Proceedings of the Third International Conference on Intelligent Sustainable Systems [ICISS 2020]

IEEE Xplore Part Number: CFP20M19-ART; ISBN: 978-1-7281-7089-3

Survey of Heart Disease Prediction and

Identification using Machine Learning Approaches

Ramya G. Franklin Dr. B. Muthukumar

2020 3rd International Conference on Intelligent Sustainable Systems (ICISS) | 978-1-7281-7089-3/20/$31.00 ©2020 IEEE | DOI: 10.1109/ICISS49785.2020.9316119

Research Scholar, School of Computing Professor and Head

Sathyabama Institute of Science and Technology Department of CSE, Mohammad Sathak College
Chennai, India Chennai, India
[email protected] m [email protected]

Abstract - Heart disease is highlighted as the major one prompt treatment and saving a human life. The requirement of
among the various death factors. Detecting heart disease tends to DM techniques aids in the collection of valuable data and
be a bit complex due to insufficient knowledge and experience of information considering multiple points of view. The current
the medical practitioners concerning warning signs of heart research work incorporates predictive mining to build a
failure. There exist innumerable data volume s in the healthcare
sector. By adopting the best appropriate data mining techniques, predictive and diagnosis system concerning heart diseases.
early detection of heart-related diseases can be achieved and also Heart disease is a term that assigns to a large number of
preventing it from occurring. Both the Machine Learning (ML) medical conditions related to the heart. The term ‘medical
and Data Mining (DM) techniques prove to be effective and conditions’ depicts the imbalanced health disorders, which
significant in the domain of the medical industry. The objective means have a direct impact on the human heart and also in the
of the current research work is to examine various risk rest of the body parts. Since heart disease portrays a prime
parameters highlighted in the investigation of Heart disease and issue of concern these days, the article targets to discover
also it targets to discover multiple techniques for the multiple techniques for the identification and prediction of
identification and prediction of heart disease along with
heart disease.
evaluating the drawbacks of the existing work. The article utilizes
the DM techniques to summarize existing researches concerning A DM technique is utilized to summarize the existing
heart disease prediction, examining a combination of DM research concerning heart disease prediction. A combination
techniques to reveal the best suitable and effective technique. The of DM techniques is to reveal the best suitable and effective
CNN, LS TM is being proposed for heart disease identification
technique. Naive Bayesian (NB) algorithm is being proposed
and prediction which yields in improvised output than the rest of
the prevailing techniques. Various levels involved in the proposed for heart disease identification and prediction. An ED&P
approach are a collection of dataset, training, and testing, approach is adopted for building a Data Mining (DM)
collection of user symptoms, securely forwarding the data by approach using Machine Learning (ML) techniques to achieve
utilizing AES , and eventually, result is being generated in PDF prior detection and prevention of malignancy of heart disease.
format. Comparative performance in datasets concerning The NB approach yields in improvised output than the rest of
medicine is adopted in predicting heart disease techniques in the prevailing techniques. The present research work has the
comparison to the rest of other ML approaches. following stages: assembles online datasets from the UCI
medical dataset, then is the process of classification which
Keywords—Heart Disease, DM techniques, MI algorithms,
LSTM, CNN involves training and testing, collection of user symptoms,
data entered by the user is being securely forwarded via AES
as the data is private and saved in open databases, lastly the
I. INT RODUCT ION results are generated in PDF format. Comparative
According to today’s imbalanced lifestyle, the human heart performances in datasets concerning medicine are adopted in
is prone to critical sort of diseases. It’s resulting in critical predicting Heart disease techniques in comparison to the rest
diseases and problems arising due to diabetes, stress , and of other ML approaches . The proposed technique is highly
intense smoking. All these factors seriously affect the human important that handles classification very effectively, thereby
heart and lead to a variety of heart diseases. The major reason approximating ML regarding NB model.
for heart disease is blockage in coronary arteries, which
The following is the class ification of the work: Section 2
mainly supplies blood to the heart. Further, this leads to illustrates the workings of the previous author along with its
reduced nutrition to the heart muscle i.e., myocardial cells. benefits and drawbacks. Section 3 discusses the proposed
Generally, the blood supply reaching the heart is done by three technique of Machine Learning approaches for the detection
primary arteries. Blockage in any one of the artery results in of heart-related diseases along with the aspects of various
Heart Attack or Myocardial Infarction. As a result, this can levels. Finally, Section 4 presents the conclusion and puts
lead to life-threatening issues, sometimes death. The early forwards research work for the near future.
these symptoms are detected, the more are the chances for

