INTERNATIONAL JOURNAL OF MULTIDISCIPLINARY RESEARCH AND TECHNOLOGY

ISSN 2582-7358
Impact Factor 6.328
Peer-Reviewed Journal
Effective prediction of Heart disease using Machine Learning Techniques
1Rajesh Kumar Pradhan, 2Priyabrata Sahu, 3Suvendu Ku Jena
1
Department of Computer Science Engineering and Applications
Indira Gandhi Institute of Technology, India
2
Department of Computer Science Engineering and Applications
Indira Gandhi Institute of Technology, India
3
Department of Computer Science Engineering &Applications
Indira Gandhi Institute of Technology, India

ABSTRACT–
One of the most essential research areas has always been applied in the area of medical development. And now days
globally, heart disease (HD), particularly coronary artery disease (CAD) commonly known as atherosclerosis is the leading
cause of death. However, the early detection system for cardiac disorders or heart disease, is one of these medical
applications to avoid the advanced cases, and reduce treatment costs. As a result, we presented a medical support system for
predicting heart disease, which would aid physicians in diagnosis and decision-making. In this work, we used machine
learning techniques such as Extra-Tree Classifier, Support Vector Machine, Logistic Regression, K-Nearest Neighbour and
Decision Tree are used to predict heart disease using data from medical files. Using the UCI data set, several experiments
have been undertaken to predict HD, and the results show that Extra-Tree Classifier beats both cross-validation and train-
test split approaches, with accuracy of 95.00 %, respectively.
Keywords—Heart Disease, coronary artery disease, atherosclerosis, machine learning, UCI heart disease data set.

I.INTRODUCTION

According to the World Health Organization (WHO), heart disease is one of the leading causes of mortality when the heart is
unable to pump oxygenated blood throughout the body. One kind of cardiovascular illness is coronary artery disease (CAD),
often known as atherosclerosis (CVD)[1].This condition causes plaque accumulation in the arteries due to cholesterol in the
circulation. Because of plaque buildup, this illness results in restricted or obstructed blood vessels and coronary arteries.
Cholesterol, calcium, and other chemicals make up this plaque. The plaque restricts blood flow to the coronary arteries as it
builds up. As a result, myocardial blood flow declines. This can result in symptoms like angina. Chest, shoulder, abdomen,
arms, and neck pain are all possibilities. The amount of oxygenated blood in the body decreases during this pain. Myocardial
ischemia is the medical term for this condition. The myocardium tissue dies when the coronary artery is nearly entirely
restricted, resulting in a heart attack (myocardial infarction) [2,3].

Developing and implementing a medical diagnostic support system (MDSS) to automate the categorization and prediction of
CVD looks to be crucial at this time. In order to give the best clinical suggestions, medical diagnostic research requires a better
level of precision and efficiency. Despite the fact that traditional MDSS has proved its capacity to handle the majority of
diagnostic problems, it has a lower accuracy factor and is unable to provide a correct diagnosis [4].

Machine learning techniques are currently being used to help produce more accurate HD forecasts based on medical data such
as demographic, symptom and assessment, ECG, and laboratory information. Several studies have been undertaken to identify
and predict heart disease. Heart disease was studied using machine learning techniques [5].

53
Volume 4 Issue 1
 INTERNATIONAL JOURNAL OF MULTIDISCIPLINARY RESEARCH AND TECHNOLOGY
ISSN 2582-7358
Impact Factor 6.328
Peer-Reviewed Journal

