A

Project Report On
Fake News Detection Using Hybrid Machine Learning Techniques
Submitted in partial fulfillment of the requirements For the award of the degree of

Bachelor of Technology
in
Computer Science and Engineering

By

Shashank Kumar Singh (2100970100105)

Shivam Kumar (2100970100106)
Sujal Chauhan (2100970100118)

Under the Supervision of

Dr. Aditya Dev Mishra

Galgotias College of Engineering and Technology Greater Noida,

Uttar Pradesh
Affiliated to

Dr. A.P.J Abdul Kalam Technical University Lucknow,

(Session 2024-25)
 GALGOTIAS COLLEGE OF ENGINEERING & TECHNOLOGY
GREATER NOIDA, UTTAR PRADESH, INDIA- 201306 .

CERTIFICATE

This is to certify that the project report entitled “ Heart Disease Prediction Using Machine
Learning Techniques” submitted by [Link] Kumar(2100970100106), [Link]
Chauhan (2100970100118), Mr. Shashank Kumar Singh (2100970100105) to the Galgotias
College of Engineering & Technology, Greater Noida, Uttar Pradesh, affiliated to Dr. A.P.J. Abdul
Kalam Technical University Lucknow, Uttar Pradesh in partial fulfillment for the award of Degree of
Bachelor of Technology in Computer science & Engineering is a Bonafide record of the project work
carried out by them under my supervision during the year 2024-2025.

[Link] Dev Mishra [Link] Chaudhary

[Link] CSE [Link] Head [Link] Dept

2
 ACKNOWLEDGEMENT

We have taken efforts in this project. However, it would not have been possible without the
kind support and help of many individuals and organizations. We would like to extend my
sincere thanks to all of them.

We are highly indebted to [Link] Dev Mishra for his guidance and constant supervision.
Also, we are highly thankful to them for providing necessary information regarding the
project & also for their support in completing the project.

We are extremely indebted to Dr. Pushpa Chaudhary , HOD, Department of Computer

Science and Engineering, GCET and Mr. Manish Kumar Sharma and Dr. Sanjay Kumar,
Project Coordinator, Department of Computer Science and Engineering, GCET for their
valuable suggestions and constant support throughout my project tenure. We would also like
to express our sincere thanks to all faculty and staff members of the Department of
Computer Science and Engineering, GCET for their support in completing this project on
time.

We also express gratitude towards our parents for their kind cooperation and encouragement
which helped me in completion of this project. Our thanks and appreciation also go to our
friends in developing the project and all the people who have willingly helped me out with
their abilities.

(Shashank Kumar Singh)

(Shivam Kumar)
(Sujal Chauhan)

3
 ABSTRACT

Heart disease is a major cause of death throughout the world. It is difficult to predict by medical practitioners as it requires
expertise and higher knowledge of prediction. The environment in healthcare sector is information rich but lacks knowledge.
A lot of data is available in healthcare systems over the internet but there is a lack of effective analysis tool to discover hidden
patterns in data. An automated system will enhance medical efficiency and reduce cost and time. This software intends to
predict the occurrence of a disease based on the data which is gathered from kaggle. The objective is to extract the hidden
patterns by applying data mining techniques on the dataset and to predict the presence value on a scale. The prediction of heart
disease requires a huge size of data which is too massive and complex to process and analyse by conventional technique. Our
aim is to find out an suitable technique that is efficient and accurate for prediction of cardiac disease.

Keywords-prediction heart disease, machine learning, algorithm, analysis

4
 CONTENTS

Title Page

CERTIFICATE 2

ACKNOWLEDGEMENT 3

ABSTRACT 4

CONTENTS 5

LIST OF TABLES 6

LIST OF FIGURES 6

CHAPTER 1: INTRODUCTION 8

CHAPTER 2: LITERATURE REVIEW 11

CHAPTER 3: PROBLEM FORMULATION 15

CHAPTER 4: PROPOSED WORK 18

CHAPTER 5: SYSTEM DESIGN 25

CHAPTER 6: IMPLEMENTATION 32

CHAPTER 7: RESULT ANALYSIS 36

CHAPTER 8: CONCLUSION, LIMITATION AND FUTURE SCOPE 37

LIST OF PUBLICATION 39

CONTRIBUTION OF PROJECT 40

5
LIST OF TABLES

Table Page
Table No.
Title

1 Accuracy comparison of three algorithm 6

2 Testing of modules 33

LIST OF FIGURES

Figure No. Figure Title Page

Fig1.1 Support vector machine graph 4

Fig2.1 17
Block Diagram of Problem Formulation

Fig 5.1 28
Structural Design of the model

Fig 5.2 Scalability design of the model 29

Fig 5.3 System Architecture Diagram 29

Fig 5.4 Classification Module Workflow 30

Fig 5.5 Overall system design 32

6
Fig 6.1 Workflow of the Implementation Model 33

Fig 6.2 34
Comparative Result Analysis Between Logistic
Regression, Random Forest and Decision Tree

7
 CHAPTER 1

Introduction
1.1 Introduction

Heart disease, a leading cause of mortality worldwide, is a pressing health concern that demands
innovative approaches for early detection and prevention. In recent years, advancements in medical
science and technology have highlighted the critical role of data-driven methods in predicting,
diagnosing, and managing diseases. Among these methods, Machine Learning (ML) has emerged as a
powerful tool, offering the ability to process vast datasets, uncover hidden patterns, and provide reliable
predictions. This report explores the application of machine learning techniques in predicting heart
disease, emphasizing its potential to revolutionize healthcare delivery.

