Using Machine Learning and Deep Learning for Enhanced Prediction

and Early Detection of Heart Disease Risk

R. Kamali

, K. Hemalatha

, A. Vaishnavi Dali

, P. Naresh Kumar

A. Gomathi

and N. Aishwarya Rani

Department of Computer Science and Engineering, Knowledge Institute of Technology, Salem, Tamil Nadu, India

Department of Information Technology, Sona College of Technology, Salem, Tamil Nadu, India

Department of Artificial Intelligence and Data Science, Knowledge Institute of Technology, Salem, Tamil Nadu, India

Keywords: Machine Learning, CNNs, LSTM, Deep Learning Techniques.

Abstract: Heart disease is still some the primary causes of mortality worldwide. Proper detection and accurate risk

prediction are critical to effective prevention and therapy. Typical risk evaluation for heart disease models

frequently uses simple statistical methodologies or regression analysis, which might not be able to grasp the

intricate and non-linear interactions between many cardiovascular risk variables. As the difficulty of

healthcare data develops, established methods are becoming unable to provide reliable forecasts. However,

ML and DL techniques have demonstrated considerable promise in dealing with complex data and discovering

detailed patterns that human specialists may ignore. These techniques are mostly helpful for predicting heart

disease because age, heart rate, and levels of cholesterol, and lifestyle decisions all interact in complex,

nonlinear ways. This study investigates how sophisticated ML and DL methods are decision trees, random

forests, neural networks, and cutting-edge algorithms similar CNNs and LSTM networks, might increase

prediction accuracy. The suggested method predicts the likelihood to acquire heart disease using a change of

modern ML and DL approaches. Below, we briefly detail each strategy and how they are used to the prediction

job. Decision trees are a simple but efficient method for machine learning that divides data into subsets

according to feature values, making decision routes simple to see and comprehend. To increase accuracy and

decrease overfitting, random forests, an ensemble technique, construct several integrating the predictions of

decision trees. This approach is effective for predicting cardiac disease since it can handle both continuous

and categorical data.

1 INTRODUCTION

One of the leading causes of death worldwide, heart

disease imposes a major burden on patients and on

health care systems. It encompasses a wide range of

disorders such as heart failure, arrhythmias, and

coronary heart disease which are influenced by

complex interactions between behavioural,

environmental, and genetic factors. Prompt

identification of those when you are at risk for

cardiac events is crucial so you can intervene and

prevent before the disease becomes severe so

drastically limiting its impact on you. Traditional

models predicting risk have been used extensively in

medical research, including logistic regression and

the Framingham Risk Score. While these models

have benefits, they can be constrained by their

inability to capture complex, non-linear relationships

in data. Healthcare must become more precise and

scalable in its predictive systems, as evidenced by

advancements in technology.

Machine learning (ML) and deep learning (DL)

are revolutionizing the domain of healthcare

analytics. These approaches are capable of analysing

large, high-dimensional datasets, discovering hidden

patterns, and providing predictions that are often

more accurate. The use of algorithm techniques like

Random Forests, SVM, CNN, using ML and DL

provide a potential solution for heart disease risk

prediction.

The proposed methodology in this study is to

formulate and implement a hybrid prediction system

based on ML and DL approaches for the prediction of

heart disease. This model is composed to combine the

396

Kamali, R., Hemalatha, K., Dali, A. V., Kumar, P. N., Gomathi, A. and Rani, N. A.

Using Machine Learning and Deep Learning for Enhanced Prediction and Early Detection of Heart Disease Risk.

DOI: 10.5220/0013913900004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 4, pages

396-401

ISBN: 978-989-758-777-1

benefits of prevailing and modern methods so as to

provide clinicians with a robust and accurate decision

support tool. Such findings could have a major

impact on heart disease prediction, enhancing patient

care and health system efficiency.

2 RELATED WORKS

Jian Ping Li, et al., 2020 Cardiac disease is a

complicated illness that affects a huge number of

people globally. In healthcare, particularly in

cardiology, early and precise detection is essential.

An effective machine learning-based method for

diagnosing Cardiac problems is presented in this item

in this regard eliminate insufficient or redundant

features, the system employs methods for choosing

characteristics including Relief, MRS, Lasso, and

Local Learning in accumulation to techniques for

classification such as Support Vector, Logistic

Regression, Neural Networks, K-Nearest

Neighbours, Naïve Bayes, and Decision Trees. In

order to growth precision and decrease execution

time, we also provide a innovative feature variety

method known as FCMIM. The system evaluates the

model and adjusts the hyperparameters using leave-

one-out cross-validation. Classifier performance on

specific features is assessed using performance

measures. According to experimental data, the

FCMIM-SVM system is a good result for the proof of

identity of Cardiac disease in healthcare since it

works well and provides good accuracy.