II. LITERATURE SURVEY

We have included a few selected articles from a literature review on heart disease diagnosis in this section. These studies
employed the same well-known databases, which we will compare performance with later.
Machine learning algorithms are used to predict many diseases, and many researchers, such as Kohali et al., have worked on
this. [7] Work was done on heart disease prediction using logistic regression, diabetes prediction using support vector machine,
and breast cancer prediction using Adaboot classifier, with the results showing that logistic regression has an accuracy of 87.1
%, support vector machine has an accuracy of 85.71 %, and Adaboot classifier has an accuracy of 98.57 %, which is good for
prediction.
This neural network integration method is described in and is used to create new models by combining anticipated values from
prior models. The accuracy rate was 89.01 percent when compared to the ML algorithm. Another study published in proposed
a medical decision support system (MDSS) for predicting cardiac disease using Weighted Fuzzy Rules (WFR). They employed
two evaluation scenarios: the first automates the WFR generating process, while the second constructs a fuzzy rule-based
MDSS. They used Cleveland's heart disease database to evaluate their MDSS. The best precision value attained by this
technology is 62.35 % when compared to the system based on a neural network.
The authors of used Fast Decision Tree (FDT) and C4.5 tree pruning techniques. This method intends to integrate the findings
of machine learning analysis into several CAD databases. The classification accuracy was found to be 78.06 %, which is
greater than the 75.48 % average classification accuracy of distinct datasets. In 2017, the authors in suggested a Hybrid Neural
Network-Genetic (HNNG) to reinforce the neural network's initial weights using a genetic approach. Using the Z-Alizadeh
Sani data set and Cleveland's heart disease database, the maximum accuracy percentage is 93.85%.
The author [12] used the Alizadeh Sani data set to create a hybrid method that improved the performance of a neural network
using a genetic algorithm and achieved a 93 % accuracy. The author [9] approved a comparative study utilising four different
classifiers, including SVM, KNN, and Neural network. Using 14 different attributes but distinct data sets, he reached a high
accuracy of 93.02 %. Despite a large body of research, there is no gold-standard model for predicting HD. As a result, there is
still scope for improvement. The data set used, the number of attributes and the output class, as well as the algorithm used, all
have an impact on the building of the HD prediction model.
Using a Hungarian data set, this research seeks to develop a medical decision support system that can predict the risk level of
HD. To find trends in existing HD patient data, a classification approach is suggested. The methodology we utilised is outlined
in the next part, along with a brief description of the data set we used. The experiments and various representations of
outcomes are presented in Section 3. Finally, in section 4, you'll find the conclusions.

III. PROPOSED METHOD

A. MACHINE LEARNING

Machine Learning is an efficient technology that is based on two terms: testing and training. The system takes training directly
from data and experience and then applies this training to different types of needs based on the algorithm required.
There are three type of machine learning algorithms:

54
Volume 4 Issue 1
 INTERNATIONAL JOURNAL OF MULTIDISCIPLINARY RESEARCH AND TECHNOLOGY
ISSN 2582-7358
Impact Factor 6.328
Peer-Reviewed Journal

Fig.1 Classification of machine learning

A. Supervised Learning

Supervised learning is described as learning with a proper guide or learning in the presence of a teacher. We get a training
dataset that acts as the trainer for prediction on the data set, so there is always a training dataset when testing a dataset. The
concept of supervised learning is "train me." The following processes are involved in supervised learning:

▪ Classification
▪ Random Forest
▪ Decision tree
▪ Regression

The phenomenon of regression is the recognition of patterns and the measurement of the probability of uninterruptible
outcomes. The system is capable of identifying numbers, their values, and grouping sense of numbers, which means width and
height, among other things. The supervised machine learning algorithms are as follows:

▪ Linear Regression
▪ Logistical Regression
▪ Support Vector Machines (SVM)
▪ Neural Networks
▪ Random Forest
▪ Gradient Boosted Trees
▪ Decision Trees
▪ Naive Bayes

B. Unsupervised Learning

Unsupervised learning is described as learning without supervision, in which no trainer is present to provide assistance. When
a dataset is supplied to Unsupervised learning, it automatically works on the dataset to uncover patterns and relationships
between them, and when fresh data is given, it classifies it and stores it in one of the relationships. Unsupervised learning is
founded on the concept of "self-sufficiency."
For example, assume that you have a mixture of fruits such as mango, banana, and apple, and you use Unsupervised learning
to classify them into three separate clusters based on their relationship to one another, and whenever new data is received, it is
automatically sent to one of the clusters.
55
Volume 4 Issue 1
 INTERNATIONAL JOURNAL OF MULTIDISCIPLINARY RESEARCH AND TECHNOLOGY
ISSN 2582-7358
Impact Factor 6.328
Peer-Reviewed Journal
Supervisor learning says there are three clusters: mango, banana, and apple, whereas Unsupervised learning says there are
three. The procedure for unsupervised algorithms is as follows:
▪ Dimensionality
▪ Clustering

There are following unsupervised machine learning

algorithms:

• t-SNE
• k-means clustering
• PCA

C. Reinforcement

The agent's ability to interact with the environment and determine the outcome is known as reinforced learning. It is based on
the principle of "hit and trial." In reinforced learning, each agent is given positive and negative points, and on the basis of
positive points, reinforced learning produces a dataset output. On the basis of positive rewards, reinforced learning trains and
test datasets. The procedure for reinforced learning is as follows:
▪ Q learning
▪ R learning
▪ TD learning

B. METHODOLOGY OF SYSTEM

The system's processing begins with data collecting, for which we use the UCI repository dataset, which has been thoroughly
confirmed by a number of researchers and the UCI authority.

