See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/379309484

A Study on Heart Disease Prediction Using Machine Learning Algorithms

Article · March 2024

CITATIONS

1 author:

Shambhu Rai
Bharati Vidyapeeth University
14 PUBLICATIONS 28 CITATIONS

SEE PROFILE

All content following this page was uploaded by Shambhu Rai on 27 March 2024.

The user has requested enhancement of the downloaded file.

NATURALISTA CAMPANO
ISSN: 1827-7160
Volume 28 Issue 1, 2024

A Study on Heart Disease Prediction Using

Machine Learning Algorithms
Dr. Shambhu Shankar Rai1, Dr. Uma Durgude2, Dr. Avinash Dhavan3, Payal Singh4, Dr. Sadhana Ojha5
1
Assistant Professor, Bharati Vidyapeeth’s Institute of Management and Information Technology, Navi
Mumbai, India
2,3,4
Assistant Professor, Bharati Vidyapeeths Institute of Management Studies and Research, Navi Mumbai,
India.
5
Coordinator at St. Wilfred's College of Computer Sciences. Navi Mumbai, India.

Email: [email protected], [email protected], [email protected],

[email protected], [email protected]

Abstract: The incidence of heart disease is rapidly increasing, underscoring the critical need to proactively
identify potential illnesses. This diagnostic task is intricate, demanding precision and efficiency. The central focus
of the research paper lies in discerning, based on diverse medical characteristics, which patients are more
predisposed to heart disease. By leveraging the patient's medical history, we devised a method to assess the
likelihood of a heart disease diagnosis. Employing a range of machine learning algorithms such as KNN and
logistic regression, we aimed to predict and classify patients at risk of heart disease. The study also explores how
the model's application can enhance the accuracy of predicting heart attacks.

Keywords: Machine Learning, Heart Disease, Prediction, Detection, Naïve Bayes.

1. Introduction

Machine learning serves as a powerful technique for extracting implicit, unknown, or known information from
data, offering broad and diverse applications with expanding scope. The application of machine learning spans
various classifiers, including those from supervised, unsupervised, and ensemble learning, contributing to the
forecasting and accuracy determination of given datasets. The knowledge derived from machine learning can be
effectively applied to projects like the Heart Disease Prediction System (HDPS), benefiting a wide range of
individuals.
Cardiovascular diseases, encompassing a variety of disorders that pose potential threats to the heart, are a prevalent
concern. According to the World Health Organization, there are 17.9 million deaths globally attributed to
cardiovascular diseases, making it a primary contributor to adult mortality. Leveraging a person's medical history,
the HDPS initiative aims to identify individuals most likely to be diagnosed with a cardiac condition. This
proactive approach allows for the identification of patients exhibiting symptoms like chest pain or high blood
pressure, facilitating timely and efficient diagnosis with fewer medical procedures. The ultimate goal is to
streamline therapies, ensuring appropriate treatment for the patient. The study focuses on three data mining
approaches, particularly emphasizing Logistic Regression and KNN, to enhance the understanding and predictive
capabilities of the HDPS initiative.
The prognostication of an individual's likelihood of having heart disease has been significantly enhanced through
a notably effective approach. In comparison to previously employed classifiers like naive Bayes, the suggested
model demonstrates a commendable accuracy in predicting indications of heart disease in specific individuals.
This achievement is attributed to the utilization of KNN and Logistic Regression. Consequently, relying on the
proposed model to assess the probability of the classifier accurately and reliably identifying heart disease has
alleviated considerable concerns. The system for predicting heart disease, as presented, not only enhances patient
treatment but also proves cost-effective. This research yields valuable insights that can be harnessed for predicting
individuals prone to developing heart disease, utilizing the .pynb file type.
Project’s predictive accuracy to 87.5%. This surpasses the accuracy of the previous system, which solely relied
on a single data mining technique. The inclusion of the Random Forest Classifier, along with other methods,
signifies a notable improvement in predictive performance. This highlights the efficacy of employing a diverse
set of data mining techniques, underscoring the potential for enhanced outcomes in predictive modeling.
Precision and efficacy in Heart Disease Prediction System (HDPS) are crucial considerations in supervised
learning, which encompasses logistic regression, a technique specifically tailored for discrete values. The primary

https://museonaturalistico.it 915
NATURALISTA CAMPANO
ISSN: 1827-7160
Volume 28 Issue 1, 2024

