet Kaur. "A Research Review On
Comparative Analysis Of Data Mining Tools, T echniques And
The proposed neural network approach, as shown in Figure Parameters." International Journal of Advanced Research in Comput er
4(c), has higher-level F-measure values than the current Naives Science 8.7 (2017).
Bayes and SVM methods . [8] Wisaeng, Kittipol. "An empirical comparison of data mining techniques
in medical databases." International Journal of Computer Applications
VI. CONCLUSION 77.7 (2013).
The difficulty of restricting and summarizing diverse data [9] David, Satish Kumar, et.al,... "Comparative analysis of data mining tools
and classification techniques using weka in medical bioinformatics."
mining techniques used in the field of med ical prediction is Computer Engineering and Intelligent Systems 4.13 (2013): 28-38.
examined in this work. For intelligent and successful heart
[10] Pon Periasamy. "A Review on Health Data Using Data Mining
disease prediction using data mining, the focus is on combining Techniques." International Research Journal of Engineering and
various methodologies and combinations of several target T echnology (IRJET) 2.07 (2015): 2395-0056.
criteria. Data mining is a useful tool for deriving meaningful [11] Vijayakumar, T. "Posed Inverse Problem Rectification Using Novel
medical rules fro m medical data, and it plays a significant role Deep Convolutional Neural Network." Journal of Innovative Image
in illness prediction and clinical d iagnosis. There is a growing Processing (JIIP) 2, no. 03 (2020): 121-127.
interest in utilizing categorization to determine whether or not [12] Pandian, A. Pasumpon. "Performance Evaluation and Comparison using
a disease is present. The current study used a huge sample of Deep Learning Techniques in Sentiment Analysis." Journal of Soft
Computing Paradigm 3, no. 2: 123-134.
hospitalized patients to demonstrate classification. The
[13] Hamdan, Yasir Babiker. "Construction of Statistical SVM based
classification technique is extremely sensitive to noisy data. If Recognition Model for Handwritten Character Recognition." Journal of
there is any noisy data present, it presents major challenges in Information Technology 3, no. 02 (2021): 92-107.
terms of classification processing capacity. It not only slows [14] Tripathi, Milan. "Sentiment Analysis of Nepali COVID19 T weets Using
down but also impairs the performance of the classification NB, SVM AND LSTM." Journal of Artificial Intelligence 3, no. 03
algorithm. As a result, before using a classification method, all (2021): 151-168.
attributes that will subsequently act as noisy attributes must be [15] Swain, Debabrata, Santosh Pani, and Debabala Swain. 2019, April. “An
removed from datasets. We can apply preprocessing processes efficient system for the prediction of Coronary artery disease using
dense neural network with hyper parameter tuning”. International
and classification rule algorithms, such as Multi-layer Journal of Innovative T echnology and Exploring Engineering (IJITEE)
perceptron, in this research effort for classifying datasets that 8 (6S). 689-695.
are supplied by users. The Multi-layer perceptron technique [16] Miranda E, Irwansyah E, Amelga AY, Kom S, Maribondang MM, Kom
produces superior outcomes than other strategies, according to S, Salim M and Kom S.2016. Detection of cardiovascular disease risk’s
the testing data. Other data mining techniques and algorithms level for adults using Naive Bayes classifier. Healthc Inf Res. 22(3)
are likely to be used in the future to improve performance 196–205.
efficiency.

978-1-6654-8328-5/22/$31.00 ©2022 IEEE 1404

Authorized licensed use limited to: National Institute of Technology. Downloaded on July 27,2023 at 16:37:20 UTC from IEEE Xplore. Restrictions apply.