International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 10 | Oct 2025
p-ISSN: 2395-0072
www.irjet.net
Heart Disease Prediction Using Hybrid Machine Learning Model Deepika G1, Amsa S2 1Student ,Department Of
MCA ,Jaya College Of Arts and Science, Thiruninravur, Tamilnadu, India
2 Assistant Professor, Department of MCA , Jaya college of arts and Science, Thiruninravur, Tamilnadu, india
---------------------------------------------------------------------***-------------------------------------------------------------------
Abstract - The persistent global challenge of cardiovascular
Our study addresses these limitations through a novel combined approach that brings together the complementary capabilities of Random Forest and XGBoost. We propose that merging RF's ability to manage variability through random feature selection with XGBoost's iterative error correction process will create a superior forecasting system. The primary novel contribution of this work is the empirical demonstration of a synergistic effect achieved by a soft voting ensemble of these algorithms for heart disease prediction. While both algorithms are well-established, their strategic hybridization addresses a key gap in the literature. This approach uniquely leverages RF's robustness to variance through bagging alongside XGBoost's sequential error-correction via boosting, creating a composite model that mitigates the individual limitations of each. This paper details our integrated model design and provides comprehensive testing results that validate its improved performance compared to conventional single-algorithm methods.
diseases (CVDs) underscores an urgent demand for innovative early-warning systems. This investigation designs and validates a hybrid machine learning framework that unites Random Forest (RF) and Extreme Gradient Boosting (XGBoost) for heart disease prognosis. Leveraging a soft voting ensemble, the model processes standard clinical indicators— including patient age, cholesterol, and peak heart rate—to assess risk. The architectural synergy of this hybrid lies in its coupling of RF's variance suppression via bagging with XGBoost's bias minimization through gradient boosting, which collectively fosters a generalized and resilient classifier. Empirical assessment on the UCI Heart Disease dataset reveals a marked superiority of the hybrid model, which attained 91% accuracy. This performance eclipsed that of standalone models: Logistic Regression (83%), Decision Tree (85%), Random Forest (87%), and XGBoost (88%). The evidence positions this RF-XGBoost ensemble as a potent decisionsupport instrument for clinicians, promising to bolster proactive cardiac care.
2. LITERATURE REVIEW
Key Words Heart Disease Prediction, Hybrid Model,
The field of cardiology has seen extensive use of machine learning, with recent work showing a steady progression in predictive modeling. For instance, Lakshmi et al. [1] applied a Random Forest classifier to a heart disease dataset and reached an accuracy of 85%. In another investigation, Ratna Kumari et al. [2] found that Support Vector Machines (SVM) had higher sensitivity than Decision Trees, a valuable trait when the cost of a false negative is high.
Ensemble Learning, Random Forest, XGBoost, Clinical Decision Support.
1. INTRODUCTION Cardiovascular illnesses continue to pose one of the most significant threats to public health worldwide, standing as a top cause of death and creating substantial economic impacts. This pressing reality drives the need for more advanced, trustworthy early detection approaches. In recent years, artificial intelligence techniques, particularly machine learning, have become essential tools in medical diagnostics, showing remarkable ability to identify complex patterns and relationships in patient information that conventional statistical approaches often miss.
Moving into deep learning, Ingole et al. [3] used Long ShortTerm Memory (LSTM) networks to assess heart risk sequentially and reported 87% accuracy. A broad study by Ahmad Hammoud et al. [4] gave a useful side-by-side look at several ML algorithms, like Naive Bayes and K-Nearest Neighbors, for predicting coronary heart disease. A recent review by Patel et al. [5] further consolidated the performances of various individual and ensemble models, confirming the dominance of tree-based methods but also noting the limited exploration of specific hybrid ensembles in clinical settings.
While various predictive models including Logistic Regression, SVM, and basic Decision Trees have been tested for heart disease assessment, these methods commonly face limitations. They may become too specialized to training data or struggle to maintain accuracy across different population groups. Combined model strategies,
A trend seen in much of the prior work is a focus on singlemodel designs. Although these models set important benchmarks, their results can be unpredictable. This very inconsistency is what prompts the need for hybrid or ensemble models.
which integrate multiple algorithms, offer a promising solution by producing more consistent and dependable predictions.
© 2025, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 692