Heart disease encompasses a range of conditions that affect the heart, including coronary artery disease,
arrhythmias, and heart valve disorders. Early detection is vital as it allows for timely interventions,
reducing the risk of severe complications. Traditionally, diagnosis relies on clinical expertise, physical
examinations, and various diagnostic tests, which can be time-consuming and resource-intensive.
Moreover, human error and subjective interpretation of test results can sometimes lead to misdiagnosis..

With the advent of machine learning, healthcare providers can analyze large datasets collected from
medical records, wearable devices, and diagnostic tools. ML models, trained on these datasets, can predict
the likelihood of heart disease based on various risk factors such as age, gender, blood pressure,
cholesterol levels, smoking habits, and family medical history.

Objectives of the Study

 Data Analysis: Analyzing the key risk factors and patterns associated with heart disease.
 Model Development: Applying machine learning algorithms such as Logistic Regression, Decision
Trees, Random Forest, and Neural Networks to develop a predictive model.
 Performance Evaluation: Assessing the accuracy, sensitivity, and specificity of the model using appropriate
 Metrics.
 Practical Application: Demonstrating how the model can assist healthcare professionals in identifying high-risk
individuals and preventive measures.

8
 Importance of ML in heart disease prediction

Machine learning enables the creation of predictive models that are not only fast but also scalable, making
them ideal for integration into clinical workflows. These models can provide:

 Improved Diagnostic Accuracy: Reducing false positives and negatives by considering subtle patterns
in the data.
 Personalized Healthcare: Tailoring treatment plans based on individual risk profiles.

 Resource Optimization: Automating routine diagnostic processes to save time and reduce costs.

1.2 Scope of the Report

The report delves into the methodology of building an ML-based heart disease prediction model,
including data preprocessing, feature selection, algorithm selection, and model evaluation. It also
discusses the challenges encountered during model development, such as data quality issues, overfitting,
and the interpretability of results. Finally, the report highlights the ethical considerations and future
prospects of using machine learning in healthcare.

In conclusion, the application of machine learning to heart disease prediction represents a significant step
forward in the fight against cardiovascular illnesses. By integrating data science with medical expertise,
we can enhance our ability to predict and prevent heart disease, ultimately saving lives and

9
 CHAPTER 2: LITERATURE REVIEW

2.1 Related Work

The prediction of heart disease using machine learning has gained significant attention in recent years.
This section highlights several key studies that have contributed to this field, comparing the effectiveness
of various algorithms and approaches used for heart disease prediction.

Gupta et al. (2021) utilized a combination of Logistic Regression and Support Vector Machine (SVM) to
predict heart disease. Their study reported an accuracy of 85% on the UCI Heart Disease dataset. The authors
observed that Logistic Regression excelled at handling linear relationships, while SVM effectively captured
complex, nonlinear patterns. This hybrid approach demonstrated that combining algorithms could improve
prediction accuracy by leveraging their individual strengths【1】.

Figure 1.1 : Support vector machine graph.[1]

Sharma et al. (2022) explored the application of Decision Trees and Random Forests for heart disease
prediction. They concluded that Random Forests outperformed Decision Trees, achieving an accuracy of
88%. The ensemble nature of Random Forests helped mitigate overfitting, which is a common limitation
of Decision Trees. This study emphasized the importance of ensemble methods in improving model
reliability and generalization【2】.

10
Kumar et al. (2023) integrated neural network models with clinical feature selection to enhance
prediction performance. By using techniques such as Principal Component Analysis (PCA) for
dimensionality reduction, they achieved an accuracy of 90%. Neural networks were particularly effective
in identifying subtle interactions among features, making them suitable for complex datasets. However,
the authors noted that these models required significant computational resources and careful tuning to
avoid overfitting【3】.

Ali et al. (2022) focused on data preprocessing techniques, including handling missing values and
balancing datasets using the Synthetic Minority Oversampling Technique (SMOTE). Their study applied
Gradient Boosting algorithms, which achieved an accuracy of 92%. The results highlighted the critical
role of data quality and preprocessing in ensuring robust model performance【4】.

Ravi et al. (2023) compared the performance of Logistic Regression, SVM, and k-Nearest Neighbors (k-
NN) for heart disease prediction. They found that SVM consistently outperformed the other models,
particularly in high-dimensional datasets, achieving an accuracy of 89%. This study demonstrated the
effectiveness of SVM in finding optimal decision boundaries in complex feature spaces【5】.

Wang et al. (2023) proposed a model integrating clinical data with real-time data from wearable devices
using Gradient Boosting Machines. Their model achieved an accuracy of 91%, showing that
incorporating real-time health metrics, such as heart rate and physical activity, could enhance the
accuracy of predictions. This approach emphasized the potential of IoT-integrated ML models in
personalized healthcare【6】.

Singh and Kaur (2022) conducted a review of various machine learning algorithms for heart disease
prediction, focusing on their scalability and real-world applicability. They highlighted that ensemble
methods, such as Random Forests and Gradient Boosting, consistently outperformed simpler algorithms
due to their ability to handle complex interactions and imbalanced datasets【7】.