978-1-7281-7089-3/20/$31.00 ©2020 IEEE 553

Authorized licensed use limited to: UNIVERSITY OF CONNECTICUT. Downloaded on May 18,2021 at 08:46:51 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Third International Conference on Intelligent Sustainable Systems [ICISS 2020]
IEEE Xplore Part Number: CFP20M19-ART; ISBN: 978-1-7281-7089-3

II. RELAT ED W ORKS employing DM tools, it’s revealed that a different result is
Purushottama. C, Kanak Saxenab, Richa Sharma[1] being generated with every DM tool even though with the
proposes an effective framework that can detect the views for same dataset and with a separate classification algorithm.
predicting risk factors in patients based on their available ChalaBeyene, Pooja Kamat[5] survey reveals that not all
health factors. The study aims to provide an insight into the medical practitioners possess equivalent skills and experience
specialized medical practitioners in taking the best measures for taking a precise decision. Some doctors lack in taking
concerning risk level related to heart disease. The algorithms appropriate decisions which results in a critical situation for
used such as SVM, CMAR, Bayesian Classifiers and C4.5. the patients. To resolve such issues, predicting disease
The advantages of this method are the proposed system occurrence becomes mandatory. The algorithms used such as
formulates the rules that are listed as Original-Rules, Pruned- SVM, Decision Tree (DT), K-Nearest Neighborhood (KNN),
Rules, Rules without duplicates, Classified-Rules, Sorted- Naïve Bayes (NB) and Artificial Neural Network (ANN). The
Rules, and Polish. The disadvantage of this method is the disadvantage of this method is that it does not predict the
system is being trained and tested by utilizing10 fold methods. performance of every algorithm for implementing the
The proposed framework is evaluated for arrangement proposed system. The advantages of this method are to
accuracy and the result illustrates that it’s highly capable of enhance and improvises the present system in decision making
accurately predicting the risk level related to heart/coronary by incorporating various feature selection methods and
diseases. algorithms. The proposed approach helps in the prediction of
Chaithra N and Madhu B [2] examines different sorts of heart disease occurrence to achieve a prior automatic
DM approaches used to build up a predictive model diagnosis and fetching result within a short time span. This
concerning cardiovascular diseases based on the data retrieved aids in achieving Quality of Services (QoS) and minimizes the
from transthoracic echocardiography. The algorithms used cost factor in saving patient’s life.
such as, Decision Tree (DT),J48, Naive Bayes (NB) and Manpreet Singh, Levi Monteiro Martins, Patrick Joanis,
Neural Network (NN). The advantages of this method are to and Vijay K. Mago [6] proposed HDPS (heart disease
measure the performance of the Neural Network (NN) is much prediction system) that relies upon SEM -Structural Equation
better in anticipating cardiac diseases. The disadvantage of Modelling and FCM - Fuzzy Cognitive Map. The dataset
this method is that the algorithm offers a low true negative rate utilized is CCHS - (Canadian Community Health Survey)
the junior cardiologists have ease to access it. Here, all the 2012.The algorithms used such as Structural Equation
three algorithms performance revealed the best true negative Modelling (SEM), and Fuzzy Cognitive Map (FCP). The
rate, making it a convenient tool for training junior disadvantage of this method is to predict the time involved for
cardiologists and medical students for diagnosing heart heart disease is quiet high. The advantages of this method
disease. reveals that the proposed model yields an accuracy of 74%. To
Kipp W. Johnson, BS, Jessica Torres Soto, MS, Benjamin predict the heart disease 80% of the dataset was utilized for
S. Glicksberg [3] offers a benchmark for medical practitioners training Structural Equation Modelling and 20% of the dataset
on significant attributes of Artificial Intelligence (AI) and for testing the Fuzzy Cognitive Map model.
Machine Learning (ML), reviews-selected applications of the Jaymin Patel, Prof.Tejal Upadhyay, Dr. Samir Patel [7]
above techniques in cardiology till date thereby determining as perform a comparison among various techniques of Decision
to how cardiovascular medicine can adopt AI in the coming Tree (DT) classification revealing the one providing effective
future. The algorithms used such as Neural Networks (NN) performance concerning the diagnosis of heart disease by
and Deep Learning (DL). The disadvantage of this method is adopting WEKA. Algorithms that are being tested includes:
poised to impact all possible phase of human state including Logistic model tree algorithm, J48, Decision Tree (DT) and
cardiology. The advantage of this method is the supervised Random Forest algorithm. The disadvantage of this method is
learning approach and reviews -selected applications in accuracy and scalability of the system need to be improvised.
cardiology are in the similar fields. At present this technique The advantage of this method is to classify the system being
of Machine Learning (ML) is carried out by individuals ideal and capable enough for applying to sizeable data. The
having specialized training. Although in near future adopting system aims to fetch or mine hidden patterns with the help of
this approach would be much feasible and hassle free- DM techniques and also anticipating heart disease in patients
Sarangam Kodati & Dr. R. Vivekanandam [4] examines where the presence is rated from, no presence to likely
some factors to anticipate heart diseases and proposes Heart presence.
Diseases Prediction System (HDPS) relying upon the Data Shalet K.S, V. Sabarinathan, V. Sugumaran, V. J. Sarath
Mining (DM) techniques. The algorithms used such as SVM, Kumar [8] proposes Heart Disease Prediction System (HDPS)
Naïve Bayes (NB), Random Forest and KNN algorithm. The having the potential to estimate an individual’s heat condition.
disadvantage of this method is the DM approach uses only The algorithms used such as Decision Tree (DT), and Support
WEKA and ORANGE tools can be detected. The advantages Vector Machine (SVM) Classifier. The disadvantage of this
of this method are when implemented on patients the valuable method is that to reveal the resting electrocardiographic output
decisions are withdrawn based on this hidden information and sex need not enhance overall results performance. The
using data mining. According to the output achieved by

