Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR444

Heart Failure Prediction using

Machine Learning Algorithms
R. Renugadevi1 Nivethitha. A2
Assistant Professor (Sr. G), Student
M.E. Department of CSE, M.E. Computer Science and Engineering,
KIT – Kalaignar Karunanidhi KIT – Kalaignar Karunanidhi
Institute of Technology (Autonomous), Institute of Technology (Autonomous),
Coimbatore, TN, India Coimbatore, TN, India

Abstract:- This day and age individuals are increasingly I. INTRODUCTION

giving precedence to their material needs as opposed to
self-care, leading to physical and mental strain. Machine learning is a powerful method that's used for
Cardiovascular diseases (CVDs) present a significant analyzing vast datasets, digging up patterns, and grasping
menace worldwide, causing about 17.9 million deaths relationships within the data. This involves the collection,
annually which is roughly 32% of global mortality. cleaning, pre-processing, and analysis of data to extract
Heart failure, which impacts over 550,000 individuals on insights that are valuable. Situated within the domain of
a yearly basis, emerges as an urgent global health Artificial Intelligence (AI), machine learning operates on the
concern. The formulation of effective prediction principle that machines can autonomously learn and
techniques for heart failure proves to be imperative in improve from data without explicit programming. One of
lessening its repercussions. Linear and machine learning the strengths of machine learning is its capability to decipher
models are put into service to forecast heart failure patterns within datasets that are immense, which enables
utilizing a myriad of inputs, comprising clinical data. machines to learn from experiences and make predictions.
With the burgeoning population, the early detection and Detecting coronary heart diseases is becoming increasingly
intervention for heart disease grow more complex. Heart challenging in emerging countries due to the lack of
disease prevalence has escalated to concerning levels, specialized medical professionals and outdated examination
culminating in untimely deaths due to arterial plaque tools, which make the process slow, and complex, this is a
accumulation. The premature pinpointing of heart significant issue worldwide. Several factors contribute to
disease holds the potential to rescue many lives by coronary heart disease, such as smoking, high blood
upholding arterial wellness. Our research integrates pressure, high cholesterol, and diabetes. Smoking, in
supervised machine learning algorithms to predict heart particular, has become exceedingly prevalent among both
disease presence, underscoring methods to enhance young as well as old individuals and is currently trending
classifier efficacy. Null values within the dataset are among youth. It constricts the coronary arteries, leading to
managed through mean value imputation, whereas irregular heartbeats and an increase in blood pressure. High
irrelevant attributes are expunged utilizing information- blood pressure poses numerous risks, as it compels the heart
gain feature selection. By wielding breakthroughs in to pump harder to circulate blood throughout the body,
machine learning (ML), the key aim of this study is to leading to thickening of the lower-left heart chamber;
design prognostic models for cardiovascular disease consequently, increasing the risk of heart failure. Elevated
utilizing 12 clinical attributes. By capitalizing on a blood sugar levels can also potentially cause damage to
dataset offered by Davide Chicco and Giuseppe Jurman, nerves or blood vessels that regulate heart function.
encompassing 12 clinical features and 299 data points, However, we must be aware of the importance of
the efficacy of three ML algorithms: Support Vector maintaining a healthy heart through proper diet and exercise.
Machine (SVM), Random Forest, and Logistic This capability is notably useful for handling complex tasks
Regression is evaluated. Our examination discloses that beyond human capacity, resulting in significant time and
Logistic Regression showcases the most outstanding cost savings. Its applications range across various domains,
accuracy and likelihood in foretelling cardio vascular including self-driving cars, cyber fraud detection, facial
disease presence. This predictive model exhibits potential recognition, and personalized recommendations on
in aiding healthcare experts in curtailing heart disease- platforms like Facebook. Notably, prominent companies
linked fatalities. such as Netflix and Amazon utilize machine learning
models for analyzing user behavior and customizing product
Keywords:- Random Forest, Support Vector Machine, recommendations. Machine learning algorithms are trained
Logistic Regression, Machine Learning Model, Heart using datasets to build models, which in turn make
Failure Prediction, Disease Prediction, Accuracy. predictions when new data is presented. These predictions
are evaluated for accuracy, and, if found satisfactory, the
machine learning algorithm is deployed. In scenarios where
accuracy is lacking, the algorithm undergoes iterative