Comparative Analysis

The following table summarizes the key findings from the reviewed studies and compares the
performance of different machine learning techniques for heart disease prediction:

11
 Study Techniques Used Dataset Accuracy Key Insights

Hybrid models leverage strengths

Logistic Regression, SVM of linear and nonlinear algorithms

Gupta et al. (2021) UCI Heart Dataset 85% for better prediction accuracy.

Ensemble methods mitigate

88%
Sharma et al. (2022) Decision Tree, Random Clinical Records overfitting and improve reliability.
Forest

Feature reduction enhances

Kumar et al. (2021) Neural Networks, PCA Clinical + Lab 90%
model efficiency and accuracy
Data

Data preprocessing and balancing

Ali et al. (2022) Gradient Boosting, SMOTE UCI Heart Dataset 92% are crucial for robust performance.

SVM excels in high-dimensional

Logistic Regression, SVM, k-NN feature spaces with superior decision

Ravi et al. (2023) UCI Dataset SVM: 89% boundary optimization.

.
Real-time IoT data integration
Wang Gradie 91%
nt Wearable + Clinical enhances prediction models.
[Link]
Boosti
(2023) ng +
IoT

Ensemble methods are preferred

Multiple Al gorithms
Singh & Kaur (2022) Review of Studies NA for scalability and handling
complex interactions.

12
 The reviewed studies demonstrate a growing preference for ensemble models and hybrid approaches,
which consistently outperform single-model techniques. Data quality, preprocessing, and the integration
of diverse features (e.g., wearable data) significantly influence model accuracy and generalizability.

References

[1] Gupta, R. et al. "A Hybrid Approach for Heart Disease Prediction Using Logistic Regression and SVM,"
International Journal of Healthcare Analytics, 2021.

[2] Sharma, P. et al. "Ensemble Methods for Heart Disease Prediction," Medical Informatics Journal, 2022.

[3] Kumar, A. et al. "Deep Learning for Predicting Cardiovascular Conditions," Clinical AI Journal, 2023.

[4] Ali, M. et al. "Data Preprocessing and Gradient Boosting for Heart Disease Prediction," AI in Medicine
Proceedings, 2022.

[5] Ravi, S. et al. "Performance Comparison of ML Algorithms for Heart Disease Prediction," Journal of Data
Science, 2023.

[6] Wang, X. et al. "Real-Time IoT Data Integration in ML Models for Healthcare," IEEE IoT Journal, 2023.

[7] Singh, A. & Kaur, J. "Scalability of ML Algorithms in Medical Predictions," Journal of Computer Applications
in Medicine, 2022.

13
 CHAPTER 3

PROBLEM FORMULATION

Description of Problem Domain

Cardiovascular diseases (CVDs), particularly heart disease, are the leading cause of mortality worldwide,
claiming millions of lives annually. Despite advances in healthcare, early detection remains a significant
challenge. Timely diagnosis is critical for preventing severe complications, reducing mortality, and
improving patient quality of life. Traditional methods for diagnosing heart disease, such as physical
examinations, electrocardiograms (ECG), and imaging, are effective but resource-intensive, time-
consuming, and often inaccessible in low-resource settings. These limitations underscore the urgent need
for innovative, scalable solutions.

Machine Learning (ML) has emerged as a transformative tool for healthcare, offering the ability to
process large datasets, uncover hidden patterns, and make accurate predictions. In the context of heart
disease prediction, ML models can analyze clinical and lifestyle data, including age, cholesterol levels,
blood pressure, and smoking habits, to assess an individual’s risk. These models not only improve
diagnostic accuracy but also facilitate personalized interventions. Furthermore, integrating ML with
wearable devices enables real-time monitoring, making it possible to detect early warning signs and
provide timely alerts.

To combat this pressing issue, automated solutions leveraging advancements in artificial intelligence (AI)
and natural language processing (NLP) are critical. These systems can analyze vast datasets in real-time,
identifying fake news before it spreads widely. By leveraging interdisciplinary approaches, including
social sciences and technological advancements, this research contributes to mitigating the societal impact
of fake news, restoring public trust in digital platforms, and ensuring the integrity of shared information.

Problem Statement

Current diagnostic methods for heart disease are expensive, time-consuming, and heavily reliant on expert
analysis, making them inaccessible for many, especially in under-resourced settings. These approaches
often fail to detect the disease in its early stages, leading to delayed interventions and poorer outcomes.

This research aims to design an automated heart disease prediction system leveraging advanced machine
learning techniques. The system will analyze patient data to provide accurate, early risk predictions,
supporting clinicians in decision-making and enabling timely interventions. By improving early detection
and incorporating real-time data monitoring, the proposed system seeks to enhance healthcare
accessibility, reduce the burden on medical professionals, and improve patient outcomes..

14
Depiction of Problem Statement

1. The challenge of detecting heart disease lies in its silent progression, where early symptoms are
often missed due to reliance on traditional diagnostic methods like physical exams and imaging.
These approaches are resource-intensive, requiring specialized expertise and infrastructure, which
limits their accessibility in under-resourced settings. Additionally, conventional methods struggle
to analyze the complex interplay of clinical and lifestyle factors, such as age, cholesterol levels,
and smoking habits, which significantly influence disease risk.

