Proceedings of the International Conference on Sustainable Computing and Data Communication Systems (ICSCDS-2022)

IEEE Xplore Part Number: CFP22AZ5-ART; ISBN: 978-1-6654-7884-7

Prediction of Risk in Cardiovascular Disease using

Machine Learning Algorithms
Nami Susan Kurian S.Sajithra F.Anne Jenefer
Assistant Professor Assistant Professor Assistant Professor
Department of Electronics and Department of Computer Science and Department of Electronics and
Communication Engineering Engineering Communication Engineering
Rajalakshmi Institute of Technology R.M.K Engineering College Panimalar Engineering College
2022 International Conference on Sustainable Computing and Data Communication Systems (ICSCDS) | 978-1-6654-7884-7/22/$31.00 ©2022 IEEE | DOI: 10.1109/ICSCDS53736.2022.9760879

Chennai, India Tamilnadu, India Chennai, India

namisusan777@gmail.com

Yuvasree R Swetha G
Kathu Sara Renji UG Scholar
UG Scholar
Assistant Professor Department of Electronics and
Department of Electronics and
Department of Biomedical Engineering Communication Engineering
Communication Engineering
V.S.B Engineering College Rajalakshmi Institute of Technology
Rajalakshmi Institute of Technology
Karur,India Chennai, India
Chennai, India

Abstract—Prediction of heart disease is one of the most the death risk by using prediction techniques of Machine
complex tasks in medical field and prior detection of heart Learning.
disease become an area of research to save patient lives. During
the pandemic period, the number of cardiac arrest cases at home Machine Learning is a buzzword for the past few years,
has drastically increased due to inaccurate predictions and
the reason might be its applications and increased number of
delay in seeking medical attention. The health care industry
works on processing of huge data and the solution for this is computation power, designing and implementation of better
machine learning. Data science process large amount of data to algorithms. In today’s world Market field, Machine Learning
make intelligent health care decisions thereby avoids risk and [1] plays a vital role. Its application and techniques
alert the patients. In this paper, a comparative analysis of implemented from self-driving cars to predicting severe and
different machine learning classifiers based on dataset to predict deadly diseases. It involves in building and implementing a
the chance of heart disease with minimal attributes. The ML Predictive model which is used to find a solution for a
algorithms used for HD prediction are K- Nearest Neighbour, Problem Statement. Heart is used to maintain continuous
Gradient Boosting Classifier, Support Vector Machine, Naive blood circulation in our body. There are many cases in the
Bayes, Logistic Regression and Random Forest algorithm. The
world related to heart diseases. The term Heart attack is one
paper also finds the correlation between different attributes and
hence using them efficiently for prediction of heart attack. of the most common diseases in recent days. It shows up
various symptoms like irregular Heartbeats, chest pain, and
Keywords—Decision Tree, Support Vector Machine, Random so on. people from all over the world suffers from
Forest; Naïve Bayes, K-means, Artificial Neural Network cardiovascular diseases which can cost their lives even.

In order to predict this disease, Machine Learning helps

I. INTRODUCTION people to detect the cardio related diseases with its algorithms
Prediction is one of vast area where this machine learning and techniques. In this paper, using some symptoms like,
concept is applied. Heart disease can be predicted by cholesterol level, chest pain, age factor and some other
collecting, analyzing the patient’s database and determining factors, Machine Learning predicts the heart disease for a
the possibility of occurrence of heart attack of that patient. person. Also, a comparative analysis of different classifiers
The field of data science has seen explosive development in was performed for the classification of the heart disease
the field of machine learning, a subfield of artificial dataset in order to predict cases with minimal attributes. The
intelligence. ML algorithms are used to perform a multiple ML algorithms used for heart disease prediction are K-
and large number of tasks such as prediction, classification, Nearest Neighbor, Decision tree, SVM, Naive Bayes,
decision making and several other factors. a survey says, that Stochastic Gradient Decent (SGD) and Random Forest
almost 35% of death is due cardio-related diseases all over algorithm [2]. This paper gives the idea of predicting heart
the world. Due to lack of diagnostic centres, certified doctors disease using ML algorithms. This various kind of effective
and sources that affect the accuracy of prediction, the earlier algorithms such as support vector classifier, random forest,
detection and treatment of cardio diseases becomes a KNN, naive bayes, decision tree and logistic regression are
complex process. With this concern, in recent years machine used to predict the heart disease with higher accuracy. The
learning techniques are used to make medical aid software as accuracy of each of these algorithms were found and
a supporting system for early diagnosis of heart disease. To compared. This not only provides better understanding of the
understand the design of data structure of disease and to patients’ health condition but also helps to avoid risks and
provide earlier treatment for the deceased people, numerous deaths.
machine learning algorithms are used. This can further reduce