objective of this project is to ascertain the likelihood of a patient being diagnosed with cardiovascular heart
diseases based on various medical characteristics, including gender, age, chest discomfort, fasting blood sugar
level, among others. The dataset, sourced from the UCI repository, comprises the patient's medical history and
characteristics, serving as the basis for predicting the potential for heart disease.
In this predictive modeling endeavor, we leverage three algorithms—KNN, Random Forest Classifier, and
Logistic Regression—to train on the 14 medical characteristics. Notably, KNN emerges as the most effective
algorithm, achieving an accuracy rate of 88.52%. The final step involves categorizing individuals based on their
risk of developing a cardiac condition, demonstrating a cost-effective and efficient procedure. This comprehensive
approach aims to enhance the precision and cost-effectiveness of anticipating heart conditions in individuals
through the utilization of diverse algorithms and medical characteristics.

Literature Review
Using the UCI Machine Learning dataset, extensive research has been done to predict cardiac disease. Varied data
mining approaches have been used to achieve varied accuracy levels, which are detailed below.
Avinash Golande and colleagues investigate various ML algorithms that can be used to categorise cardiac disease.
An investigation was conducted to examine the accuracy of the classification algorithms Decision Tree, KNN,
and K-Means. The study found that Decision Trees had the highest accuracy, and it was concluded that by
combining various methodologies and fine-tuning its parameters, it might be made more effective.
A system that combined the MapReduce algorithm with data mining techniques has been suggested by T.
Nagamani et al. For the 45 instances in the testing set, the accuracy obtained according to this article was higher
than the accuracy obtained using a traditional fuzzy artificial neural network. Here, the usage of dynamic schema
and linear scaling increased the algorithm's accuracy.
AI ML model created by Fahd Saleh Alotaibi compares five alternative methods. When compared to Matlab and
Weka, the Rapid Miner tool performed more accurately. This study compared the classification accuracy of
Decision Tree, Logistic Regression, Random Forest, Naive Bayes, and SVM algorithms. The most accurate
algorithm was the decision tree algorithm.
A system that employs NB (Naive Bayesian) approaches for dataset categorization and the AES (Advanced
Encryption Standard) algorithm for safe data transport was proposed by Anjan Nikhil Repaka, et al.
A survey was conducted by Theresa Princy, R., et al., using various classification algorithms for heart disease
prediction. The classifiers' accuracy was examined for a variety of variables using Naive Bayes, KNN (K-Nearest
Neighbor), Decision Trees, and Neural Networks as the classification methodologies.
Heart disease was predicted by Nagaraj M. Lutimath et al. using Naive Bayes classification and SVM. (Support
Vector Machine). Mean Absolute Error, Sum of Squared Error, and Root Mean Squared Error are the performance
measurements utilized in analysis. It has been determined that SVM outperformed Naive Bayes in terms of
accuracy.
After reading the aforementioned publications, the fundamental idea behind the suggested system was to build a
heart disease prediction system based on the inputs presented in Table 1. By comparing the accuracy, precision,
recall, and f-measure scores of the classification algorithms Decision Tree, Random Forest, Logistic Regression,
and Naive Bayes, we were able to determine which classification algorithm would be most effective at predicting
heart disease.
Shah et al.'s study from 2020 [18] sought to create a model for predicting cardiovascular illness using machine
learning methods. The 303 cases and 17 attributes of the Cleveland heart disease dataset, which was sourced from
the UCI machine learning repository, were utilised to generate the data for this project. The authors used a range
of supervised classification techniques, including k-nearest neighbor, naive Bayes, decision trees, and random
forests. (KKN). The study's findings showed that, at 90.8% accuracy, the KKN model had the best level of
precision.
The study emphasizes the potential value of machine learning methods in anticipating cardiovascular illness and
the the significance of choosing the right models and methods to get the best results.
In a study by Drod et al. (2022), the goal was to identify the most important risk factors for cardiovascular disease
(CVD) in patients with metabolic-associated fatty liver disease using machine learning (ML) approaches.
(MAFLD). 191 MAFLD patients had their blood biochemically analyzed, and subclinical atherosclerosis was
evaluated. Using ML techniques, such as multiple logistic regression classifier, univariate feature ranking, and
principal component analysis, a model to identify those with the highest risk of CVD was created. (PCA). The
most important clinical traits, according to the study, were hypercholesterolemia, plaque scores, and length of
diabetes. With an AUC of 0.87, the ML method worked well, correctly classifying 114/144 (79.17%) low-risk
patients and 40/47 (85.11%) high-risk patients. The results of the study show that using straightforward patient
criteria, an ML technique is beneficial for identifying MAFLD patients with extensive CVD.
The author of a study by Alotalibi (2019) [19] set out to look into the effectiveness of machine learning (ML)