2. This research proposes an automated system for heart disease prediction utilizing advanced
machine learning algorithms, including Logistic Regression, Random Forest, and Support Vector
Machines (SVM). The system leverages structured data from clinical and lifestyle metrics,
applying preprocessing, feature selection, and classification techniques to provide accurate, early
risk assessments.

3. Designed for scalability and real-time analysis, the system integrates predictive models with
capabilities for early detection and personalized healthcare recommendations. Challenges such as
data quality, interpretability, and generalizability are addressed through robust validation and
optimization strategies. The workflow encompasses data acquisition, preprocessing, feature
engineering, model training, and evaluation

Fig. 2.1: Block Diagram of Problem Formulation

15
Objectives

The objectives of the proposed research are as follows:

● Develop a machine learning-based system to predict heart disease with high

accuracy by analyzing clinical and lifestyle datasets..

● Design the system to support real-time risk assessment, enabling early

detection and timely medical interventions.

● Build a robust and scalable framework capable of handling large, diverse

datasets and ensuring generalizability across populations.

● Improve detection accuracy and reliability to restore trust in digital platforms.
● Leverage advancements in AI, NLP, and social sciences to provide holistic

solutions to the fake news problem.

● Mitigate the societal effects of misinformation in sectors like healthcare,

politics, and public safety by curbing its spread.

16
 CHAPTER 4

PROPOSED WORK
The proposed research aims to develop a machine-learning-based system for predicting heart disease,
focusing on accuracy, scalability, and real-time risk assessment. This system combines clinical
and lifestyle data analysis with advanced algorithms, enabling timely intervention and reducing
the societal burden of heart disease. Below are the detailed steps and methodologies involved:
.

Data Collection and Preprocessing

The data will undergo several preprocessing steps to ensure its quality and suitability for
machine learning models:

a. Data Collection
i. Collect structured datasets from reliable sources like UCI Machine Learning
Repository or Kaggle, which contain attributes such as age, gender, cholesterol
levels, blood pressure, and exercise habits.
.
b. Data Cleaning
i. Remove missing, redundant, or inconsistent values to ensure the dataset's quality.

c. Future Scaling
i. Standardize numerical features like cholesterol levels and blood pressure to prevent
dominance
ii. of higher magnitude features during model training.

d. Categorical Encoding
i. Convert categorical variables (e.g., gender or chest pain type) into numerical
representations using one-hot encoding or label encoding.

e. Data Spilliting
i. Stop words are common words like "is," "and," "the," etc., which do not carry
significant meaning in the context of text analysis. These words are removed
from the dataset to ensure that the model focuses on the more meaningful words
that help distinguish real news from fake news.

17
 Feature Extraction
Feature extraction is a critical step in transforming text data into a numerical format that can be
understood and processed by machine learning algorithms. The following methods will be
employed:

1. Correlation Analysis
Identify features most correlated with the target variable (heart disease presence).

2. Principal Component Analysis(PCA)

Reduce dimensionality, retaining key features that explain the majority of variance in the
data.

.Model Development

1. Machine Learning Models:

Utilize a combination of the following algorithms for prediction:

 Logistic Regression: For a probabilistic model with interpretable coefficients.

 Support Vector Machine (SVM): To handle non-linear decision boundaries.
 Decision Tree: For easily interpretable decision-making structures.
 Random Forest: To leverage ensemble learning for improved accuracy and robustness.
 Naive Bayes: For handling smaller datasets with speed and simplicity.

2. Hyperparameter Tuning:
Use techniques like grid search or random search to optimize model parameters for better performance.

3. Model Training and Validation:

Train models on the training set and validate performance using metrics like accuracy, precision, recall.

18
These methods will help the system convert raw text into meaningful numerical data, which is
essential for training machine learning models.

Machine Learning Models

Algorithm for Logistic Regression

Given a training dataset D = {(a1, b1), (a2, b2), ..., (an, bn)}, where ai is the ith observation with
m features and bi is the corresponding class label variable.
1. Define the logistic function (sigmoid function) as:
P(b=1 | a) = 1 / (1 + exp(-w * a + b))
where w is the weight vector and b is the bias.
2. The goal is to learn the optimal values of w and b by minimizing the cost function,
which is the log-likelihood:
L(w, b) = -Σ [b_i * log(P(b_i | a_i)) + (1 - b_i) * log(1 - P(b_i | a_i))]
3. Use gradient descent or a similar optimization algorithm to minimize the cost function:
w_new = w_old - α * ∇ L(w, b), where α is the learning rate.
4. Once the optimal weights are found, classify a new observation by computing:
P(b=1 | a) = 1 / (1 + exp(-w * a + b))
and assign the class label b=1 if P(b=1 | a) > 0.5, else b=0.
The Logistic Regression algorithm can be expressed mathematically as:
P(b=1 | a) = 1 / (1 + exp(-w * a + b))

Algorithm for Support Vector Machine (SVM)

Given a training dataset D = {(a1, b1), (a2, b2), ..., (an, bn)}, where ai is the ith observation with
m features and bi is the corresponding class label.
1. Define the hyperplane as the decision boundary separating different classes:
w * a + b = 0, where w is the weight vector and b is the bias term.
2. The goal is to find the optimal hyperplane that maximizes the margin (distance
between the hyperplane and the nearest data points from each class).
3. The optimization problem can be formulated as:
min (1/2) * ||w||^2 subject to y_i * (w * a_i + b) ≥ 1, for all i.
4. Use a technique like the **Lagrange Multiplier method** to solve the constrained
optimization problem.
5. Once the optimal hyperplane is found, classify a new observation by determining
which side of the hyperplane it lies on:
f(a) = sign(w * a + b)
where sign(x) returns 1 for positive values and -1 for negative values.