978-1-6654-7884-7/22/$31.00 ©2022 IEEE 162

Authorized licensed use limited to: University of Wollongong. Downloaded on October 19,2024 at 06:06:07 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Sustainable Computing and Data Communication Systems (ICSCDS-2022)
IEEE Xplore Part Number: CFP22AZ5-ART; ISBN: 978-1-6654-7884-7

II. RELATED WORKS In [12], a machine learning method to enhance heart

failure patient support was proposed by authors. In this study,
In [3], authors proposed a paper in which the machine they provided a clinical decision support system (CDSS) to
learning methods are diagnosed for the diabetic patients. In assess the heart failure (HF) patients, delivering various
this paper, analysis of the medical data by applying data outputs such as an HF severity rating, HF-type prediction, as
mining techniques and considering the existing relationships well as a management interface that compares all different
between variables. In [4] two algorithms are used MAFIA patients' follow-ups.
(Maximal Frequent Item set algorithm) and K-Means
clustering for accuracy efficiency. III. MACHINE LEARNING TECHNIQUES FOR PREDICTION
In [5], the prediction was based on ID3 algorithm, which
uses gradient descent algorithm and information gain is the The numerous machine learning algorithms for the
criteria followed in this method. This method was analysis of parameters are explained below:
implemented for television and mobile phones. This A. Logistic Regression
suggested strategy enables the people to have a better
understanding of the disease but also it may aid to lessen the The supervised learning method includes well-known
death rate and count of afflicted persons. In [6] five different techniques like logistic regression. A set of independent
algorithms are compared and the simulation is done using the variables is utilised to predict the categorical dependent
rapid miner tool that results in higher accuracy than Matlab
variable. Categorical or discrete values must be returned by
and weka tool. The different models are compared based on
logistic regression. However, rather than providing the precise
the accuracy rate such as, Decision Tree, Random Forest,
value as 0 and 1, the probabilistic values that lie in 0 and 1 are
logistic regression, Naive Bayes and SVM classification
algorithms In [7], created a survey comprising different presented instead. Logistic regression is fundamentally
classification method which is utilized for predicting heart similar to linear regression algorithm but the logistic
disorders. The classification approaches employed where regression is widely used for classification problems., whereas
decision tree, neural networks, naive bayes, KNN algorithm linear regression finds solution for regression problems. It is
and accuracy of the classifiers was assessed for varied values an important algorithm because it has the capability of
of parameters. providing and classifying both continuous and discrete
datasets. The logistic regression function forms a curve like
In the paper [8] the training samples are taken as weighted shaped called sigmoid function or logistic regression function
in supervised locality preserving projection. The experimental because the between o and 1 must be a logistic regression's
results are done on three hyperspectral images and it is found value. Linear regression may be used to obtain the logistic
that FEWT has higher performance and better speed than regression equation. Regression equation can be derived by
other supervised feature extraction methods from the available using the straight-line equation:
set of training samples. In [9], heart disease classification
based on Principle component analysis was put forward for 𝑌 = 𝑏0 + 𝑏1 𝑋1 + 𝑏2 𝑋2 + 𝑏3 𝑋3 … 𝑏𝑁 𝑋𝑁
dimensionality reduction. Feed forward neural network using
back propagation algorithm was also put forward. The dataset The value of Y can be present between 0 and 1. So divide the
from UCI repository was taken. Accuracy was calculated with above equation by (1-Y)
reduced number of features.
𝑌 0 𝑓𝑜𝑟 𝑌 = 0
In [10] this suggested method, they contrasted the ways which = {
(1 − 𝑌) ∞ 𝑓𝑜𝑟 𝑌 = 1
utilizes the ECG signal is its own fact that it is totally based
on the HRV signal which can get from perhaps a very noisy Taking logarithm of the resulted equation results in logistic
ECG signal with a significantly better accuracy. regression equation.