IJISRT24MAR444 www.ijisrt.com 372

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR444

training with augmented datasets to boost performance. They employed various machine learning algorithms to
Machine learning, being a subset of artificial intelligence, detect and predict human heart disease, utilizing a heart
focuses on developing and studying statistical algorithms disease dataset. Performance evaluation of these algorithms
that are capable of executing tasks without explicit was conducted using metrics such as classification accuracy;
instructions. Key algorithms like Support Vector Machine F-measure, sensitivity, and specificity!!! They wanted to
(SVM), Random Forest, and Logistic Regression play ascertain the effectiveness of these approaches in predicting
essential roles in constructing machine learning models. heart failure accurately. Furthermore, the team explored
other possibilities and experimented with different models to
II. RELATED WORK enhance the predictive capabilities of the machine learning
systems being utilized. Despite some challenges faced
Numerous researchers have utilized machine learning during the research, the team remained resilient and focused
techniques, including Support Vector Machine (SVM), to on achieving their goals. In summary, their research sheds
devise strategies for predicting heart disease. For example, light on the potential benefits of using machine learning in
in their study titled "Heart Failure Prediction Using Machine predicting heart failure, which can ultimately aid in early
Learning Algorithm," presented at the 3rd International detection and proactive intervention strategies for
Conference on Computation, Automation and Knowledge individuals at risk.
Management (ICCAKM) in 2022, Pandey and Kaur
developed a system aimed at reducing heart diseases by The study conducted by Montu Saw, Tarun Saxena,
addressing the challenge of similar symptoms, resulting in Sanjana Kaithwas, Rahul Yadav, and Nidhi Lal, titled
an impressive accuracy of 94.56% in diagnosis. During the "Estimation of Prediction for Getting Heart Disease Using
research process, Pandey and Kaur employed a Support Logistic Regression Model of Machine Learning," released
Vector Machine (SVM) to analyze the data collected from in January 2020, reported achieving an 87% accuracy using
patients. The machine learning algorithm they used had the logistic regression technique. Their research highlighted
proven to be effective in identifying patterns and trends, that men are more prone to cardiovascular disease compared
hence why it was chosen for this study. The findings of their to women. Additionally, factors such as aging, daily
study show that with proper utilization of machine learning cigarette consumption, and systolic blood pressure were
algorithms, healthcare professionals can potentially enhance identified as influencing heart disease risk. Interestingly, the
their ability to diagnose heart diseases accurately and study found that total cholesterol alone did not significantly
efficiently. Additionally, the development of such systems alter the likelihood of coronary heart disease (CHD),
can lead to early detection and intervention in individuals at suggesting that the level of HDL within the total cholesterol
risk, ultimately improving patient outcomes. Although value might be a contributing factor. Furthermore, the
machine learning algorithms can be greatly beneficial, it is effects of glucose on CHD risk were deemed
important to note that they are not without limitations. insignificant.The researchers suggested that further data
Researchers must continue to refine these techniques and collection and the utilization of additional machine learning
algorithms to ensure their accuracy and reliability. In models could potentially improve the predictive
conclusion, the utilization of machine learning techniques, performance of the model. This underscores the importance
such as the Support Vector Machine (SVN), presents of continuous refinement and validation of predictive
promising opportunities for the field of healthcare, models in healthcare research, particularly for complex and
particularly in the prediction and diagnosis of heart diseases. multifactorial conditions such as heart disease.
Further research and development in this area will
undoubtedly contribute to improved patient care and The study "Heart Disease Prediction Using Random
outcomes. Forest Algorithm," conducted by Kompella Sri Charan and
Kolluru S S N S Mahendranath, was released in March
Boukhatem, Youssef, and Nassif proudly presented 2022. In this study, the researchers evaluated the accuracy
their latest paper "Heart Disease Prediction Using Machine scores of several machine learning algorithms, including
Learning" at the 2022 Advances in Science and Engineering Decision Tree, Random Forest, Support Vector Machine
Technology International Conferences (ASET) in Dubai, (SVM), AdaBoost, and Gradient Boosting, for identifying
UAE!!! They showcased the implementation of four heart diseases. The study found that the Random Forest
classification techniques like Multilayer Perceptron (MLP), algorithm outperformed other algorithms, achieving an
Support Vector Machine (SVM), Random Forest (RF), and impressive accuracy rating of 92.16% in forecasting heart
Naïve Bayes (NB) to build predictive models for heart disease. This suggests that Random Forest is the most
problems. Ahead of pattern creation, they completed data effective machine learning method for the identification of
preprocessing and carefully selected features. Surprisingly, heart illnesses among those evaluated. Despite the
the SVM model stood out as the frontrunner, achieving an promising results, the study indicates that there is still room
excellent accuracy rate of 91.67. for improvement and further potential for enhancing the
predictive model. This highlights the importance of ongoing
Abbas, Imran, Al-Aloosy, Fahim, Alzahrani, and research and development efforts in the field of machine
Muzaffar presented their research titled "Heart Failure learning for healthcare applications, particularly in
Prediction Using Machine Learning Approaches" at the improving the accuracy and reliability of predictive models
2022 Mohammad Ali Jinnah University International for diagnosing and managing heart diseases.
Conference on Computing (MAJICC) in Karachi, Pakistan.