19
 The Support Vector Machine (SVM) algorithm can be expressed mathematically as:
f(a) = sign(w * a + b)

Algorithm for Decision Tree

Given a training dataset D = {(a1, b1), (a2, b2), ..., (an, bn)}, where ai is the ith observation with
m features and bi is the actual class label variable.
1. Select the best feature from the training dataset D based on a specified criterion. This is
done by evaluating the information gain.
2. Split the dataset D into subsets premised on data of the selected feature f. Each subset
has observations that have the same value of f.
3. Recursively repeat steps 1 and 2 on each subset till a stopping standard is achieved.
The stopping standard can be a maximum tree depth, a minimum number of observations in a
node.
4. Assign an output value to each leaf node rooted on the maximized class or average
value of the observations in the node.
The decision tree algorithm can be expressed mathematically as a function:
T(a) = {c1, if a ∈ R1; c2, if a ∈ R2; ...; cm, if a ∈ Rm}
where T(a) is the predicted output for the input vector a, and R1, R2, ..., Rm are the
regions of the input space defined by the decision tree.
Each region corresponds to a leaf node of the tree, and ci is the class label or output value
assigned to the corresponding leaf node.

Algorithm for Random Forest

Given a training dataset D = {(a1, b1), (a2, b2), ..., (an, bn)}, where ai is the ith observation with
m features and bi is the corresponding class label variable.
1. Randomly select k subsamples (with replacement) from the training dataset D. D1, D2,
..., Dk, where |Di| = |D| and each Di contains approximately two-thirds of the observations in D.
2. For each Di, grow a decision tree T using a random subset of the features at each node
of the tree.
3. Repeat steps 1 and 2 t times to create t trees T1, T2, ..., Tt.
4. To make a prediction for a new observation a, pass it through each of the t trees and
obtain the predicted class label from each tree.
5. To get the final estimation, add together the projections from all the trees. This can be
performed for classification issues by using the planned class labels' majority approval.
This can be handled for issues with regression by determining the average of the
anticipated values.

20
 The Random Forest algorithm can be expressed mathematically as:
h(a) = 1/t * Σ (i=1 to t) Ti(a)
where h(a) is the predicted output for the input vector a, and Ti(a) is the predicted output
of the ith decision tree.

Algorithm for Naive Bayes

The Naive Bayes algorithm can be represented as follows:

Given a training dataset D = {(a1, b1), (a2, b2), ..., (an, bn)}, where ai is the ith
observation with m features and bi is the corresponding class label.
1. Estimate the prior probability P(b) of each output value in the training dataset D.
P(b) = (number of observations with output value b) / (total number of observations).
2. For each feature ai, calculate the conditional probability P(ai | b) of the feature given
each output value b in the training dataset D.
P(ai | b) = (number of observations with feature ai and output value b) / (number of
observations with output value b).
3. For a new observation x with m features, calculate the posterior probability P(b | a) of
each output value b given the features of a using Bayes' theorem:
P(b | a) = P(a | b) * P(b) / P(a).
where P(a | b) = P(a1 | b) * P(a2 | b) * ... * P(am | b) is the conditional probability of the
features given the output value, and P(a) is the negligible probability of the features.
4. Assign the new observation x to the output value with the maximum probability:
b* = argmax(P(b | a)).
The Naive Bayes algorithm can be expressed mathematically as:
y* = argmax(P(b) * Π (i=1 to m) P(ai | b))
where b* is the predicted class label for the new observation a.

Performance Evaluation Formulas

After training the models, we evaluate their performance using the following standard metrics:

1. Accuracy

Accuracy is the proportion of correct predictions (both real and fake news) made by the model. It
is the most commonly used metric to assess the overall performance of a classifier.

Formula:

Accuracy = (TP + TN) / (TP + TN + FP + FN)

21
Where:

● TP = True Positives (correctly predicted fake news)

● TN = True Negatives (correctly predicted real news)
● FP = False Positives (fake news classified as real news)
● FN = False Negatives (real news classified as fake news)

2. Precision

Precision is the ratio of correctly predicted fake news instances to the total predicted as fake
news. It focuses on how many of the predicted fake news articles are truly fake.

Formula:

Precision = TP / (TP + FP)

Where:

● TP = True Positives
● FP = False Positives

3. Recall (Sensitivity)

Recall is the ratio of correctly predicted fake news instances to the total number of actual fake
news instances in the dataset. It measures how well the model identifies fake news.

Formula:

𝑅𝑒𝑐𝑎𝑙𝑙 = 𝑇𝑃 / (𝑇𝑃 + 𝐹𝑁)

Where:

● TP = True Positives
● FN = False Negatives

4. F1-Score

The F1-Score is the harmonic mean of precision and recall, providing a single measure that
balances both. It is particularly useful when the dataset is imbalanced (i.e., when fake and real
news articles are not equally represented).

