An Overview of the Heart Disease Prediction Using Machine Learning and its Application

Cardiovascular diseases are among the most prevalent health conditions worldwide, contributing to significant morbidity and mortality. According to the World Health Organization, early detection and intervention are crucial in mitigating risks and improving patient outcomes. Advances in data science and machine learning have paved the way for predictive systems that can assist healthcare providers in diagnosing and managing such conditions more effectively. The heart is a vital organ responsible for pumping blood throughout the body, ensuring the proper functioning of all other organs. If the heart fails to operate correctly, critical systems such as the brain and other organs cease functioning, leading to death within minutes. Changes in lifestyle, work-related stress, and unhealthy dietary habits have significantly contributed to the rise of heart-related diseases globally. Heart diseases have emerged as a leading cause of mortality worldwide. According to the World Health Organization (WHO), cardiovascular diseases account for approximately 17.7 million deaths annually, representing 31% of all global deaths. In India, CVDs account for a significant proportion of deaths, ranging from 30% to 42%. The Global Burden of Disease study estimates that the age-standardized CVD death rate in India is 272 per 100,000 people, which is higher than the global average of 235 per 100,000. In India, heart-related diseases have become the primary cause of death, with 1.7 million fatalities reported in 2016, as per the 2016 Global Burden of Disease Report. The economic impact is equally severe; between 2005 and 2015, India is estimated to have

LITERATURE SURVEY

Chala Beyene et al [1] proposed a framework for the Prediction and Analysis of Heart Disease Occurrence Using Data Mining Techniques. The primary goal of their methodology is to facilitate early and automated diagnosis of heart disease, delivering results swiftly. This approach is particularly beneficial in healthcare organizations with limited expertise or insufficient specialized skills. The proposed system employs a variety of medical attributes, including blood sugar levels, heart rate, age, and sex, to determine whether an individual is at risk of heart disease. By leveraging these attributes, the framework aims to improve the accuracy of predictions and assist in timely medical intervention. Senthil Kumar Mohan et al [2], proposed a hybrid machine learning approach for predicting heart diseases using the Cleveland dataset. Their method begins with a data pre-processing step, where tuples with missing values are removed, and non-essential attributes like age and sex are excluded, as they were deemed personal and irrelevant to prediction accuracy. The remaining 11 attributes, which hold significant clinical relevance, were retained for analysis. The authors introduced a Hybrid Random Forest Linear Method (HRFLM), combining Random Forest (RF) and Linear Method (LM). The HRFLM framework comprises four main algorithms:

1. Dataset Partitioning: A decision tree partitions the input dataset into feature-specific leaf nodes.
2. Rule Application: Classification rules are applied to the partitioned dataset, resulting in labeled outputs.
3. Feature Extraction: Using a Less Error Classifier, the

METHODOLOGY

The Heart Disease Prediction System is structured into five main modules: data structure analysis, data visualization, feature engineering, model building, and prediction. Each module is designed to address a specific aspect of the data science pipeline, ensuring a comprehensive approach to data exploration and model development.

Data Loading and Exploration

The application begins by loading the heart disease dataset, which contains various medical attributes such as age, cholesterol levels, blood pressure, and others. These attributes are crucial for predicting the target variable, which indicates the presence or absence of heart disease. The dataset is loaded using Pandas and cached to improve performance. Users can explore the dataset's structure, including its shape, column names, data types, and summary statistics. This step provides a foundational understanding of the data. The first step in the methodology involves loading the dataset and performing initial exploration to understand the data structure, identify missing values, and recognize patterns. For this project, we use the Heart Disease dataset, which contains 303 instances, each representing a patient’s medical record. Each record includes 14 attributes that describe various aspects of the patient's medical condition, such as age, sex, blood pressure, cholesterol levels, and whether they suffer from heart disease.

a. Loading the Data

The dataset is loaded from a CSV file using a Python library like Pandas. Pandas provides an efficient way to read and manipulate structured data. The dataset is loaded into a Data Frame for easy

3. Random Forest Classifier

Overview

Random Forest is an ensemble learning method that builds multiple decision trees during training and merges their outputs to improve accuracy and reduce overfitting. It is a robust model that performs well on many tasks and is particularly effective for classification problems.

Working

Bootstrap Aggregating (Bagging): Random Forest uses a technique called bagging, where multiple decision trees are trained on different random subsets of the data. These subsets are generated by bootstrapping, which means randomly sampling with replacement from the training data. Feature Randomization: At each node, the algorithm randomly selects a subset of features for the split, reducing the correlation between trees and ensuring diversity in the forest. This hyperplane that best separates the data into two classes. The model assigns labels based on which side of the hyperplane the data points fall on. Non-linear SVM: For non-linearly separable data, SVM uses kernel functions to map the data into a higher-dimensional space where a linear separator can be found. Common kernel functions include:

SVM is highly effective for binary classification tasks and works well in high-dimensional spaces. However, it can be computationally expensive, especially with large datasets or when using complex kernels. Regularization (parameter C) and kernel choice are crucial for the model’s performance. Cross-validation is employed to evaluate the model’s generalization ability and to avoid overfitting.

OUTCOME OF

Prediction Module

The prediction module allows users to input patient-specific information, such as age, blood pressure, cholesterol levels, etc., which is processed and fed into a selected machine learning model, such as Support Vector Machine (SVM). The model predicts the likelihood of heart disease, with visual feedback provided for clarity.

Steps in Prediction

1. Input Data: The user provides essential details like age, cholesterol, ECG, etc.
2. Preprocessing: The data is cleaned, normalized, and encoded to ensure it’s suitable for model input.
3. Model Prediction: The preprocessed data is fed into the machine learning model (SVM), which outputs the likelihood of heart disease.

to achieve its objectives:

• Python: The primary programming language for implementing data processing, visualization, and machine learning functionalities.
• Streamlit: A Python-based framework for building interactive web applications, allowing seamless integration of visualizations and user interfaces.
• Pandas and NumPy: Libraries for data manipulation and numerical computations.

Matplotlib and Seaborn: Visualization libraries for creating graphs and plots that facilitate data exploration and feature analysis.

• Scikit-learn: A machine learning library offering tools for model implementation, evaluation, and pre-processing.
• StandardScaler: Used for feature scaling to standardize numerical variables.

• Machine Learning Models: KNN, Decision Trees, Random Forests, and SVM are implemented for predictive