978-1-7281-7089-3/20/$31.00 ©2020 IEEE 554

Authorized licensed use limited to: UNIVERSITY OF CONNECTICUT. Downloaded on May 18,2021 at 08:46:51 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Third International Conference on Intelligent Sustainable Systems [ICISS 2020]
IEEE Xplore Part Number: CFP20M19-ART; ISBN: 978-1-7281-7089-3

advantages of this method is that certain factors such as chest Illustrating secure execution and examination reveals
pain, age, type, thalassemia, colored fluoroscopy and exercise feasibility in the Cloud Computing (CC) situation. The
induced chest pain are categorized by employing the classifier. algorithms used such as Naive Bayes (NB), and AES. The
This research yield an accuracy of 82.35%and the algorithm of disadvantages of this method are allowing controlled sharing
DT is adopted for the process of feature selection and the of acquired information with patient’s family using the benefit
SVM classifier is adopted for classification. of the social networking platform. The advantage of this
method is that the AES tends to be an encryption algorithm
KaanUyar, Ahmet İlhan [9] proposes a few computational
which secures sensitive but unclassified information. CAM -
techniques for examining heart diseases. The algorithms used
Cloud assisted privacy-preserving mobile health monitoring
such as GA - Genetic algorithm, RFNN - recurrent fuzzy
system aids in transferring the computational overload to the
neural networks. The disadvantage of this method are needs to
cloud by incorporating the newly framed technique of key
be accompanied by medical practitioners for considering
private proxy re-encryption.
various parameters due to which decision-making ability of
the method can get affected. The advantages of this method is K. Rajalakshmi and K. Nirmala[13] discuss that the
that altogether 297 instances of patient data is used, from medical industry can't mine an entire set of information and
which 252 are adopted for training and 45 are being adopted knowledge from databases related to heart disease. The tool
for testing. An accuracy of 97.78% accuracy is achieved from being utilized is MapReduce using Hive Database within the
the testing set. The investigation is carried out successfully Hadoop open-source framework. The algorithms used such as
with the help of the heart disease testing dataset. The factors K-Means, WAC - Weighted Associative Classifier, and
being computed are: accuracy, RMSE - Root Means Square Prediction Tree C5.0. The disadvantage of this method is
Error, probability of the misclassification error, sensitivity, cardiologist lacks in providing first aid advice. The advantages
specificity, F-score, and precision. of this method is WAC offers an effective medium for
imbibing the classification process using Association Rule
S. Bagavathy, V. Gomathy, S. Sheeba Rani, Sujatha. K,
Discovery that aids in enhancing the prediction accuracy for
Bhuvana. M.K, Monica Murugesan [10] surveys multiple
classification. The integrated technique of DT along with
techniques of data mining is to fetch information from a huge
WAC and K-Means offers improvised prediction outcomes
dataset. The study is carried out on K-Means Clustering
concerning Heart Disease symptoms. The proposed techniques