IJISRT24MAR444 www.ijisrt.com 373

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR444

 Existing System  Steps in Proposed System

World Health Organization (WHO) has recently
released some astonishing statistics concerning the global  Data Collection:
burden of heart diseases. Shockingly, an estimated 12 Starting with obtaining the dataset from Kaggle, you
million deaths worldwide are yearly associated with heart have patient details which is crucial for your project.
diseases! Unbelievable, right? Heart disease supposedly
represents around 25% of deaths among individuals aged 25  Attribute Selection (Feature Selection):
to 69 years - a truly startling figure. In urban areas, this This step involves identifying and selecting the most
percentage seems to jump to a whopping 32.8%, while in relevant features from your dataset that contribute the most
rural areas, the numbers remain high at 22.9%. Furthermore, to your model's predictive performance. It helps in reducing
it's mind-boggling to know that over 80% of deaths dimensionality and improving model interpretability.
worldwide are said to be due to heart disease WHO seems to
believe that by 2030, the number of deaths because of heart  Data Pre-Processing:
disease could potentially reach nearly 23.6 million Following attribute selection, you'll move on to data
individuals. These statistics clearly shine a light on the pre-processing. This step involves cleaning the data by
pressing need for effective strategies to tackle and alleviate handling missing, null, or corrupted values, and removing
the impact that heart diseases have on global health. unnecessary features that were not selected in the attribute
selection step.
 Proposed System
This study utilized machine learning methodologies to  Model Development:
predict occurences of heart disease, showcasing the Once the data is pre-processed, you'll proceed to
extensive potential of machine learning in healthcare sectors develop machine learning models. You've selected Support
for prediction and categorization tasks. Notably, within Vector Machine, Random Forest, and Logistic Regression
healthcare, data mining precision has notably improved with due to their known accuracy and efficiency. This step
the advent of machine learning techniques.The heart disease involves training these models on your pre-processed data.
prediction model leverages machine learning, a field poised
to revolutionize healthcare. Machine learning has  Model Evaluation and Selection:
demonstrated significant promise in improving prediction After training the models, you'll evaluate their
and classification accuracy, particularly in healthcare performance using suitable evaluation metrics (such as
applications. Among the diverse algorithms explored, accuracy, precision, recall, F1-score, etc.) on a separate test
logistic regression stands out for its high accuracy rate in dataset. Then, you'll select the model with the highest
predicting heart disease.The model employs a variety of accuracy among the three algorithms for deployment.
techniques and algorithms, including logistic regression,
support vector machine (SVM), and random forest, to III. MODULE DESCRIPTION
determine whether an individual is affected by heart disease.
Through the integration of these algorithms and tools, the  Collection of Clinical Data
system aims to accurately classify individuals based on the The dataset used in this project originates from the
presence or absence of heart disease. research conducted by Davide Chicco and Giuseppe Jurman,
titled "Machine learning can predict survival of patients with
heart failure from serum creatinine and ejection fraction
alone," published in BMC Medical Informatics and Decision
Making in 2020. This dataset comprises 13 features and 299
rows, specifically designed for predicting mortality due to
heart failure. The attributes within the dataset are numeric,
with one of the key features being the age of the patient.
Age holds significant importance as it is considered the most
critical risk factor in the development of heart diseases.
Notably, the risk of heart disease doubles during
adolescence, highlighting the relevance of age in predicting
heart failure.