Tsatsral Amarbayasgalan., et al., 2020 The

primary cause of death is heart disease worldwide,

and its prevalence is increasing. Early detection of

heart Problems before a cardiac event occurs is

challenging. While large amounts of heart disease

data are available in healthcare settings like clinics

and hospitals, this data is often not effectively

analysed to uncover hidden patterns. Machine

learning methods can be beneficial transform this

medical data into useful insights. These techniques

are utilized to create decision support systems (DSS)

that learn and improve from experience. Both

industry and academics are now showing interest in

deep learning. This research aims to accurately

diagnose cardiac disease using a Keras-based DL

methods with a dense neural network. The model is

tested with different configurations of hidden layers,

ranging from 3 to 9 layers, with 100 neurons in each

layer and the ReLU activation function. Various heart

disease Tests are conducted using datasets and both

individual and ensemble models are evaluated. The

model's performance is evaluated using the F-

measure, precision, sensitivity, and efficiency across

all datasets. The output display that the suggested

deep learning techniques outperforms each method

along with other ensemble strategies in terms of

precision, sensitivity, and specificity.

G. Madhukar Rao., et al., 2020 Many lives can be

saved by early detection of heart condition, Among

the primary factors of mortality globally. By

examining huge amounts of medical data to identify

secreted designs using ML can helps in the

recognition of cardiac disorders. This study uses

systems for massive amounts of data, such as Apache

Hadoop to provide A hybrid approach to deep

learning for detecting heart disease. After eliminating

outliers using an enhanced k-means clustering

technique, Using the SMOTE, information is stable.

Recursive feature elimination (RFE) is used to

identify key traits, and an attention-based automated

recurrent unit model and a bio-inspired hybrid

mutation-based swarm intelligence (HMSI) are used

to forecast illness. Four more machine learning

algorithms—naïve Bayes, logistic regression (LR),

K-nearest neighbor (KNN), and sparse autoencoder +

artificial neural network (SAE + ANN) will be used

to match the model. According to the statistics, a

hybrid approach performs better than alternative

methods and closes research gaps with a 95.42%

precision rate.

Santosh Maher., et al., 2020 Because of their

capacity to track heart activity and associated

conditions, a diversity of sensors and devices, such

the Microsoft Band, Apple Watch, and MI HRV

band, have become more and more popular. The poor

survival rate of These days, sudden cardiac death that

happen away from hospitals pose a serious threat to

healthcare. More individuals die from cardiac

conditions each year than from other illnesses

including cardiac attacks and strokes, making it the

world's leading cause of mortality. The WHO

estimates that heart disease claimed 17.9 million lives

in 2016, accounting for 31% of all fatalities

worldwide. Smoking, eating poorly, not exercising,

and taking excessive amounts of alcohol are the

leading causes of cardiac attacks and strokes. Heart

attacks and strokes account for 85% of these fatalities.

Among the primary reasons for shorter lifespans is

cardiac disease. For prompt, precise outcomes, a lot

of people depend on healthcare systems. This paper's

objective is to apply machine learning methods to a

dataset that is regularly gathered by KVK research

labs and healthcare institutions. The study

recommends employing distinctive traits to increase

accuracy in identifying and predict heart illness to

lessen the chance of death.

Using Machine Learning and Deep Learning for Enhanced Prediction and Early Detection of Heart Disease Risk

397

3 EXITING SYSTEM

Traditional systems for cardiac disease prediction

primarily rely on statistical methods, decision trees,

and rule-based algorithms that focus on structured

data such as patient demographics, medical history,

and clinical test results. These systems often use a

limited set of risk factors, such such as heritage,

nicotine intake, lipid levels, and years of age, With

the objective to determine the risk of cardiovascular

disease.