Formula:

𝐹1 − 𝑆𝑐𝑜𝑟𝑒 = 2 * (𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛 * 𝑅𝑒𝑐𝑎𝑙𝑙) / (𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛 + 𝑅𝑒𝑐𝑎𝑙𝑙)

22
Where:

● Precision = Precision
● Recall = Recall

5. Confusion Matrix

A confusion matrix is a table that summarizes the performance of a classification algorithm. It

shows the number of true positives, true negatives, false positives, and false negatives.

True Positives (TP) True Positives (TP)

False Negatives (FN) True Negatives (TN)

Model Comparison and Selection

After computing the metrics for each model, we compare their performance based on these
evaluation metrics. The best-performing model will be chosen for deployment based on:

1. Highest accuracy: A higher accuracy generally indicates a better overall performance.

2. Balanced Precision and Recall: In case of an imbalanced dataset, a higher F1-Score
could be considered more relevant than accuracy alone.
3. Low False Positives (FP): Minimizing false positives is critical in fake news detection,
as incorrectly classifying real news as fake could be more damaging.
4. Low False Negatives (FN): Ensuring that fake news is detected is equally important, and
minimizing false negatives ensures no fake news articles are missed.

The model with the best combination of these metrics will be selected to ensure the system is
both reliable and effective in detecting fake news.

23
 CHAPTER 5

SYSTEM DESIGN
This chapter details the functional specifications, architectural design, and dynamic workflow of
a heart disease prediction system using machine learning. The system is designed to predict the
likelihood of heart disease in individuals based on health-related parameters, aiming to provide
accurate, scalable, and user-friendly diagnostics.

Functional Specification of the System

The primary objective of the system is to predict heart disease using a set of medical attributes
and demographic details. It leverages machine learning techniques to analyze structured medical
data and outputs a prediction score indicating the risk of heart disease..

KEY FEATURES.

1. Data Collection

Data is sourced from publicly available datasets, such as the Cleveland Heart Disease dataset
from UCI Machine Learning Repository. These datasets include patient records with
attributes like age, gender, chest pain type, resting blood pressure, cholesterol levels, fasting
blood sugar, resting ECG results, maximum heart rate, and more

2. Processing

Data preprocessing ensures that raw medical data is cleaned, normalized, and formatted for machine
Learning [Link] step involves:

 Handling missing values using imputation techniques.

 Encoding categorical variables (e.g., chest pain type) using one-hot encoding or label encoding.
 Scaling numerical data with methods like Min-Max scaling or standardization.
 Removing outliers using techniques like the Z-score or IQR-based filtering.

3. Feature Selection

Key features influencing heart disease prediction are identified using statistical tests, correlation matrices, or
feature importance metrics from models like Random Forest. Dimensionality reduction techniques such as
Principal Component Analysis (PCA) may also be employed to enhance model efficiency.

4. Modeling:
 Logistic Regression: For baseline binary classification.
 Decision Trees: For interpretable decision-making.

24
  Random Forest: To reduce overfitting and improve accuracy.
 Support Vector Machines (SVM): For high-dimensional data separation.
 K-Nearest Neighbors (KNN): For instance-based learning.
 Gradient Boosting Algorithms: Such as XGBoost or LightGBM for highly accurate predictions.

5. Model Training and Optimization:

 The dataset is split into training (80%) and testing (20%) subsets.
 Cross-validation ensures the model's generalizability.
 Hyperparameter tuning, using techniques like Grid Search or Random Search, optimizes model .

6. Evaluation Metrics:

 Accuracy: Overall correctness of predictions.

 Precision: Correct positive predictions out of total positive predictions.
 Recall: Correct positive predictions out of actual positives.
 F1-Score: Balances precision and recall.
 ROC-AUC Curve: Measures the model's ability to distinguish between classes.

Structural and Dynamic Modeling

Structural Architecture
 Input Layer:
Accepts patient data in structured format (e.g., CSV files or form submissions through a web interface).
 Preprocessing Module:
Cleans and prepares data for analysis.
 Feature Extraction Module:
Transforms input features into a format suitable for machine learning models.
 Classification Module:
Employs trained machine learning models for predictions.

25
 Output Layer:
Provides the likelihood of heart disease (e.g., percentage risk) and associated recommendations.

Dynamic Workflow
1. Input Stage:
New patient data is entered into the system via an interface or batch file upload.
2. Preprocessing:
Data is cleaned and transformed into feature vectors.
3. Feature Transformation:
Features are extracted and scaled for consistency.
4. Prediction:
Data is passed through the selected machine learning model, which outputs a prediction score and risk category.
5. Feedback Loop:
New predictions are added to the dataset for continuous learning and refinement of the models.

Scalability and Adaptability

Scalability
The system is designed to handle an increasing number of patient records and variables:

 Cloud Integration: Allows deployment on platforms like AWS or Azure for on-demand resource allocation.
 Batch Processing: Enables efficient handling of large datasets.

26
 fig 5.1 Structural Design of the model

The system includes a feedback loop for continuous learning. As the nature of misinformation
evolves, the system is updated with new data to refine its models and enhance accuracy. This
adaptive mechanism ensures that the system remains effective in identifying novel types of fake
news.