A. Data Collection

The first stage in developing a prediction system is gathering data and settling on a training and testing dataset. In this
research, we used 70 % of the training dataset and 30 % of the testing dataset to develop the system.

B. Attribute Selection

Attributes of datasets are properties of datasets that are applied for systems, and for the heart, various attributes such as the
person's heart bit rate, gender, age, and many more are shown in TABLE.1 for the prediction system.

TABLE.1 Attributes of the Dataset

56
Volume 4 Issue 1
 INTERNATIONAL JOURNAL OF MULTIDISCIPLINARY RESEARCH AND TECHNOLOGY
ISSN 2582-7358
Impact Factor 6.328
Peer-Reviewed Journal

C. Data pre-processing

Pre-processing is required for machine learning algorithms to produce prestigious results. For example, the Random Forest
technique does not handle datasets with null values, so we must manage null values from the original raw data.
We need to convert some category values to fake values in the form of "0" and "1" for our project.

Fig.2 Data pre-processing

D. Data Balancing
57
Volume 4 Issue 1
 INTERNATIONAL JOURNAL OF MULTIDISCIPLINARY RESEARCH AND TECHNOLOGY
ISSN 2582-7358
Impact Factor 6.328
Peer-Reviewed Journal

Data balancing is necessary for reliable results since we can tell from the data balancing graph that both target classes are
equal. The target classes are depicted in Fig.3, with "0" representing patients with heart disease and "1" representing patients
without heart disease.

Fig.3 Target class view

E. Disease Prediction

SVM, Naive Bayes, Decision Tree, Random Tree, Logistic Regression, Ada-boost, and Gradient Boosting are machine
learning methods used for classification. A comparison of algorithms is conducted for heart disease prediction, and the
algorithm with the best results is chosen, and the algorithm with the highest accuracy is employed.

Fig.4 Disease Prediction

IV. MACHINE LEARNING ALGORITHMS

A. Logistic Regression

Logistic regression is a statistical analysis method to predict a binary outcome, such as yes or no, based on prior observations
of a data set. A logistic regression model predicts a dependent data variable by analyzing the relationship between one or more
existing independent variables.

Fig.5 Logistic Regression

58
Volume 4 Issue 1
 INTERNATIONAL JOURNAL OF MULTIDISCIPLINARY RESEARCH AND TECHNOLOGY
ISSN 2582-7358
Impact Factor 6.328
Peer-Reviewed Journal

B. Decision tree

A decision tree, on the other hand, is a graphical representation of data and a type of supervised machine learning technique.

Fig.6 Decision tree

We use the entropy of the data attributes to build the tree, and we draw the root and other nodes based on the attribute root.
Entropy= -Σ Pij log Pij (1)
The entropy of each node is derived using the following equation of entropy (1), where Pij is the probability of the node. The
root node is chosen based on the highest entropy calculation, and this procedure is repeated until all of the tree's nodes have
been calculated or the tree has been created.
When the number of nodes in a tree is unbalanced, the tree develops an overfitting problem, which is bad for calculations and
one of the reasons why decision trees are less accurate than linear regression.

C. Support Vector Machine

It is a type of machine learning technique that works on the concept of the hyper plan, which means that it classifies data by
creating a hyperplane between them.
The training sample dataset is (Yi, Xi), where i=1,2,3,.....n and Xi is the ith vector and Yi is the target vector. The number of
hyper plans determines the type of support vector; for example, if a line is used as a hyperplane, the method is known as a
linear support vector.