In [11] writers worked on Deep learning technique for the 𝑌

log (1−𝑌) = 𝑏0 + 𝑏1 𝑋1 + 𝑏2 𝑋2 + 𝑏3 𝑋3 … 𝑏𝑁 𝑋𝑁
efficient and active categorization of ECG data. In this
research, they offered a paper in unique Methodological
technique deep learning based in active categorization of ECG B. Support Vector Machine(SVM)
waves. They illustrate an acceptable feature representation Among the ML algorithms, one of the widely utilized
from the raw ECG data in an unsupervised method using supervisory algorithms in classification and regression
stacked denoising auto - encoders (SDAEs) with sparsity evaluation is the support vector machine. An important
constraint. In [12], a machine learning method to enhance function performed by n-dimensional space by the SVM
heart failure patient support was proposed by authors. In this algorithm is deciding where fresh data values should be
study, they provided a clinical decision support system stored in the future. To generate a hyperplane, SVM [13]
(CDSS) to assess the heart failure (HF) patients, delivering selects the most local maxima. This approach is referred
various outputs such as an HF severity rating, HF-type described as a support vector machine algorithm because of
prediction, as well as a management interface that compares
the name given to these extreme points: support vectors. It is
all different patients' follow-ups.
necessary to discover the optimal value for the decision
boundary in order to locate the best data point in n number of

978-1-6654-7884-7/22/$31.00 ©2022 IEEE 163

Authorized licensed use limited to: University of Wollongong. Downloaded on October 19,2024 at 06:06:07 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Sustainable Computing and Data Communication Systems (ICSCDS-2022)
IEEE Xplore Part Number: CFP22AZ5-ART; ISBN: 978-1-6654-7884-7

dimensional spaces. So, the best boundary is based on the most influential algorithms. The prediction model is the form
identifying the hyperplane of SVM. The datapoints nearer to of decision trees and it outperforms random forest. The
the hyperplane is called support vector. SVM is more algorithm is so called gradient boosted trees when the DT is
commonly used for image classification and face detection. a weak learner. The most vital step in gradient boosting
The most challenging issue in SVM is the kernel selection method is regularization by shrinkage. The base estimator is
and also the method selection to avoid overfitting and fixed in this case and it is Decision Stump. This algorithm is
underfitting of dataset. The training data is mapped with used for prediction of both continuous target variable and
function called kernel. categorical target variable. The cost function varies based on
whether it is used for regression or classification. The cost
C. Random Forest Algorithm functions are Mean Square Error (MSE) for former and Log
loss for latter.
Automated decision-making is supported by supervised F. Naive Bayes Algorithm
learning in the Random Forest Algorithm. This technique is The algorithm follows Bayes rule for calculating probabilities
employed in machine learning to solve classification and and conditional probabilities with the assumption that
regression problems. Because of the idea of "ensemble attributes are statistically independent of each other. It is
learning," this is possible. Multiple classifiers are combined classifier that provides better accuracy and is widely used in
in order to find solution for a complicated problem that computer vison applications. It is used for very large dataset
improves the overall performance. The accuracy of the analysis and it outperforms other classification methods in
dataset's prediction can be improved by using a random terms of accuracy.
forest, which uses an average of many decision trees from the
given dataset. Random forest takes less time and it predicts 𝑑
𝑃 ( ) ∗ 𝑃(ℎ)
high accurate outputs even for large data set and runs 𝑃(ℎ|𝑑) = ℎ
efficiently and maintain accuracy though large amount of 𝑃(ℎ)
data is missed. Steps for random forest are:
P(h|d) - class posterior probability
P(h) - class prior probability
Step-1: X data points are randomly selected from the
P(d|h) - likelihood probability
original input training set
P(d) - prior probability of predictor
Step-2: Develop the decision trees as per the data points that
were selected in step one.
Step-3: Decide on how many decision trees you want to IV. METHODOLOGY
create.
A. Data Collection
Step-4: Repeat the Steps 1 & 2
The Cleveland HD dataset downloaded from Kaggle UCI
D. KNN Algorithm repository is used in this research work. The database consists
The supervised machine learning technique of k-nearest of 76 attributes among that 14 are taken into consideration for
neighbours may be used to address problems involving the research purpose. The 14 attributes are explained in Table
classification and regression analysis. Naive bayesian 1.
classification is used in the development of Decision Support
in the Prediction of Heart Disease System . A database of TABLE 1. DATASET FOR HEART DISEASE PREDICTION
heart disease cases from the past is mined by the algorithm to
unearth previously unknown information. Heart disease S. No Attribute Description
patients may be accurately predicted using this approach. 1 age In years
This algorithm is non-parametric and does not rely on any 2 sex male or female
assumptions or previous values. To put it another way: It's a 3 cp The kind of chest pain
lazy learning algorithm because it doesn't actually do 4 trestbps blood pressure level
anything during the classification process itself. The main
steps involved in this are: 5 chol serum cholestoral (mg/dl)
6 fbs blood sugar level - fasting
Step-1: Choose K neighbours 7 restecg resting ecg
Step-2: Find the distance between the K nearest neighbours
8 thalach threshold heart rate achieved
by using the Euclidean distance formula
Step-3: Determine the Euclidean distance of K nearest 9 exang exercise angina (induced)
neighbours 10 oldpeak State of depression
Step-4: Compute the total number of the data points 11 slope the slope of ST segment
Step-5: Allocate the newly attained datapoints to the one with 12 Thal Major vessels
greatest neighbour counts 13 ca Status of heart
E. Gradient Boosting Classifier 14 target Output class
It is a machine learning algorithm for evaluating
regression and classification problems, and it is one of the

