A Comprehensive Review of Machine Learning Algorithms in Predicting Cardiovascular Diseases

Mahnoor Kashif
Freelance Writer, Jumma Khalsa Branch, Near Khurshid, Pakistan
Correspondence to: Mahnoor Kashif, mahnoorkashif92@gmail.com

DOI: https://doi.org/10.70389/PJC.100015

Additional information

• Ethical approval: N/a
• Consent: N/a
• Funding: No industry funding
• Conflicts of interest: N/a
• Author contribution: Mahnoor Kashif – Conceptualization, Writing – original draft, review and editing
• Guarantor: Mahnoor Kashif
• Provenance and peer-review: Unsolicited and externally peer-reviewed
• Data availability statement: N/a

Keywords: Random forest, XGBoost, Graph neural networks, Smote resampling, Federated learning.

Peer Review
Received: 25 July 2025
Accepted: 16 December 2025
Version accepted: 1
Published: 22 January 2026

Plain Language Summary Infographic

Introduction

Cardiovascular diseases (CVDs) are the reason why millions of lives are lost each year and they continue to be the number one cause of death globally.¹ Even after years of progress in both prevention and treatment of CVDs, early detection remains a major challenge. This is specifically evident in cases where symptoms develop silently or overlap with other conditions. These conditions can range from coronary artery disease and heart failure to arrhythmias or stroke. Many traditional diagnostic tools rely on predefined thresholds or expert judgment, which, while useful, can sometimes miss subtle warning signs or produce delayed diagnoses.² Figure 1 shows that ischemic heart disease and stroke alone accounted for over 15 million deaths in 2019, highlighting the urgent need for more effective strategies in diagnosis and prevention.

In the last decade, however, a growing number of researchers and clinicians have started adopting machine learning algorithms as a way to improve cardiovascular risk prediction. The ML approach has a unique ability to find patterns in large datasets that may not be obvious even to experienced physicians. ML models can evaluate a wide range of information in context of CVDs. This information includes standard lab results, ECG readings, and even data from wearable health monitors.³ With the help of appropriate data and ML integration, health practitioners can help identify people at elevated risk of cardiovascular issues long before serious symptoms appear.⁴

One predominant factor driving growing interest in applying machine learning approaches to cardiovascular care is its flexibility in handling different types of patient data. The ML models can sift through lab results, imaging, and clinical records to uncover patterns that help tailor treatment plans. Research comparing algorithms like logistic regression, support vector machines, and deep neural networks shows that they can predict heart disease risk with strong and reliable accuracy.^5,6 A recent study involving over 143,000 hypertensive patients found that ensemble models combining Random Forest, XGBoost, and deep learning outperformed traditional logistic regression in predicting cardiovascular risk.⁷ In many cases, these machine learning methods even exceed traditional statistical tools, particularly when feature selection is done thoughtfully. However, many prediction models come under scrutiny in terms of model interpretability by clinicians, dealing with incomplete or inconsistent data, and maintaining their effectiveness outside of controlled environments.⁸ For these advanced tools to facilitate healthcare providers and patients, they must not only be technically sound but also ethically responsible and easy to incorporate into routine clinical workflows.

Background

Overview of Cardiovascular Diseases

Cardiovascular diseases (CVDs) represent a diverse group of heart and blood vessel disorders which are difficult to be detected in early stages because they often progress silently over time. While the symptoms can vary greatly depending on the specific condition, the underlying mechanisms frequently involves long-term damage to the vascular or cardiac system caused by a combination of lifestyle, genetic, and metabolic factors.⁹

Coronary artery disease (CAD) is the most well-known and widely studied form of CVDs which results from the buildup of plaque in the arteries that supply blood to heart. With time, this buildup can lead to restricted blood flow, angina, or myocardial infarction. Cardiac arrhythmias, such as atrial fibrillation basically accounts for the abnormal electrical activity in heart. They are associated with increased risks of stroke and heart failure. Other major types include congestive heart failure, valvular diseases, congenital heart defects, and peripheral arterial disease, all of which contribute to the broader burden of CVDs across different populations.¹⁰ In fact, coronary heart disease alone accounted for 43% of all cardiovascular-related deaths in the U.S. in 2016, making it the most fatal form by far. Stroke followed at 17%, with heart failure and high blood pressure contributing 9% and 10% respectively, while other CVD types made up 21% of cases (See Figure 2).