https://museonaturalistico.it 916
NATURALISTA CAMPANO
ISSN: 1827-7160
Volume 28 Issue 1, 2024

approaches for diagnosing heart failure condition. The study made use of using a dataset from the Cleveland Clinic
Foundation, we developed prediction models using a variety of ML algorithms, including decision trees, logistic
regression, random forests, naive bayes, and support vector machines (SVM). During the model development
process, a 10-fold cross-validation strategy was used. The findings showed that the decision tree algorithm, which
had a rate of 93.19%, and the SVM method, which had a rate of 92.30%, had the highest accuracy in predicting
heart disease. This work highlights the decision tree algorithm as a potential useful tool for forecasting heart
failure disease and sheds light on the possibilities of ML approaches as such.

2. Methodology

This study seeks to estimate the likelihood of developing heart disease using computerised heart disease prediction,
which may be useful for patients and medical professionals. We used a dataset and many machine learning
methods to accomplish this goal, and the findings are presented in this study report. We intend to sanitise the data,
get rid of extraneous details, and add new characteristics like MAP and BMI to improve the technique. The dataset
will then be divided depending on gender, and k-modes clustering will be used. Finally, we will use the cleaned
data to train the model. As shown in Figure, the revised process will result in more accurate results and greater
model performance.

2.1 Data Collection and Pre-Processing

This study seeks to estimate the likelihood of developing heart disease using computerized heart disease
prediction, which may be useful for patients and medical professionals.
We used a dataset and many machine learning methods to accomplish this goal, and the findings are presented in
this study report. We intend to sanitize the data, get rid of extraneous details, and add new characteristics like
MAP and BMI to improve the technique. The dataset will then be divided depending on gender, and k-modes
clustering will be used.
Finally, we will use the cleaned data to train the model. As shown in Figure, the revised process will result in
more accurate results and greater model performance.

2.2 Data Collection

The process of gathering, measuring, and analysing precise insights for study is known as data collection. A
researcher can assess their hypothesis using the data that they have gathered. Regardless of the field of study, data
gathering is typically the first and most crucial phase in the research process. A structured data set of Algerians
who have completed analyses at the Mohand Amokrane EHS Hospital ex CNMS in Algiers, Algeria, is used in
this study. It has 1200 rows and 20 columns, and the variables age, sex, cp, trestbps, chol, Ex-Ang, Col-Ves, fbs,
restecg, thalach, exang, oldpeak, slope, RBP, ca, thal, smoking, alcohol use.

Fig: Datasets for the study

2.3 Manual Investigation

Data exploration, also known as manual exploration, is the first stage of data analysis, during which users examine

https://museonaturalistico.it 917
NATURALISTA CAMPANO
ISSN: 1827-7160
Volume 28 Issue 1, 2024

a sizable data collection informally to find the first patterns, traits, and areas of interest. This approach is intended
to help establish a broad picture of significant trends and key areas to investigate in more detail rather than to show
every piece of information that a data set contains. To begin the pre- processing for our study, we add a column
to our data collection called Target that has the values 0 or 1 (0 = not sick, 1 = sick). Figure explains this

Fig: Manual Investigation

2.4 Data Preprocessing

We prepare the data to be implemented before beginning the application of machine learning algorithms. This
phase is accomplished in two steps:
Features choice the correlation matrix is the basis for this stage. Initially, we possessed 20 of the previously
mentioned qualities. We identified 13 attributes (age, sex, cp, trestbps, chol, fbs, restecg, thalach, exang, oldpeak,
slope, ca, thal) that are connected to and reliant on one another after using the Pearson correlation matrix. Table
explains the specifics of the chosen characteristics.

Table: The details of selected Features

https://museonaturalistico.it 918
NATURALISTA CAMPANO
ISSN: 1827-7160
Volume 28 Issue 1, 2024

3. System Implementation

3.1 Split Data Set

Dividing a data set our data set is divided into two sections: The test portion is 20% larger than the training data
set in the first half. Fig. shows the division of our data set.