978-1-6654-7884-7/22/$31.00 ©2022 IEEE 164

Authorized licensed use limited to: University of Wollongong. Downloaded on October 19,2024 at 06:06:07 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Sustainable Computing and Data Communication Systems (ICSCDS-2022)
IEEE Xplore Part Number: CFP22AZ5-ART; ISBN: 978-1-6654-7884-7

The total instances in the dataset are 1025 and is used for of 200. The only one attribute that is outside the outlier is
the data analysis in this work. The dataset undergoes pre- chol, having the value more than 400. Each of the attributes
processing to handle the missing values through statistical closer look is shown in Fig.5 and Fig.6.
techniques. The dataset is separated into two – training data
(80 percent ) and testing data (20 percent ).

B. Data Pre-processing

The original dataset cannot be used directly for analysis

as it contains many missing values that may results in
inaccuracies and wrong predictions. To improve the
efficiency in terms of accuracy of the model, missing values
should be replaced by some optimal way by using the “mean
of column” method. This works by taking the average value
of the neighborhood values. The 0 value gets replaced by the
calculated new value. Numeric values are converted to
nominal values to make the dataset compactible for the
analysis of ML techniques.

C. Building Model

The model was built using python software. The block

diagram of proposed protocol is shown in Fig 1 below. The
classification techniques are implemented over the training
data. The system is trained with 80% of training data. Testing Fig. 2. Histogram of different attributes
the model with the test data and the obtained result is matched
with the trained data to check the accuracy of the model.
Different ML algorithms have different method of
implementation and different accuracy scores.

Feature Pre- Machine

selection processing learning
dataset
algorithms

Input Prediction
from
patient

Fig. 3. Correlation Matrix

Comparison

Fig. 1. Workflow of heart disease prediction

The histogram image of different attributes is shown in

the Fig.2. All 14 attributes are taken into consideration for the
analysis. The different attributes are correlated together to
find the relation between each of the features and its
dependency. Fig.3 shows the correlation matrix of different
features. This image gives the value of correlation between
different features thereby provides a better understanding of
the attribute relations. The higher correlation between target
and chest pain is visible in Fig.4. There are few attributes
without any correction with other attributes whose value are
near to zero. Using the box plot finds any outliers if present.
In the above plot, most of the attributes are within the range

978-1-6654-7884-7/22/$31.00 ©2022 IEEE 165

Authorized licensed use limited to: University of Wollongong. Downloaded on October 19,2024 at 06:06:07 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Sustainable Computing and Data Communication Systems (ICSCDS-2022)
IEEE Xplore Part Number: CFP22AZ5-ART; ISBN: 978-1-6654-7884-7

predicted correctly and it is one of the most important

Fig. 4. Box plot to find the outlier parameters in ML analysis.