3.1 Disadvantages

• Limited Scope and Inaccuracy

• Inability to Handle Unstructured Data

• Over-Simplification of Risk Factors

• Lack of Real-Time Prediction

• Scalability Issues

• Lack of Personalization

4 PROPOSED SYSTEM

The future system leverages modern ML and DL

methods to increase the precision and effectiveness of

heart disease risk assessment. This approach is

designed to report the restrictions of traditional

systems, which often rely on simplistic models and a

narrow range of input features. The goal line of the

proposed system is to deliver a more precise,

dynamic, and Individualized evaluation of cardiac

risk determined by combining unstructured data from

digital health records (DHRs) with organized clinical

information.

5 PROBLEM DESCRIBTION

Data Integration and Preprocessing: The system

integrates structured (eg ECG readings, age, heart rate

and cholesterol levels) and unstructured data (eg,

physician notes, medical history). Unstructured

clinical works are processed with advanced Natural

Language Processing (NLP) techniques to extract

relevant information. This enables the system to

consider a broader range of features, enhancing the

accuracy of the predictions. G. Madhukar Rao., et al.,

2020 Hybrid ML and DL Model: The system uses the

hybrid approach which combines traditional machine

learning with the latest deep learning architectures.

This hybrid approach is believed to seize both linear

and non-linear connections of characteristics

producing more correct predictions.

Real-Time Risk Reports: The system can analyze

real-time patient data, which can be constantly

collected from body-worn technology or other up to

date health records, enabling timely interventions and

personalized health monitoring, allowing healthcare

providers to make preemptive actions based on the

individual patient's current state of risk. Cross-

Validation and Performance Optimization: To

increase the robustness of the model and its ability to

generalize To do this, cross-validation techniques are

implemented to ensure the model generalizes well

across different data and patient characteristics.

Hyperparameter optimization methods like grid

analysis and random optimization are employed to

tune the model for ideal performance.

Evaluation Metrics Basic metrics like precision,

recall, F1-score, and AUC-ROC (Area Under the

Receiver Operating Characteristic Curve) which are

used to validate the performance of the system. These

metrics demonstrate the improved precision for

prediction when compared to conventional cardiac

risk prediction methods.

Implementation of a User-Friendly Interface: The

system should use a user-friendly interface that

allows medical professionals to seamlessly input

patient data and view the risk prediction outputs and

facilitate decision making. Graphical representation

of risk factors and prediction outcomes can be

accomplished with visualization tools such as

Tableau or Power BI and provide a means to help

clinicians understand the reasoning behind the model.

Ongoing Retraining with New Data: There is a

feedback mechanism built into the system to allow for

new data and findings to be used to retrain and inform

the model to keep it updated as newer research on

heart disease is conducted and the patient population

becomes more diverse.

6 RESULT

The results of the investigation for predicting the risk

level of heart attack using machine learning and deep

learning techniques are displayed in this section. The

results demonstrate how well the proposed methods

predict risk categories for heart disease given

structured patient data and medical images.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

398

6.1 Machine Learning-Based Risk

Assessment

Table 1 gives the structured patient data included sex,

number of blood vessels, type of thrombotic, angina

precipitated by physical activity, blood pressure,

cholesterol, maximal heart rate, depression of ST,

slope of ST, type of chest pain. The trained model

generated predictions of the risk levels for heart

disease using these features.

Key Results

• Status: Success

• Predicted Value: 37.86

• Risk Level: High Risk

Table 1: Machine learning prediction results.

Feature Value

Sex Male (1)

Chest Pain Type 0 (Asymptomatic)

Resting Blood Pressure 125 mmHg

Cholesterol 258 mg/dL

Maximum Heart Rate 141 bpm

Exercise-Induced Angina Yes (1)

ST Depression 2.8

ST Slope 1 (Upsloping)

Major Vessels 1

Thalassemia Type 3 (Reversible Defect)

Table 2 gives the model predicted a significant

chance of cardiac disease due to the patient’s input

values. Factors such as high decrease maximal

heartbeat and cholesterol, and ST depression

contributed significantly to the risk prediction. The

machine learning model effectively classified the

heart disease risk with a rapid response time, making

it suitable for early screening. Performance indicators

like precision, recollection, and F1-score They were

utilized to evaluate the precision of the model.

Table 2: Machine learning model performance.

Metric Value

Accuracy 91.2%

Precision 89.4%

Recall 90.1%

F1-Score 89.7%

The model's high success rate shows how reliable

it is at identifying the risk of cardiovascular illness.

However, its performance could be further improved

by incorporating a larger dataset and powerful

selection of features methods.

6.2 Deep Learning-Based Image

Classification for Heart Disease

In the deep learning approach, a medical image

(angiography or echocardiogram) was provided as

input, and a CNN-based model analysed it for signs

of stenosis (artery narrowing).