Fig 1 Architecture Diagram

IJISRT24MAR444 www.ijisrt.com 374

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR444

Fig 2 Overview of Dataset

 Data – Pre-processing
Data pre-processing is indeed a crucial step in data
mining, as it involves preparing and transforming data into a
suitable format for further analysis. This process serves to
enhance the quality and effectiveness of data mining
procedures. Data pre-processing encompasses several key
tasks:

 Data Cleaning:
This involves identifying and rectifying errors or
inconsistencies in the data, such as missing values, duplicate Fig 3 The Correlation between Different Columns
records, or inaccuracies.
 Feature Extraction:
 Data Integration: Feature extraction involves selecting or creating a
In this step, data from multiple sources or formats are subset of relevant features from the dataset. This helps
combined into a unified dataset. This ensures that all reduce the dimensionality of the data and focuses on the
relevant data is available for analysis. most important aspects for analysis.
 Data Reduction:  Feature Engineering:
Data reduction techniques are applied to reduce the Feature engineering involves transforming or creating
size of the dataset while preserving its informational new features based on existing ones to improve the
content. This may involve techniques such as sampling, performance of machine learning algorithms.
aggregation, or dimensionality reduction.
Overall, data pre-processing plays a critical role in
 Normalization: ensuring the quality, reliability, and effectiveness of data
Normalization is performed to scale the numerical mining procedures, ultimately leading to more accurate and
features of the dataset to a standard range, typically between meaningful insights from the data.
0 and 1. This ensures that all features contribute equally to
the analysis and prevents biases due to differences in scale. IV. METHODOLOGIES

 Outlier Detection and Removal:  Support Vector Machine

Outliers, or data points that deviate significantly from A Support Vector Machine (SVM) is a supervised
the rest of the dataset, can skew the results of data analysis. learning algorithm utilized in machine learning to tackle
Hence, outlier detection techniques are used to identify and intricate classification, regression, and outlier detection
remove such outliers from the dataset. tasks. Indeed, SVM is a prominent method within
supervised learning, extensively applied in statistical
classification and regression analysis. It falls under the
category of generalized linear classifiers, distinguished by
their capacity to simultaneously minimize empirical error
and maximize geometric edge regions. Hence, SVM is often
referred to as a maximum edge region classifier.

IJISRT24MAR444 www.ijisrt.com 375

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR444

 Logistic Regression
Logistic regression is a classification model frequently
employed in binary scenarios. It serves as one of the
fundamental algorithms in machine learning for addressing
binary (0 or 1) problems, enabling the estimation of the
likelihood of certain outcomes. Linear regression is utilized
for solving regression problems, where the goal is to predict
a continuous numerical outcome based on one or more input
variables.