Algorithm which is then compared with Map Reduce
offer an effective and accurate way of predicting heart disease.
Algorithm for heart disease detection. The algorithms used
such as SVM, and K-NN. The disadvantage of this method is Yildirim et al. [14], the author has proposed a DBLSTM-
the latency found because of batch processing. The advantages WS model which gives a high recognition performance of
of this method is that the K-means clustering algorithm is 99.39%. It is observed that the wavelet-based layer proposed
being compared with that of Map Reduce Algorith m’s within the study significantly improves the popularity
implementation efficiency in both parallel and DS (distributed performance of conventional networks.
systems).Conclusion: Map Reduce Algorithm yields better
Marinho, L. B., de MM Nascimento, N., Souza, J. W. M.,
accuracy compared to the K-Means Clustering Algorithm as it Gurgel, M. V., Rebouças Filho, P. P., & de Albuquerque[15],
offers dynamic schema and linear scaling. have proposed a system. The algorithms used in the system are
Palash Ukey, Vinit More, Pravin Patil, Amruta Sankhe[11] NB, SVM and OPF. The accuracy of this proposed model is
have built a web application using DM Techniques that aids in 94.30%. Li, Z., Zhou, D., Wan, L., Li, J., & Mou, W.[17] have
detecting multiple heart-related issues, thereby offering a used a deep residual network to build their model whose
resolution for the same. The project formed is termed heart accuracy was measured as 99.06% to classify the heartbeat.
disease prediction which involves three modules. Every Pandey, S. K., & Janghel, R. R.[18]
module has a few sub-modules. Also, there is a provision of
security in each module. The algorithms used such as III. M ACHINE LEARNING APPROACHES
Artificial Neural Network (ANN), and Naive Bayes (NB). The Table I: Various approaches of ML Techniques
disadvantage of this method is that it is difficult to provide Authors Year Approach Accuracy
system maintenance. The advantage of this method is that the Purushottama 89.7%
system provides automatic results along with the concerned et al. [1] SVM, CMAR, Bayesian
2016
doctors for subsequent treatment. An online software program Classifiers and C4.5
is used to predict the heart disease which is thoroughly
automated. The ultimate benefit of the proposed system is that Chaithra N et al. Decision Tree (DT),J48, 95.66%
it can be used by every individual in the scheduling processor [2] 2018 NB (Naive Bayes) and NN
as administrators. Also, it makes user’s tasks much more (Neural Network)
secure and effective. Kipp W. 94.28%
NN (Neural Networks) and
Johnson et al. 2018
Arun. R and N. Deepa[12] put forth cardiovascular illness. Deep Learning.
[3]
The varied types of symptoms and innumerable powerful
SarangamKodati SVM, NB (Naïve Bayes), 91.37%
intercessions are believably startling and confusing. 2018
et al. [4] Random Forest and KNN

978-1-7281-7089-3/20/$31.00 ©2020 IEEE 555

Authorized licensed use limited to: UNIVERSITY OF CONNECTICUT. Downloaded on May 18,2021 at 08:46:51 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Third International Conference on Intelligent Sustainable Systems [ICISS 2020]
IEEE Xplore Part Number: CFP20M19-ART; ISBN: 978-1-7281-7089-3