Key Results

• Inference ID: c5610a62-3e41-476d-a443-

d6e42fe011ef

• Processing Time: 0.36 seconds

• Image Dimensions: 512 x 512 px

• Predicted Class: Stenosis

• Confidence Score: 75.82%

• Bounding Box Coordinates:

• X: 267.5

• Y: 176

• Width: 27

• Height: 22

6.3 Discussion

The deep learning model detected stenosis with a

confidence of 75.82% showing potential arterial

stenosis. Next, the bounding box is there to highlight

the region detected, so doctors can focus their

attention on areas of concern.

The deep learning approach has the following

benefits over traditional manual diagnosis:

• Performances:

The latency was 0.36

seconds which granted on-the-fly decisions.

• Precision: The model reached a strong

(75.82%) co-efficiency of detected

confidence, with more data

inferences,

detection will receive a better indicative.

• Automating Detection: System

overlays

affected areas with visual bounding box,

helps radiologists in diagnosis

However, limitations include potential false

positives and reliance on high-quality images. In the

future, there are plans to train the model with a greater

range of datasets and improve precision by using

more sophisticated augmentation methods.

6.4 Comparative Analysis of ML and

DL Approaches

Table 3 gives the information of the following:

The machine learning techniques is well-suited

for structured data analysis and provides an instant

risk level assessment.

Using Machine Learning and Deep Learning for Enhanced Prediction and Early Detection of Heart Disease Risk

399

Table 3. Performance Comparison.

Method Input Type

Processing

Time

Prediction Output Confidence Best Use Case

Machine

Learning

Structured

Data

~0.2s Risk Level: High 91.2%

Accuracy Risk

Prediction

Deep

Learning

Medical

Images

0.36s Stenosis Detection 75.82%

Confidence

Visual

Diagnosis

Discussion

The deep learning model is highly effective in

image-based diagnosis, identifying heart

abnormalities with bounding box visualizations.

Combining the two strategies could result in to a

more comprehensive heart disease assessment

system, integrating clinical parameters with medical

imaging insights.

7 CONCLUSIONS

The results demonstrate that both ML and DL

approaches provide valuable data about the risk of

cardiac disease assessment. Machine learning excels

in structured data-based predictions, while deep

learning is effective in image-based diagnosis. Future

work should focus on integrating both models into a

hybrid system to improve overall predictive accuracy

and clinical applicability.

8 FUTURE DISCUSSION

To increase forecast accuracy, use slower data. Use

feature engineering strategies to improve the

weighting of risk factors. To increase confidence

levels, train the model on a bigger dataset. To

distinguish between several cardiac diseases other

than stenosis, use a multiple-class system.

REFERENCES

Essam H. Houssein, Rehab E. Mohamed, and Abdelmgeid

A. Ali, "Heart Disease Risk Factor Detection from

Electronic Health Records Using Advanced NLP and

Deep Learning Techniques," IEEE Journal of

Biomedical and Health Informatics, vol. 24, no. 11, pp.

3201-3209, 2020, doi: 10.1109/JBHI.2020.3003928.

G. Madhukar Rao, Dharavath Ramesh, Vandana Sharma,

Anurag Sinha, Md. Mehedi Hassan, and Amir H.

Gandomi, "AttGRU-HMSI: Enhancing Heart Disease

Diagnosis Using Hybrid Deep Learning Approach,"

IEEE Transactions on Biomedical Engineering, vol. 67,

no. 5, pp. 1234-1245, 2020, doi:

10.1109/TBME.2019.2953456.

J. Deepika, J. Akilandeswari , Lokeshwaran B ,

Mukilarasan N, Naveen S, Prabhu A , “An Intelligent

HealthCare System Using Blockchain Techniques”, 7th

International Conference on Electronics,

Communication and Aerospace Technology ISBN:979-

8-3503-4060-0 https://doi.org/10.1109/ICECA

58529.2023.10 395852 , pp 538 – 545, Feb-

2024[SCOPUS INDEXED CONFERENCE]

Jian Ping Li, Amin Ul Haq, Salah Ud Din, Jalaluddin Khan,

Asif Khan, and Abdus Saboor, "Heart Disease

Identification Method Using Machine Learning

Classification in E-Healthcare," Journal of Medical

Systems, 2020, pp. 1-10, doi: 10.1007/s10916-020-

01543-7.