Fig.7 SVM

D. K-Nearest Neighbour

K-Nearest Neighbour Classification is a non-generalizing type of instance-based learning that stores instances of the training
data. It finds a predetermined number of training samples that are closest in distance to predict a new location. Weights can

59
Volume 4 Issue 1
 INTERNATIONAL JOURNAL OF MULTIDISCIPLINARY RESEARCH AND TECHNOLOGY
ISSN 2582-7358
Impact Factor 6.328
Peer-Reviewed Journal
also be assigned to new points when they are classified. Each neighbour in scikit-learn can be given a uniform weight or
weights proportional to the inverse of the distance from the query point or a user defined weight.

E. Random Forest

A supervised learning algorithm is Random Forest. It is an extension of machine learning classifiers that includes bagging to
improve Decision Tree performance. It combines tree predictors, and the trees are dependent on a random vector that is
sampled independently. It can be used both for classification and regression.

F. Extra Tree Classifier

Extremely Randomized Trees (or Extra-Trees) is an ensemble learning method. The method creates extra trees in sub-samples
of datasets and applies majority voting to improve the predictivity of the classifier. By this approach, the method reduces the
variance. The method applies a random threshold for each feature of sub-samples to obtain the best of the thresholds as a
splitting rule.

G. Gradient Boosting

Gradient Boosting is a well-known forward learning ensemble method in machine learning. It is an effective method for
developing predictive models for regression as well as classification tasks.

Fig.8 Gradient Boosting

V. Result Analysis

A. About Jupyter Notebook

Jupyter notebook is used as a simulation tool, and it is user-friendly for Python programming projects. Jupyter notebook
includes rich text features as well as code, such as figures, equations, links, and many more. Because of the combination of
rich text components and code, these documents are ideal for bringing together an analysis description and its findings, as well
as executing data analysis in real-time.
Jupyter Notebook is a web-based interactive visual, maps, charts, visualizations, and narrative text application that is free
source.

B. Accuracy calculation

The algorithms' accuracy is determined by four values: true positive (TP), false positive (FP), true negative (TN), and false
negative (FN) (FN).
Accuracy= (FN+TP) / (TP+FP+TN+FN) (2)
The numerical value of TP, FP, TN, and FN is defined as follows:
TP = Number of individuals with heart disease
TN = Number of individuals with and without heart disease.
FP = Number of individuals who do not have heart disease.
FN denotes the number of individuals who do not have heart disease and those who do have heart disease.

60
Volume 4 Issue 1
 INTERNATIONAL JOURNAL OF MULTIDISCIPLINARY RESEARCH AND TECHNOLOGY
ISSN 2582-7358
Impact Factor 6.328
Peer-Reviewed Journal

Fig.9 Confusion matrix

C. Result

After using the machine learning technique for testing and training, we discovered that the accuracy of the Extra Tree
Classifier is substantially higher than that of other algorithms. Accuracy should be calculated with the help of the confusion
matrix of each algorithm, where the number of counts of TP, TN, FP, and FN is given and using the equation (2) of accuracy,
value has been calculated and it is concluded that Extra Tree Classifier is the best among them with 95 % accuracy.

VI. CONCLUSION AND FUTURE SCOPE

Because the heart is such an important and critical organ in the human body, and the prediction of heart problems is also a
major worry for people, algorithm reliability is one of the parameters used to evaluate the efficiency of the algorithm. The
dataset used for training and testing purposes determines the accuracy of machine learning algorithms. We discover that Extra-
Tree Classifier is the best algorithm when we compare algorithms using a dataset with the properties indicated in TABLE.1
and a confusion matrix.
In the future, more machine learning approaches will be utilized for the best analysis of cardiac illnesses and for early disease
prediction, so that the incidence of mortality cases may be lowered via disease awareness.

