Research paper on

HEART DISEASE PREDICTING

By
Mr. Aniruddha Ambre
Roll No: 31031523001
MSC CS Part - II

Under the Guidance of

Dr. Swati Maurya

S K Somaiya College
Department of Information Technology and Computer Science
Somaiya vidyavihar University, Mumbai

2024-25
##Abstract
Heart disease Remain one of the leading causes of
mortality globally. new diagnosing get importantly
better endurance rates and cut discourse costs. In the
research we explore Machine learning techniques to
predict heart disease based on clinical Information.
exploitation associate in nursing open-source dataset
we apply and value respective sorting Procedures
including logistical regression ,decision trees, suport
vector machines (svm). Our results demonstrate that
machine learning can effectively identify potential
heart disease cases providing a promising tool for
healthcare Uses.

##Introduction
Heart disease encompasses a range of conditions that
affect the heart's , Including coronary artery disease
arrhythmias and heart failure. Identifying high-risk
patients early is crucial for effective management.
Traditional diagnostic methods rely on clinical expertise
and invasive Checks but they can be time-consuming
and costly. Machine learning offers amp non-invasive
data-driven approach to predict heart disease using
patient data.
 This report investigates the feasibleness of Machine
learning Procedures inch predicting Harth disease and
compares their operation along amp publically free
data set

##Background and Related Work

Machine learning has gained traction in healthcare due to
its ability to Examine Complicated data sets and uncover
Layouts. respective studies bear explored its diligence inch
predicting heart disease:
Framingham Heart study: researchers old logistical regress to
call cardiovascular risks founded along Goal and demographic
Information
Deep learning approaches: Recent works employ neural
networks for high-dimensional ekg point Method
comparative studies: Procedures such as Random Forest and
SVM have shown promise in detecting heart conditions with
high accuracy.

##Model and important Concepts

1.Model Used
Logistic Regression: A simple yet effective bas for binary
classification.
Decision Trees: A rule-based Representation that splits
Information iteratively to classify samples.
Random Forest: An ensemble of decision trees reducing
overfitting and improving Precision.
Support Vector Machine (SVM): Finds the optimal hyperplane
to separate classes in high-dimensional space.

2. important Concepts
Characteristic Selection: Identifying difficult Characteristics
such as age cholesterol level and blood pressure.
Data Pre-processing: Handling missing values scaling and
encoding categorical Information.
Evaluation Metrics: Using accuracy ,precision recall and F1-
score to compare Representations.

##Implementation
1. Dataset
The Data set used is sourced from the UCI Heart Disease
Storage. it includes 303 instances with cardinal attributes
such as age, chest pain type resting blood pressure, and
cholesterol levels.
2. information pre processing
handled lost values away mean/mode imputation normalized
perpetual variables for break Check Effectiveness one-hot
encoded flat Characteristics
3. model training and evaluation
 Split information into 80% education and 20% examination
subsets performed hyperparameter tuning exploitation
gridiron look for Random Forest and support vector
machine(svm)
Evaluated models using a confusion matrix, ROC-AUC curve,
and cross-validation.

##Results:

Logistic Regression
Accuracy: 85.7%
Logistic Regression is a baseline classification Model that
calculates the probability of a binary outcome (heart disease
or not). inch this suit it right expected 857% of the instances
precision: 830%
precision is the part of bold predictions that are extremely
right. For Logistic Regression 83.0% of the time when it
predicted heart disease the prediction was correct.
Recall: 84.2%
 Recall measures the ability of the Representation to correctly
identify positive cases (true positives). inch this suit it right
known 842% of complete the fast cases of hearth disease
f1-score: 836%
f1-score is the sympathetic base of preciseness and think.
Logistic Regression strikes a good balance between precision
and recall with an F1-score of 83.6%.

Decision Tree
 Accuracy: 82.1%
Decision Trees work by splitting the data into subsets
based on the feature that best separates the classes.
Here, it achieved 82.1% accuracy, meaning that around
82% of the predictions were correct.
 Precision: 80.2%
This model predicted heart disease correctly 80.2% of
the time when it made a positive prediction.
 Recall: 81.4%
Recall indicates that the Decision Tree model identified
81.4% of the actual heart disease cases.
 F1-Score: 80.8%
The F1-score of 80.8% shows that while the Decision
Tree is effective, it’s slightly less balanced than Logistic
Regression.

Random Forest
  Accuracy: 88.3%
Random Forest, an ensemble model based on multiple
decision trees, provided the highest accuracy at 88.3%. It
aggregates the results of many decision trees, making it
more robust and less prone to overfitting.
 Precision: 85.4%
Precision for Random Forest is 85.4%, meaning that
when it predicted heart disease, it was correct 85.4% of
the time.
 Recall: 86.7%
Recall for this model is 86.7%, indicating it identified
86.7% of the actual heart disease cases.
 F1-Score: 86.0%
The F1-Score is the highest among all models, suggesting
that Random Forest is the most balanced in terms of
precision and recall.
Support Vector Machine (SVM)
 Accuracy: 84.5%
SVM is a powerful model that finds the optimal
hyperplane separating classes. It achieved 84.5%
accuracy, correctly predicting heart disease in most
cases.
 Precision: 82.7%
When SVM predicted heart disease, it was correct 82.7%
of the time.
  Recall: 83.9%
The model identified 83.9% of all true heart disease
cases.
 F1-Score: 83.3%
With an F1-score of 83.3%, SVM performs well in
balancing precision and recall, though it slightly lags
behind Random Forest in overall performance.
##Conclusion
This research demonstrates the potential of Machine learning
in predicting heart disease using clinical Information. Random
Forest proved to be the most effective model, offering high
accuracy and interpretability. While Machine learning
provides a promising tool for healthcare further work is
needed to Combine these methods into clinical workflows.
prospective search might search advance sound acquisition
techniques and corroborate Representations along big
different data sets