Fig 4 Confusion Matrix for SVM

 Random Forest
The Random Forest classifier is composed of
numerous decision trees, and its output class is determined
by the mode of the output classes of the individual trees. In
mathematics, a random walk, also referred to as a drunkard's
walk, is a stochastic process that depicts a trajectory
comprising a series of random steps within a mathematical
space. The process typically begins at a starting point, often
denoted as 0, and at each subsequent step, it randomly Fig 6 Confusion Matrix for Logistic Regression
moves either +1 or −1 with equal probability.During the
training process, each decision tree in the Random Forest V. CONFUSION MATRIX
independently makes predictions, and then the final
prediction is determined by aggregating the predictions of The confusion matrix offers a detailed insight into
all the trees. In classification tasks, the most commonly model performance. Despite its name, you'll discover that
occurring class among the predictions of all trees is chosen the confusion matrix is a straightforward yet impactful
as the final prediction (mode). In regression tasks, the concept. It comprises an N x N matrix utilized for assessing
average of all predictions is taken as the final prediction. By the performance of a classification model, with N
averaging the predictions of multiple decision trees trained representing the number of target groups. Within this
on different subsets of data, Random Forest can effectively matrix, the actual target values are compared against the
reduce variance and produce more accurate and robust predictions generated by the machine learning model. This
predictions compared to individual decision trees. holistic approach provides a comprehensive understanding
of the model's performance and the nature of errors it may
produce.

 Accuracy:
The accuracy metric measures the proportion of correct
predictions relative to the total number of predictions made.
It is computed by dividing the sum of true positives and true
negatives by the total number of predictions.

 Precision:
Precision evaluates the accuracy of positive predictions
by calculating the ratio of true positives to the sum of true
positives and false positives.

 Recall (Sensitivity):
Recall assesses the model's ability to identify actual
positives by determining the ratio of true positives to the
sum of true positives and false negatives.