𝑇𝑃 + 𝑇𝑁
𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦 =
𝑇𝑃 + 𝑇𝑁 + 𝐹𝑁 + 𝐹𝑃

ROC AUC: ROC curve is a graph that gives the trade-off

between TPR and FPR. In order to check how good, the curve
is, area under the ROC curve need to be calculated. Higher
the area under the curve, higher will be the ROC AUC score.

𝑇𝑃𝑅
𝑁𝑜. 𝑜𝑓 𝑇𝑟𝑢𝑒 𝑃𝑜𝑠𝑖𝑡𝑖𝑣𝑒𝑠 (𝑇𝑃)
=
𝑁𝑜. 𝑜𝑓 𝑇𝑟𝑢𝑒 𝑃𝑜𝑠𝑖𝑡𝑖𝑣𝑒𝑠 (𝑇𝑃) + 𝐹𝑎𝑙𝑠𝑒 𝑁𝑒𝑔𝑎𝑡𝑖𝑣𝑒𝑠 (𝐹𝑁)

𝑇𝑃𝑅
Fig. 5. Closer look of each attribute outlier -scatter plot
𝑁𝑜. 𝑜𝑓 𝐹𝑎𝑙𝑠𝑒 𝑃𝑜𝑠𝑖𝑡𝑖𝑣𝑒𝑠 (𝑇𝑃)
=
𝑁𝑜. 𝑜𝑓 𝐹𝑎𝑙𝑠𝑒 𝑃𝑜𝑠𝑖𝑡𝑖𝑣𝑒𝑠 (𝐹𝑃) + 𝑇𝑟𝑢𝑒 𝑁𝑒𝑔𝑎𝑡𝑖𝑣𝑒𝑠 (𝑇𝑁)

TABLE II. COMPARISON OF RESULT OF DIFFERENT ML ALGORITHMS

Model Accuracy Score ROC AUC Score

Logistic 0.857 0.846
Regression
Support Vector 0.857 0.846
Machine
Classifier
Random Forest 0.885 0.892
Classifier
KNN 0.714 0.643
Gradient 0.885 0.892
Boosting
Algorithm
Naive Bayes 0.942 0.933
Fig. 6. Closer look of each attribute – whisker plot

V. RESULT AND DISCUSSION

The analysis shows that SVM outperforms other ML
classification methods in terms of accuracy. The parameters Accuracy Score
taken into consideration for the analysis are discussed in
Table II. Comparison of accuracy and ROC AUC for 1
different ML algorithms are shown in Fig.7 and Fig.8
0.9

0.8
Precision or Positive Predictive Value : If the model predicts
the positive class correctly, it is called True Positive and if 0.7
the model predicts the negative class correctly, it is called
0.6
True Negative.
0.5
𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛 (𝑝) LR SVM RFC KNN GBA Naive
𝑁𝑜. 𝑜𝑓 𝑇𝑟𝑢𝑒 𝑃𝑜𝑠𝑖𝑡𝑖𝑣𝑒𝑠 (𝑇𝑃) Bayes
=
𝑁𝑜. 𝑜𝑓 𝑇𝑟𝑢𝑒 𝑃𝑜𝑠𝑖𝑡𝑖𝑣𝑒𝑠 (𝑇𝑃) + 𝐹𝑎𝑙𝑠𝑒 𝑃𝑜𝑠𝑖𝑡𝑖𝑣𝑒𝑠 (𝐹𝑃)

Fig. 7. Comparison of Accuracy Score

Recall: It is the average probability of complete retrieval..
Accuracy : It is the measure of how many observations are