27
.

Fig: 5.2 Scalability design of the model

Fig 5.3:System Architecture Diagram

28
 Fig 5.4 :Classification Module Workflow

Overall System Design

The heart disease prediction system is a robust, machine-learning-powered diagnostic tool

capable of delivering accurate predictions of heart disease risk. It integrates modular components
for data preprocessing, feature engineering, classification, and feedback, ensuring scalability and
adaptability. The design emphasizes user accessibility, efficiency, and continuous improvement
to remain relevant in clinical and research settings.

29
Fig 5.5 :Overall system design

30
 CHAPTER 6
IMPLEMENTATION PLAN

The implementation plan for the fake news detection system is divided into distinct phases,
ensuring a clear and systematic approach to development. Each phase focuses on specific

objectives, building upon the previous one to achieve a comprehensive solution.

Fig 6.1 Workflow of the Implementation Model

Phase 1: Data Preparation

The first phase involves preparing and standardizing the medical data for analysis. This
step ensures data consistency, cleanses the dataset of noise, and extracts relevant features
for model training.

1.1: Preprocessing Tasks:.

• Handle missing values using techniques like mean/median imputation or KNN imputation.
• Encode categorical variables such as chest pain type and ECG results using one-hot or label encoding.
• Scale numerical data (e.g., age, cholesterol levels) using Min-Max scaling or standardization.
• Remove outliers based on statistical thresholds (e.g., Z-score > 3 or IQR-based filtering).

31
 1.2 Feature Engineering
• Extract features like age groups or BMI categories for better insights.
• Perform correlation analysis to retain significant attributes for predicting heart disease.

Phase 2: Model Training

In this phase, machine learning models are developed and trained to recognize patterns associated with heart disease.

Model Training Process:

o Implement baseline models like Logistic Regression to establish performance benchmarks.

o Train advanced models such as Random Forest, SVM, and Gradient Boosting algorithms to capture
complex relationships in the data.
o Use ensemble techniques like bagging and boosting to improve predictive accuracy and minimize
variability.

Training Split:

o Divide the data into an 80:20 ratio for training and testing.
o Use stratified sampling to ensure balanced class distributions in training and testing

Phase 3: Model Optimization

The third phase focuses on refining the performance of machine learning models through optimization techniques.

 Optimization Techniques:

• Perform hyperparameter tuning using methods such as Grid Search and Random Search to determine the best
configurations.
• Apply cross-validation to evaluate the generalizability of models on unseen data.

Evaluation Metrics:

• Accuracy: Measures the percentage of correct predictions.

• Precision: Assesses the proportion of true positives among predicted positives.
• Recall: Evaluates the proportion of true positives identified from all actual positives.
• F1-Score: Provides a harmonic mean of precision and recall.
• ROC-AUC Curve: Visualizes the model’s ability to separate classes.

 Outcome:

• The optimized models are fine-tuned for high accuracy and consistency, ensuring reliable predictions.

32
 ● Phase 4: Real-time Deployment
The final phase involves integrating the optimized model into a user-friendly platform for real-time predictions.

 Deployment Process:
• Host the model on cloud platforms (e.g., AWS, Azure, or Google Cloud) for scalability and reliability.
• Develop a web or mobile interface that allows users to input medical data and receive risk predictions
instantly.
• Set up APIs to connect the front-end application with the backend model for real-time interactions.

 Feedback Loop:
• Monitor system performance using real-world data and collect user feedback.
• Periodically retrain the model with updated datasets to maintain accuracy and adaptability to
evolving data patterns.

This structured approach ensures that the heart disease prediction system is effective, scalable, and adaptable, catering to the needs of both
healthcare professionals and patients.

33
 Chapter 7
Result Analysis
Performance Measure -
Performance measure in our project is accuracy.

Result Analysis -
Our study achieved a good accuracy in Logistic Regression that is 98.57%
followed by Random Forest that is 98.86 and then decision tree that is
99.54% overall. We conclude that if we have a lot of data set than we have
a more accurate prediction of our disease.

Fig: 7.1 Comparative Result Analysis Between Logistic Regression, Random Forest and Decision Tree

34
 CHAPTER 8

CONCLUSION, LIMITATION AND FUTURE WORK

Conclusion

This study underscores the transformative potential of machine learning in heart disease prediction,
providing a reliable and efficient solution for early diagnosis. By employing algorithms like Logistic
Regression, Random Forest, and Decision Trees, the system achieves outstanding accuracy, with Decision
Trees emerging as the most effective model. The research validates the importance of leveraging data-
driven insights to support clinical decisions, enhance patient outcomes, and reduce the burden on
healthcare systems. The adaptability of the system to diverse datasets and its real-time processing
capabilities positions it as a valuable tool in preventive cardiology. Additionally, this work contributes to
the broader application of AI in healthcare, bridging technological innovation and patient-centric care.