Fig 5 Confusion Matrix for Random Forest

IJISRT24MAR444 www.ijisrt.com 376

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR444

 F1 Score: evaluates the dataset using three different algorithms,

The F1 score, also known as the F1 measure, offers a comparing their results. Through this methodology, the
balanced assessment of precision and recall. It is computed research achieves a prediction accuracy of 89.93% in
as the harmonic mean of precision and recall, providing determining whether a patient is suffering from heart disease
equal weight to both metrics. Actually, precision in the (0 denoting absence, and 1 denoting presence). Notably, the
context of a confusion matrix refers to the proportion of Logistic Regression algorithm yields the highest accuracy of
correctly predicted positive instances among all instances approximately 89.93% compared to other algorithms.
predicted as positive by the model, regardless of whether
they are true positive or false positive. It represents the VII. CONCLUSIONS
accuracy of the model in identifying positive instances
among all instances it predicted as positive. To summarize, In our study, we've highlighted the significance of
precision measures the model's ability to correctly identify managing null values and conducting feature selection to
positive instances among all instances predicted as positive, enhance the accuracy of classification models. Through
regardless of whether they are true positive or false positive. comparisons of various classification methods, our aim was
to identify the most effective classifier. By effectively
Table 1 Final Accuracy of Three Algorithms managing null values, researchers can improve the overall
Name of Classification Confusion Accuracy quality of the dataset and enhance the reliability of the
Algorithm Matrix classification models' outcomes. This, in turn, contributes to
SVM TP=46 FP=2 83.33% more accurate predictions and better-informed decision-
TN=9 making in healthcare or related domains.We found that each
FN=9 classification method demonstrated commendable
LOGISTIC TP=47 89.39% performance in handling observations with null values
REGRESSION FP=1 through mean imputation and employing the information
TN=12 gain feature selection strategy. However, it's worth noting
FN=6 that the dataset's lower-contributing features and presence of
RANDOM TP=46 87.88% null values may have contributed to lower classification
FOREST FP=2 accuracy. Among the three classifiers assessed, the Logistic
TN=12 Regression Algorithm emerged as the most effective,
FN=6 achieving an overall accuracy of 89.93%, along with high
recall scores. Moving forward, it's advisable to explore
VI. RESULTS AND DISCUSSIONS alternative classification algorithms that incorporate more
robust feature selection methods for improved performance.
Heart failure presents a significant global health
challenge, impacting over 550,000 individuals annually. REFERENCES
Enhancing prediction methods for this condition is pivotal in
mitigating its effects. The results of the 13-feature [1]. C. Boukhatem, H. Y. Youssef and A. B. Nassif,
classification models indicate the predictive performance of "Heart Disease Prediction Using Machine Learning,"
each model in determining the existence of certain 2022 Advances in Science and Engineering
outcomes, such as heart failure or heart disease. These Technology International Confserences (ASET),
models leverage a set of 13 features derived from clinical Dubai, United Arab Emirates, 2022, pp. 1-6, doi:
data or other relevant sources to make predictions. 10.1109/ASET53988.2022.9734880.
Leveraging advancements in machine learning (ML) within [2]. K. M. Hridoy et al., "Heart Disease Prediction Using
the healthcare sector, we can develop models capable of Machine Learning Algorithms," 2023 4th
predicting heart diseases. This research employs a dataset International Conference on Big Data Analytics and
comprising various clinical attributes such as patient age, Practices (IBDAP), Bangkok, Thailand, 2023, pp. 1-
sex, chest pain, and fasting blood sugar (Fbs), etc..,The 6, doi: 10.1109/IBDAP58581.2023.10271997.
dataset is then partitioned into two subsets: a training set [3]. S. Ibrahim, N. Salhab and A. E. Falou, "Heart disease
comprising 70% of the data and a testing set comprising Prediction using Machine Learning," 2023 1st
30%. One critical aspect to consider in the analysis of these International Conference on Advanced Innovations in
models is the presence of observations with null values. Null Smart Cities (ICAISC), Jeddah, Saudi Arabia, 2023,
values, or missing data points, can significantly impact the pp. 1-6, doi: 10.1109/ICAISC56366.2023.10085522.
reliability and accuracy of the models' predictions. It's [4]. K. M. Hridoy et al., "Heart Disease Prediction Using
essential to address these null values appropriately during Machine Learning Algorithms," 2023 4th
the pre-processing stage to ensure the robustness of the International Conference on Big Data Analytics and
models. Various techniques can be employed to handle null Practices (IBDAP), Bangkok, Thailand, 2023, pp. 1-
values, including mean imputation, median imputation, or 6, doi: 10.1109/IBDAP58581.2023.10271997.
removing observations with null values altogether. The [5]. Chicco, Davide, and Giuseppe Jurman. "Machine
choice of technique depends on the nature of the data and learning can predict survival of patients with heart
the specific requirements of the modelling task.The training failure from serum creatinine and ejection fraction
set is utilized to construct the predictive model, while the alone." BMC medical informatics and decision
testing set is employed to assess its accuracy. The research making 20.1 (2022): 16.

IJISRT24MAR444 www.ijisrt.com 377

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR444

[6]. B. Jin, C. Che, Z. Liu, S. Zhang, X. Yin and X. Wei,

"Predicting the Risk of Heart Failure with EHR
Sequential Data Modeling," in IEEE Access, vol. 6,
pp. 9256-9261,2018.
[7]. S. Mohan, C. Thirumalai and G. Srivastava,
"Effective Heart Disease Prediction Using Hybrid
Machine Learning Techniques," in IEEE Access, vol.
7, pp. 81542-81554, 2019.
[8]. Yilmaz R, Yagin FH. Early detection of coronary
heart disease based on machine learning methods.
International Medical Journal. 2022 Jan 1; 4(1): 1–6.
doi: 10.37990/medr.1011924.
[9]. Riyaz L, Butt MA, Zaman M, Ayob O. Heart disease
prediction using machine learning techniques: a
quantitative review. International Conference on
Innovative Computing and Communications, pp. 81–
94, vol. 1394, Singapore: Springer; 2022.
[10]. S. Babu et al., "Heart disease diagnosis using data
mining technique," 2017 International Conference of
Electronics, Communication and Aerospace
Technology (ICECA), Coimbatore, India, 2017, pp.
750-753, doi: 10.1109/ICECA.2017.8203643.

IJISRT24MAR444 www.ijisrt.com 378