Authors Year Approach Accuracy [3] Kipp W. Johnson, BS, Jessica T orres Soto, MS, Benjamin S.
Glicksberg (2018), “Artificial Intelligence in Cardiology”, Elsevier,
ChalaBeyene et SVM, DT (Decision Tree), 94.56% Journal Of T he American College Of Cardiology, Vol. 71, No. 23, pp.
al. [5] KNN (K-Nearest 2668 - 2679.
2018 Neighborhood), NB (Naïve [4] SarangamKodati& Dr. R. Vivekanandam (2018), “Analysis of Heart
Disease using in Data Mining T ools Orange and WEKA”, Global
Bayes) and ANN (Artificial Journal of Computer Science and T echnology, Vol. 18.
Neural Network) [5] ChalaBeyene, Pooja Kamat (2018), “Survey on Prediction and Analysis
Manpreet Singh SEM -Structural Equation 80.00% the Occurrence of Heart Disease Using Data Mining T echniques”,
International Journal of Pure and Applied Mathematics, Vol. 118, No.
et al. [6] 2016 Modelling and FCM - 8, pp. 165-174.
Fuzzy Cognitive Map [6] Manpreet Singh, Levi Monteiro Martins, Patrick Joanis, and Vijay K.
KaanUyar et al. Genetic algorithm, RFNN - 97.78% Mago (2016), “Building a Cardiovascular Disease Predict ive Model
using Structural Equation Model & Fuzzy Cognitive Map”, IEEE,
[9] 2017 recurrent fuzzy neural International Conference on Fuzzy Systems (FUZZ), pp. 1377 -1382.
networks [7] Jaymin Patel, Prof.T ejalUpadhyay, Dr. Samir Patel (2015), “ Heart
S. Bagavathy et 89.00% Disease Prediction Using Machine learning and Data Mining
2018 SVM and K-NN T echnique”, IJCSC, Vol.7, pp.129-137.
al. [10]
[8] Shalet K.S, V. Sabarinathan, V. Sugumaran, V. J. Sarath Kumar
Palash Ukey et Artificial Neural Network, 92.3% (2015), “Diagnosis of Heart Disease Using Decision T ree and SVM
2018
al. [11] NB Classifier”, International Journal of Applied Engineering Research,
Yildirim et al. Vol. 10, No.68, pp. 598-602.
2018 LSTM 99.25% [9] KaanUyar, Ahmet ?lhan (2017), “Diagnosis of heart disease using
[14] genetic algorithm based trained recurrent fuzzy neural networks”,
Marinho et Elsevier B.V, International Conference on Theory and Application of
2019 NB, SVM, OPF 94.30%
al.[15] Soft Computing, pp. 588–593.
Yildirim et [10] S. Bagavathy, V. Gomathy, S. Sheeba Rani, Sujatha. K, Bhuvana. M.K,
2019 CAE and LSTM 99.00% Monica Murugesan (2018), “Early Heart Disease Detection Using Data
al.[16] Mining T echniques with Hadoop Map Reduce”, International Journal
Li et al.[17] 2020 Deep residual network 99.06% of Pure and Applied Mathematics, Vol. 119, No. 12, pp .1915-1920.
Pandey et [11] Palash Ukey, Vinit More, Pravin Patil, Amruta Sankhe (2018), “ Heart
2020 LSTM 99.37% Disease Prediction Using Data Mining”, IOSR Journal of Engineering,
al.[18]
Vol. 4, pp. 64-67.
Zheng, Z et al [12] Arun. R and N. Deepa (2018), “Heart Disease Prediction System
2020 CNN-LSTM 99.01%
[19] Using Naive Bayes”, International Journal of Pure and Applied
Artetxe, A et al., 77.00% Mathematics, Vol. 119, No. 16, pp. 3053 -3065.
2018 Bayesian [13] K. Rajalakshmi and K. Nirmala (2016), “Heart Disease Prediction with
[21] MapReduce by using Weighted Association Classifier and K-Means”,
Krittanawong et 85.00% Indian Journal of Science and T echnology, Vol. 9, No. 19, pp.1 -7.
2018 CNN-MLP [14] Yildirim, Ö. (2018). A novel wavelet sequence based on a deep
al., C[22]
Rad, A. B et al., bidirectional LST M network model for ECG signal classification.
2019 RNN-LSTM Computers in biology and medicine, 96, 189-202.
[23] 78% [15] Marinho, L. B., de MM Nascimento, N., Souza, J. W. M., Gurgel, M.
Shakya et al., V., Rebouças Filho, P. P., & de Albuquerque, V. H. C. (2019). A novel
2020 SVM,KNN,RF electrocardiogram feature extraction approach for cardiac arrhythmia