Limitations

This study highlights the transformative role of machine learning in predicting heart disease, offering a
reliable and efficient framework for early diagnosis and improved healthcare outcomes. By leveraging
algorithms such as Logistic Regression, Random Forest, and Decision Trees, the system achieves
impressive accuracy, with Decision Trees showing the best performance. The research emphasizes the
importance of data-driven insights to enhance clinical decision-making, reduce patient risk, and alleviate
the burden on healthcare systems. Its ability to process data in real-time and adapt to diverse datasets
makes it a valuable tool for preventive cardiology. Despite its success, the system faces some limitations,
including its dependency on high-quality, labeled datasets, challenges in handling complex or multimodal
data, and scalability constraints under high patient loads. Additionally, its current framework may not
generalize well to underrepresented populations, limiting its universal applicability.

Future developments of the system could address these limitations by integrating multimodal data such
as medical imaging and patient histories, adopting advanced AI techniques like deep learning, and
expanding datasets to include diverse demographic and clinical contexts. Real-time adaptive learning
mechanisms could also enhance the system's ability to update itself based on deployment feedback,
ensuring continuous improvement. Collaboration with healthcare providers and rigorous clinical
validation would further refine its reliability, while leveraging cloud computing could address scalability
challenges. By tackling these areas, the system can evolve into a comprehensive and impactful tool for
proactive heart disease management, contributing significantly to global healthcare advancements.

35
Future Scope

The To enhance the utility and robustness of the heart disease prediction system, future advancements should focus on several key
directions. First, integrating multimodal data, such as medical imaging (e.g., ECGs and MRIs) and patient histories, can provide a richer
context for diagnosis, improving predictive accuracy. Additionally, leveraging advanced AI techniques, including transformer models and
neural networks, could enable the system to analyze complex and unstructured data with greater precision. Expanding datasets to include
diverse populations and regions is essential for ensuring the system’s generalizability and applicability across various demographic and
clinical contexts.

Real-time adaptive learning mechanisms, such as reinforcement learning, could be introduced to dynamically update the model based on
deployment feedback, enhancing its adaptability to new data patterns. Collaboration with healthcare providers for rigorous clinical testing
and validation will ensure that the system adheres to real-world standards of reliability and usability. Lastly, optimizing scalability through
cloud computing and distributed systems will improve the system’s efficiency and effectiveness under high patient loads. These
developments will collectively elevate the system’s potential as a comprehensive tool for heart disease prediction and management,
addressing critical challenges in global healthcare.

36
 REFERENCES

[1] Shetty, S. S., Shreejith, K. B., Deekshitha, K., Dhanusha, K. B., & Gagana, K. B. "Heart Disease Prediction Using Random Forest and
Logistic Regression," International Journal of Information Science, 2023.

[2] Kumar, P., et al. "Comparative Study of Machine Learning Algorithms for Predictive Healthcare Systems," International Journal of
Data Science, 2021.

[3] Smith, D., & Brown, M. "Scalability in Predictive Models for Healthcare," ACM Transactions on Information Systems, 2022.

[4] Wibowo, J. S., Wahyudi, E. N., & Listiyono, H. "Performance of SVM, Naive Bayes, and Neural Networks in Disease Prediction,"
Proceedings of the International Conference on Technology and Information Science, 2023.

[5] Rastogi, S., & Bansal, D. "A Review on Predictive Models in Healthcare: Techniques and Applications," International Journal of
Computer Science & Engineering, 2022.

[6] Zhang, L., et al. "Improving Model Accuracy in Medical Diagnosis Using Hybrid Learning Approaches," Expert Systems with
Applications, 2023.

[7] Nurhasanah, et al. "Incorporating Clinical Data and Metadata in Machine Learning for Disease Prediction," Proceedings of the
International Conference on Computer Science, 2023.

[8] Dataset Resource: "Heart Disease Dataset," Kaggle

37
 CONTRIBUTION OF PROJECT
The Heart Disease Prediction project makes a meaningful contribution to advancing healthcare by leveraging artificial intelligence to
predict heart-related ailments accurately and efficiently. This chapter discusses the project's objectives, expected outcomes, and its broader
societal implications.

Objective of Relevance of the Project

● The project aims to design a system capable of predicting heart disease risk using advanced
machine learning algorithms..
● It enhances diagnostic accuracy, addressing the challenges of manual analysis in large-scale and
diverse patient datasets.
● The system is tailored for deployment in clinics and hospitals, supporting doctors in decision-
making processes.
● By employing dynamic models, the project ensures its utility remains relevant despite the
evolving nature of medical data.

Expected Outcomes

● Enhanced Diagnostic Accuracy:

The system reduces errors, providing dependable results through robust machine
learning techniques.
● Scalable Solution:
It processes extensive datasets in real-time, enabling its use in environments with high
patient volumes.
● Contextual Insights:
By integrating patient history and other metadata, the project delivers comprehensive and
actionable predictions.
● Accessible Interface:
A user-friendly design ensures accessibility for medical professionals and researchers
without deep technical expertise.

Social Relevance

● Improved Public Health:

Early detection of heart disease can significantly reduce mortality rates and healthcare
costs by facilitating timely interventions.
● Support for Healthcare Providers:
The project augments the capabilities of healthcare systems, especially in resource-
limited settings, by offering an additional layer of diagnostic support..

38
● Promoting Preventative Care:
By identifying at-risk individuals proactively, the project encourages preventive
measures and healthier lifestyles.
● Global Impact:
With its potential for adaptation to diverse populations, the project addresses healthcare
disparities and fosters equitable access to medical advancements.

39