978-1-6654-7884-7/22/$31.00 ©2022 IEEE 166

Authorized licensed use limited to: University of Wollongong. Downloaded on October 19,2024 at 06:06:07 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Sustainable Computing and Data Communication Systems (ICSCDS-2022)
IEEE Xplore Part Number: CFP22AZ5-ART; ISBN: 978-1-6654-7884-7

Sustainable Communication Systems (ICESC), Coimbatore, India,

ROC AUC Score 2020, pp. 398-403, doi: 10.1109/ICESC48915.2020.9155850.
[8]. L. SathishKumar and A. P. A.Padmapriya, “Prediction for Common
1 Disease using ID3 Algorithm in Mobile Phone and Television,” Int.
J. Comput. Appl., vol. 50, no. 4, pp. 30–33, 2012, doi:
10.5120/7762-0830.
0.9
[9]. F. S. Alotaibi, “Implementation of machine learning model to
predict heart failure disease,” Int. J. Adv. Comput. Sci. Appl., vol.
0.8 10, no. 6, pp. 261–268, 2019, doi: 10.14569/ijacsa.2019.0100637.
[10]. R. Joshua Samuel Raj, T.Sudarson Rama Perumal,
0.7 N.Muthukumaran, ‘Road Accident Data Analytics Using Map -
Reduce Concept’, International Journal of Innovative Technology
0.6 and Exploring Engineering, Volume. 8, Issue. 11, pp. 1032- 1037,
September 2019.
0.5 [11]. M. Imani and H. Ghassemian, “Feature Extraction Using Weighted
LR SVM RFC KNN GBA Naive Training Samples,” IEEE Geosci. Remote Sens. Lett., vol. 12, no. 7,
Bayes pp. 1387–1391, 2015, doi: 10.1109/LGRS.2015.2402167.
[12]. Gautham A Nagendran, Harshmeet Singh, R Joshua Samuel Raj and
N. Muthukumaran, "Input Assistive Keyboards for People with
Disabilities: A Survey," 2021 Third International Conference on
Fig. 8. Comparison of ROC AUC Score Intelligent Communication Technologies and Virtual Mobile
Networks (ICICV), Tirunelveli, India, March 2021, pp. 829-832.
VI. CONCLUSION [13]. J. Prince Antony Joel, R. Joshua Samuel Raj and N. Muthukumaran,
"Cognitive and Cybernetics based Human Adaptive Mechatronics
Heart disease patients are increasing day by day and it System in Gait Rehabilitation Therapy," 2021 Third International
became crucial to develop a model for the accurate analysis Conference on Intelligent Communication Technologies and
of heart disease in prior. Machine learning plays a significant Virtual Mobile Networks (ICICV), Tirunelveli, India, March 2021,
pp. 516-521.
part in prediction and so establishing the effective algorithm [14]. T. Santhanam and E. P. Ephzibah, “Heart disease classification
for the previous identification of heart disease is a tough using PCA and feed forward neural networks,” Lect. Notes Comput.
issue. In this research, we have done the comparison study of Sci. (including Subser. Lect. Notes Artif. Intell. Lect. Notes
several machine learning algorithms and discovered that Bioinformatics), vol. 8284 LNAI, pp. 90–99, 2013, doi:
10.1007/978-3-319-03844-5_10.
Naive Bayes method beats other algorithm in terms of
[15]. G. P. Devaraj, R. Kabilan, J. Z. Gabriel, U. Muthuraman, N.
accuracy and ROC score. In future, the work could be Muthukumaran and R. Swetha, "Design and Analysis of Modified
extended for sensing the data from the patients to cloud Pre-Charge Sensing Circuit for STT-MRAM," 2021 Third
platform where the machine learning analysis could be International Conference on Intelligent Communication
Technologies and Virtual Mobile Networks (ICICV), Tirunelveli,
carried out in the cloud platform and the doctor can monitor India, March 2021, pp. 507-511.