3.2 Testing Algorithm

We verify the accuracy of each algorithm on the various data sets after running the three algorithms on the four
data sets (600, 800, 1000, and 1200 lines). Table 3 shows how we determined the accuracy for the three methods
based on the confusion matrix (we used the latest data set, which had 1200 lines).
Now that we have the 4 data sets, we can show the accuracy of each algorithm, In terms of accuracy and stability
against changes in the data sets, Fig. compares the three algorithms

Neural Network Dataset (1200 lines) SVM Dataset (1200 lines) KNN Dataset (1200 lines)

Sick Not Sick Sick Not Sick Sick Not Sick

Sick 94 8 90 11 84 18

Not Sick 6 92 9 90 11 87

Accuracy 93% 90% 85.50%

4. Results

4.1 Shows the Risk of Heart Attack on the basis of their Age
From these findings, it is clear that even if the majority of studies use other algorithms, such as SVC and Decision
trees, to identify patients with heart disease, KNN, Random Forest Classifier, and Logistic Regression produce a
superior outcome to them. Our algorithms are faster and more precise than those employed by earlier studies. They
also save a significant amount of money, making them very cost- effective. Furthermore, the combined maximum
accuracy of KNN and Logistic Regression is 88.5%, which is higher than or nearly equal to the accuracy of earlier
studies. So, to sum up that our accuracy is improved due to the increased medical attributes that we used from the
dataset we took. Our project also tells us that Logistic Regression and KNN outperforms Random Forest Classifier
in the prediction of the patient diagnosed with a heart Disease. This proves that KNN and Logistic Regression are
better in diagnosis of a heart disease. The following ‘figures’ shows a plot of the number of patients that are been
segregated and predicted by the classifier depending upon the age group, Resting Blood Pressure, Sex, Chest Pain:

https://museonaturalistico.it 919
 NATURALISTA CAMPANO
ISSN: 1827-7160
Volume 28 Issue 1, 2024

5. CONCLUSION

Heart disease has increased in prevalence throughout the world, including in our country. Consequently,
diagnosing the illness before contracting it reduces the danger of dying. There has been extensive research done
in this prediction field.
Our study is a component of the investigation into the identification and prognosis of cardiac disease.
It is based on the use of machine learning algorithms, of which we have selected the three most popular ones (Neural
Network, SVM, and KNN), on a real data set of Algerian people, with excellent results; we reached 93% accuracy
with Neural Network. The key finding of our study was that, after testing the algorithm's stability on a variety of
data sets of varying sizes, it was clear that neural networks produced the greatest results. Additionally, we
conducted research on feature selection and employed a correlation matrix to identify attribute dependencies. This
strategy can be improved in a number of ways, including by using deep learning algorithms, other attribute
selection techniques, and even bigger data sets.

6. REFERENCES

1. Golande, A., Kumar, P., "Prediction of Heart Disease Utilizing Effective Machine Learning Techniques,"
International Journal of Recent Technology and Engineering, Vol. 8, pp. 944-950, 2019.
2. Nagamani, T., Logeswari, S., Gomathy, B., "Heart Disease Prediction through Data Mining with
Mapreduce Algorithm," International Journal of Innovative Technology and Exploring Engineering
(IJITEE) ISSN: 2278-3075, Volume-8 Issue-3, January 2019.
3. Alotaibi, F. S., "Implementing a Machine Learning Model for Heart Failure Disease Prediction,"
International Journal of Advanced Computer Science and Applications (IJACSA), Vol. 10, No. 6, 2019.
4. Repaka, A. N., Ravikanti, S. D., Franklin, R. G., "Design and Implementation of Heart Disease Prediction
Using Naive Bayesian," International Conference on Trends in Electronics and Information (ICOEI 2019).
5. Princy, T., Thomas, J., "Human Heart Disease Prediction System using Data Mining Techniques,"
International Conference on Circuit Power and Computing Technologies, Bangalore, 2016.
6. Lutimath, N. M., Chethan, C., Pol, B. S., "Prediction of Heart Disease using Machine Learning,"
International Journal of Recent Technology and Engineering, 8(2S10), pp. 474-477, 2019.
7. Ambekar, S., Phalnikar, R., "Disease Risk Prediction using Convolutional Neural Network," 2018 Fourth
International Conference on Computing Communication Control and Automation.
8. Rjeily, C. B., Badr, G., Hassani, E., A. H., Andres, E., "Medical Data Mining for Heart Diseases and the
Future of Sequential Mining in Medical Field," in Machine Learning Paradigms, 2019, pp. 71–99.
9. Alzubi, J., Nayyar, A., Kumar, A., "Machine Learning from Theory to Algorithms: An Overview," Journal
of Physics: Conference Series, 2018.
10. Alarsan, F. I., Younes, M., "Analysis and Classification of Heart Diseases using Heartbeat Features and
Machine Learning Algorithms," Journal of Big Data, 2019; 6:81.

https://museonaturalistico.it 920

View publication stats