R. Saravana Ram, J. Akilandeswari, M. Vinoth Kumar,

“HybDeepNet: A Hybrid Deep Learning Model for

Detecting Cardiac Arrhythmia from ECG Signals”

Information Technology and Control, Volume 52, Issue

No. 2 ISSN 1392-124X https://doi.org/1 0.5755/j01.i

tc.52.2.32993, Jul-2023[WOS-SCIE]

Rohit Bharti, Aditya Khamparia, Mohammad Shabaz,

Gaurav Dhiman, Sagar Pande, and Parneet Singh,

"Prediction of Heart Disease Using a Combination of

Machine Learning and Deep Learning," Proceedings of

the International Conference on Computing,

Communication, and Intelligent Systems (ICCCIS

2020), pp. 320-328, 2020, doi:

10.1109/ICCCIS49312.2020.00065.

Rohit Bharti, Aditya Khamparia, Mohammad Shabaz,

Gaurav Dhiman, Sagar Pande, and Parneet Singh,

"Heart Disease Risk Detection Using Machine

Learning Models: A Comprehensive Survey,"

International Journal of Computer Science and

Information Security (IJCSIS), vol. 18, no. 9, pp. 234-

242, 2020, doi: 10.13140/RG.2.2.24356.00641.

S. David Samuel Azariya, V. Mohanraj, G. Jothi ,J. Jeba

Emilyn J. Akilandeswari, “Cardiac Image Analysis for

Accurate Heart Disease Diagnosis Using Deep

Learning Techniques: A Comprehensive Review”

Lecture Notes in Networks and Systems, Volume 864

Print ISBN 978-981-99 -8627-9 Online ISBN 9 78 -

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

400

981-99-8628-6 https://doi.org/10.1007/978-981-99-

8628-6_5 , Apr-2024 [SCOPUS INDEXED BOOK

CHAPTER]

Sakthipriya, G, Dr.Y.Suresh, Y Varnisha, C, Sindhu, R. ,

Shivraj, R, “ Analysis of Heart Disease Prediction

Using Various Machine Learning Algorithms” Lecture

Notes in Electrical Engineering, Volume 1095, Online

ISBN 978-981-99-7077-3, Print ISBN 978-981-99-

7076-6 https://doi.org/10.1007/978-981-99-7077-3_12

Jan-2024[SCOPUS INDEXED BOOK CHAPTER]

Santosh Maher, Shaikh Abdul Hannan, Sumegh Tharewal,

and K.V. Kale, "HRV-Based Human Heart Disease

Prediction and Classification Using Machine

Learning," International Journal of Computational

Biology and Drug Design, vol. 14, no. 6, pp. 654-667,

2020, doi: 10.1504/IJCBDD.2020.1000352

Tsatsral Amarbayasgalan, Van-Huy Pham, Nipon Theera-

Umporn, Yongjun Piao, and Keun Ho Ryu, "An

Efficient Prediction Method for Coronary Heart

Disease Risk Based on Two Deep Neural Networks

Trained on Well-Ordered Training Datasets," IEEE

Transactions on Neural Networks and Learning

Systems, vol. 31, no. 8, pp. 1-15, 2020, doi:

10.1109/TNNLS.2020.2998765

Tsatsral Amarbayasgalan, Van-Huy Pham, Nipon Theera-

Umporn, Yongjun Piao, and Keun Ho Ryu, "An

Efficient Prediction Method for Coronary Heart

Disease Risk Based on Two Deep Neural Networks

Trained on Well-Ordered Training Datasets," IEEE

Transactions on Neural Networks and Learning

Systems, vol. 31, no. 8, pp. 1-15, 2020, doi:

10.1109/TNNLS.2020.2998765.

Velusamy A , J Akilandeswari, S Vaanathi, M Priya, T

Munirathinam, S Kokila, “Enhancing Myocardial

Infarction Prediction with Machine Learning:

Leveraging Clinical and Demographic Data for

Improved Early Detection and Patient Outcomes”,

OPJU International Technology Conference (OTCON)

on Smart Computing for Innovation and Advancement

in Industry 4.0, https://doi.org/10.1109/OTCON60325

.2024.10687656, IEEE Explore , ISBN: 979-8-3503-

7379-0, Sep-2024[SCOPUS INDEXED

CONFERENCE]

Using Machine Learning and Deep Learning for Enhanced Prediction and Early Detection of Heart Disease Risk

401