the patient from remote. The remote monitoring provides [16]. B. M. Asl, S. K. Setarehdan, and M. Mohebbi, “Support vector
immediate consultation to save the lives of patient. These machine-based arrhythmia classification using reduced features of
developments in health care with machine learning not only heart rate variability signal,” Artif. Intell. Med., vol. 44, no. 1, pp.
51–64, 2008, doi: 10.1016/j.artmed.2008.04.007.
results in prior detection of diseases but also helps in accurate [17]. V. Sharma, “Heart Disease Prediction using Machine Learning
analysis resulting in better diagnosis. Techniques,” pp. 177–181, 2020.
[18]. G. Choudhary and S. Narayan Singh, “Prediction of heart disease
using machine learning algorithms,” Proc. Int. Conf. Smart
REFERENCES Technol. Comput. Electr. Electron. ICSTCEE 2020, vol. 7, pp. 197–
202, 2020, doi: 10.1109/ICSTCEE49637.2020.9276802.
[1]. M. Balakrishnan, A. B. Arockia Christopher, P. Ramprakash, and [19]. R. Kabilan, N. Muthukumaran, “A Neuromorphic Model for Image
A. Logeswari, “Prediction of Cardiovascular Disease using Recognition using SNN”, 2021 6th International Conference on
Machine Learning,” J. Phys. Conf. Ser., vol. 1767, no. 1, pp. 1–7,
Inventive Computation Technologies (ICICT), Coimbatore, India,
2021, doi: 10.1088/1742-6596/1767/1/012013. 2021, pp. 720-725.
[2]. VP. Anubala, N. Muthukumaran and R. Nikitha, ‘Performance [20]. J. Rebekah, D. C. J. W. Wise, D. Bhavani, P. Agatha Regina and N.
Analysis of Hookworm Detection using Deep Convolutional Neural
Muthukumaran, "Dress code Surveillance Using Deep learning,"
Network’, 2018 International Conference on Smart Systems and 2020 International Conference on Electronics and Sustainable
Inventive Technology, pp. 348-354, 2018, doi: Communication Systems (ICESC), Coimbatore, India, 2020, pp.
10.1109/ICSSIT.2018.8748645.
394-397, doi: 10.1109/ICESC48915.2020.9155668.
[3]. G. Parthiban and S. K. Srivatsa, “Applying Machine Learning [21]. G. Guidi, M. C. Pettenati, P. Melillo, and E. Iadanza, “A machine
Methods in Diagnosing Heart Disease for Diabetic Patients,” Int. J. learning system to improve heart failure patient assistance,” IEEE
Appl. Inf. Syst., vol. 3, no. 7, pp. 25–30, 2012, doi: 10.5120/ijais12-
J. Biomed. Heal. Informatics, vol. 18, no. 6, pp. 1750–1756, Nov.
450593. 2014, doi: 10.1109/JBHI.2014.2337752.
[4]. R. Kabilan, G. P. Devaraj, U. Muthuraman, N. Muthukumaran, J.
Z. Gabriel and R. Swetha, "Efficient Color Image Segmentation
Using Fastmap Algorithm," 2021 Third International Conference on
Intelligent Communication Technologies and Virtual Mobile
Networks (ICICV), Tirunelveli, India, March 2021, pp. 1134-1141.
[5]. S. Anantha Babu, R. Joshua Samuel Raj, Arul Xavier V.M and N.
Muthukumaran, "DCT based Enhanced Tchebichef Moment using
Huffman Encoding Algorithm (ETMH)," 2021 Third International
Conference on Intelligent Communication Technologies and
Virtual Mobile Networks (ICICV), Tirunelveli, India, March 2021,
pp. 522-527.
[6]. A. Anikumar and P. X. Shajan, “Heart Disease Detection Using
Data Miningtechniques,” no. 7, pp. 31–33, 2018.
[7]. S. Gayathri, D. C. J. W. Wise, P. B. Shamini and N. Muthukumaran,
"Image Analysis and Detection of Tea Leaf Disease using Deep
Learning," 2020 International Conference on Electronics and

978-1-6654-7884-7/22/$31.00 ©2022 IEEE 167

Authorized licensed use limited to: University of Wollongong. Downloaded on October 19,2024 at 06:06:07 UTC from IEEE Xplore. Restrictions apply.