REFERENCES

[1] OMojisola Grace Asogbon; Oluwarotimi Williams Samuel; Shixiong Chen; Pang Feng; Guanglin Li"A Hybrid Approach Based on Non-parametric
Attribute Learning Technique and Multi-layer Networks for Congestive Heart Failure Risk Prediction" 2019 IEEE 5th International Conference on
Computer and Communications (ICCC),DOI: 10.1109/ICCC47050.2019.9064070
[2] Oumaima Terrada; Bouchaib Cherradi; Soufiane Hamida; Abdelhadi Raihani; Hicham Moujahid; Omar Bouattane"Prediction of Patients with Heart
Disease using Artificial Neural Network and Adaptive Boosting techniques" 2020 3rd International Conference on Advanced Communication
Technologies and Networking (CommNet),DOI: 10.1109/CommNet49926.2020.9199620
[3] Senthilkumar Mohan; Chandrasegar Thirumalai; Gautam Srivastava"Effective Heart Disease Prediction Using Hybrid Machine Learning Techniques"
DOI: 10.1109/ACCESS.2019.2923707
[4] AniruddhaDutta, TamalBatabyal, MeheliBasu, Scott T.Acton"An efficient convolutional neural network for coronary heart disease prediction" Expert
Systems with Applications Volume 159, 30 November 2020, 113408 DOI: 10.1016/j.eswa.2020.113408
[5] E. Nasarian, M. Abdar, M.A. Fahami, R. Alizadehsani, S. Hussain, M.E. Basiri, M. Zomorodi-Moghadam, X. Zhou, P. Pławiak, U.R. Acharya, R.-S.
Tan, N. Sarrafzadegan, “Association between work-related features and coronary artery disease: A heterogeneous hybrid feature selection integrated with
balancing approach,” Pattern Recognition Letters, 133, 33–40, 2020, doi:10.1016/j.patrec.2020.02.010.
[6] M. Akhil, B. L. Deekshatulu, and P. Chandra, “Classification of Heart Disease Using K- Nearest Neighbor and Genetic Algorithm,” Procedia Technol.,
vol. 10, pp. 85–94, 2013.

61
Volume 4 Issue 1
 INTERNATIONAL JOURNAL OF MULTIDISCIPLINARY RESEARCH AND TECHNOLOGY
ISSN 2582-7358
Impact Factor 6.328
Peer-Reviewed Journal
[7] Z. Arabasadi, R. Alizadehsani, M. Roshanzamir, H. Moosaei, A.A. Yarifard, “Computer aided decision making for heart disease detection using hybrid
neural network-Genetic algorithm,” Computer Methods and Programs in Biomedicine, 141, 19–26, 2017, doi:10.1016/j.cmpb.2017.01.004.
[8] Pahulpreet Singh Kohli and Shriya Arora, “Application of Machine Learning in Diseases Prediction”, 4th International Conference on Computing
Communication And Automation(ICCCA), 2018.
[9] F Ali, S El-Sappagh, SMR Islam, D Kwak, A Ali, Muhammad Imran"A smart healthcare monitoring system for heart disease prediction based on
ensemble deep learning and feature fusion" Information Fusion Volume 63, November 2020, Pages 208-222,DOI: 10.1016/j.inffus.2020.06.008
[10] Safial Islam Ayon, Md. Milon Islam, Md. Rahat Hossain"Coronary Artery Heart Disease Prediction: A Comparative Study of Computational
Intelligence Techniques" DOI: 10.1080/03772063.2020.1713916
[11] K. H., Miao, J. H. Miao & G. Miao. “Diagnosing Coronary heart disease Using Ensemble Machine Learning,” International Journal of Advanced
Computer Science and Applications, vol 7(10), 2016.
[12] Santhana Krishnan J and Geetha S, “Prediction of Heart Disease using Machine Learning Algorithms” ICIICT, 2019.
[13] Aditi Gavhane, Gouthami Kokkula, Isha Panday, Prof. Kailash Devadkar, “Prediction of Heart Disease using Machine Learning”, Proceedings of the 2nd
International conference on Electronics, Communication and Aerospace Technology(ICECA), 2018.
[14] Senthil kumar mohan, chandrasegar thirumalai and Gautam Srivastva, “Effective Heart Disease Prediction Using Hybrid Machine Learning Techniques”
IEEE Access 2019.
[15] Himanshu Sharma and M A Rizvi, “Prediction of Heart Disease using Machine Learning Algorithms: A Survey” International Journal on Recent and
Innovation Trends in Computing and Communication Volume: 5 Issue: 8 , IJRITCC August 2017.
[16] L. M. Hung, D. T. Toan, & V. T. Lang. “Automatic Heart Disease Prediction Using Feature Selection and Data Mining Technique,”. Journal of
Computer Science and Cybernetics,vol 34(1), pp. 33-47, 2018.

62
Volume 4 Issue 1