[25]
classification. Future Generation Computer Systems, 97, 564 -577.
[16] Yildirim, O., Baloglu, U. B., T an, R. S., Ciaccio, E. J., & Acharya, U.
R. (2019). A new approach for arrhythmia classification using deep
IV. CONCLUSION coded features and LST M networks. Computer methods and programs
in biomedicine, 176, 121-133.
The proposed method illustrates an effective Heart Disease [17] Li, Z., Zhou, D., Wan, L., Li, J., & Mou, W. (2020). Heartbeat
Prediction System by employing the technique of Data classification using deep residual convolutional neural network from 2-
Mining. An online software program is used in the prediction lead electrocardiogram. Journal of Electrocardiology, 58, 105 -112.
[18] Pandey, S. K., & Janghel, R. R. (2020). Automatic arrhythmia
of heart disease automatically. Also, it makes user’s tasks recognition from electrocardiogram signals using different feature
much more secure and effective. The encryption complexity methods with long short-term memory network model. Signal, Image
can be enhanced using the proposed technique with LSTM and and Video Processing, 1-9.
CNN heart disease prediction and prior automatic diagnosis [19] Zheng, Z., Chen, Z., Hu, F., Zhu, J., Tang, Q., & Liang, Y. (2020). An
Automatic Diagnosis of Arrhyt hmias Using a Combination of CNN
can be achieved. Also, it offers reduces the time to retrieve and LST M T echnology. Electronics, 9(1), 121.
output and delivers qualities of services and minimizes cost [20] Vijaykumar, T . "Classification of Brain Cancer type using Machine
involved in saving life of an individual. Learning." Journal of Artificial Intelligence 1, no. 2 (2019): 105 -113.
[21] Artetxe, A., Beristain, A., & Grana, M. (2018). Predictive models for
hospital readmission risk: A systematic review of methods. Computer
References methods and programs in biomedicine, 164, 49-64.
[22] Krittanawong, C., Bomback, A. S., Baber, U., Bangalore, S., Messerli,
[1] Purushottama. C, Kanak Saxenab, Richa Sharma (2016), “Efficient F. H., & T ang, W. W. (2018). Future direction for using artificial
Heart Disease Prediction System”, Elsevier, Procedia Computer intelligence to predict and manage hypertension. Current hypertension
Science, No. 85, pp. 962 – 969. reports, 20(9), 75.
[2] Chaithra N and Madhu B (2018), “Classification Models on [23] Rad, A. B., Zabihi, M., Zhao, Z., Gabbouj, M., Katsaggelos, A. K., &
Cardiovascular Disease Prediction using Data Mining T echniques”, Särkkä, S. (2019). Automated Polysomnography Analysis for Detection
Journal of Cardiovascular Diseases & Diagnosis, Vol. 6, pp.1-4. of Non-Apneic and Non-Hypopneic Arousals using Feature

978-1-7281-7089-3/20/$31.00 ©2020 IEEE 556

Authorized licensed use limited to: UNIVERSITY OF CONNECTICUT. Downloaded on May 18,2021 at 08:46:51 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Third International Conference on Intelligent Sustainable Systems [ICISS 2020]
IEEE Xplore Part Number: CFP20M19-ART; ISBN: 978-1-7281-7089-3

Engineering and a Bidirectional LST M Network. arXiv preprint [25] Shakya, Subarna. "Analysis of Artificial Intelligence based Image
arXiv:1909.02971. Classification Techniques." Journal of Innovative Image Processing
[24] Bizopoulos, P., & Koutsouris, D. (2018). Deep learning in cardiology. (JIIP) 2, no. 01 (2020): 44-54.
IEEE reviews in biomedical engineering, 12, 168-193.

978-1-7281-7089-3/20/$31.00 ©2020 IEEE 557

Authorized licensed use limited to: UNIVERSITY OF CONNECTICUT. Downloaded on May 18,2021 at 08:46:51 UTC from IEEE Xplore. Restrictions apply.