Cardiovascular disease typically arises from a complex interplay of multiple factors rather than a single cause. While some of these influences can be managed or altered, others are beyond individual control. Behaviors such as poor dietary habits, tobacco use, sedentary lifestyle, hypertension, and high cholesterol significantly increase the risk. Beyond lifestyle influences, certain inherent characteristics, such as one’s genetic makeup, age, and sex also play a critical role in shaping cardiovascular risk.¹¹ In parallel, there is growing recognition of the role that biological processes like chronic inflammation, insulin resistance, and metabolic dysfunction have in initiating and advancing cardiovascular disease.¹²

Early diagnosis of cardiovascular conditions remains a major clinical challenge. Many forms of heart disease progress silently for years, and when symptoms do arise, they are often vague, overlap with other illnesses, or go unrecognized particularly in women, who may present differently than men. In under-developed countries and regions with scarce healthcare resources, limited access to diagnostic services is the major reason that delays early detection of cardiovascular diseases.¹³ Cardiovascular diseases are complex and often develop silently due to which conventional screening methods don’t always succeed in identifying them early. That is why more recent approaches like machine learning and artificial intelligence are being explored by researchers to improve how risk is assessed and diagnosis is made. Some studies now focus on methods that combine a broad range of patient information including lab tests, imaging, family history, and lifestyle factors to reveal hidden patterns of risk. Although these strategies are still evolving, they can be considered as a roadway leading towards early detection of heart diseases for enabling timely medical intervention.¹²

Machine Learning in Healthcare

In recent years, machine learning, a branch of artificial intelligence, has begun carving out a meaningful space in healthcare. What makes it so promising is its ability to make sense of large, often complex datasets that might overwhelm traditional tools. Particularly in the field of cardiovascular care, the growing power of computational systems has helped shape newer and more targeted treatment strategies for conditions ranging from arrhythmias to coronary artery disease.¹⁴ Rather than depending solely on fixed equations or assumptions, these models pull from diverse sources, such as electronic health records, lab reports, or even imaging scans, helping clinicians spot issues earlier and with greater accuracy.¹⁵ While it has certain limitations, the shift toward data-informed diagnosis is clearly advancing.

Machine learning techniques are further categorized into four classifications; one type is supervised learning where algorithms are given certain inputs and outputs, and the aim is to map an input to output. This includes identification of an image, handwriting recognition, or electrocardiogram interpretation and using the labeled data obtained from them to train the model for making predictions.¹⁶ On the other hand, unsupervised learning groups people based on shared features, which is useful for identifying patterns or clusters in patient populations. Furthermore, deep learning is a more advanced method that uses multiple layers of processing, similar to how the brain works. In case of diagnosis, it has been effective for interpreting ECGs, analyzing scans, and reviewing clinical records.¹⁷ Lastly, reinforcement learning takes a different approach by learning through feedback as it updates decisions over time and is the subject of ongoing studies for its potential use in long-term treatment planning. In cardiovascular care, machine learning supports more than just early detection by identifying high-risk individuals even before symptoms appear.¹⁸ Table 1 compares the machine learning techniques to summarize their clinical applications.

Table 1: Key machine learning types and healthcare applications.
Type	Core Idea	Data Type	Use in Healthcare	Example
Supervised	Learns from labeled data	Labeled (X, Y)	Diagnosis, prediction	Heart disease classification
Unsupervised	Finds hidden patterns	Unlabeled	Clustering, pattern discovery	Risk-based patient grouping
Reinforcement	Learns via feedback	Reward-based actions	Treatment optimization	Drug dosing strategy
Deep Learning	Neural networks, complex	Large/raw structured & unstructured	Signal & image analysis	ECG/MRI interpretation

Datasets and Preprocessing for Cardiovascular Disease Prediction

Many ML models for cardiovascular disease prediction depend upon publicly available datasets which offer structured and clinically relevant information. One of the most widely used datasets is the Framingham Heart Study (FHS) dataset. It includes decades of longitudinal data on risk factors such as blood pressure, cholesterol, diabetes, and lifestyle habits. FHS has played a foundational role in predictive modeling for heart disease, especially for estimating 10-year cardiovascular risk.¹⁹ Another notable resource is the UCI Heart Disease dataset which contains a relatively small but highly informative sample of patients (e.g. variables like age, sex, chest pain type, and ECG results). Although this dataset has a limited size, it is still frequently used in classification and detection of heart-related issues because of its clean structure and relevance to clinical tasks.²⁰ For more complex and critically ill patient cases, the MIMIC-III dataset offers ICU data from over 40,000 patients, comprising of lab reports, detailed time-series measurements, medications, and diagnostic codes. Hence, the MIMIC-III dataset is well suited for identifying potential complications or estimating mortality risk in patients with severe or rapidly changing conditions.²¹

Prior to developing a predictive model, the dataset needs careful preparation to avoid bigger issues down the line. The first step is data cleaning which involves checking for consistency, correcting odd outliers, and dealing with any missing values which is followed by choosing the right features.²² By picking only the most relevant inputs, the model avoids unnecessary noise and focuses on valuable information. Then, it’s important to scale or normalize different measurements so that blood pressure, cholesterol, heart rate, and other variables don’t disproportionately influence the model, simply due to their differing scales. Proper feature scaling not only speeds up model training but also helps prevent certain features from dominating the results, hence laying a solid foundation for the predictive model.²³ In many heart-related datasets, there’s a bigger problem; only a small number of patients actually experience events like heart attacks or strokes, so the data can be off-balanced. Such issues can be handled by using techniques like SMOTE or under-sampling which help the model learn from both common and rare cases, without leaning too heavily towards one.²⁴ 

Overview of Applied ML Algorithms

Classical Machine Learning Algorithms

Some classical machine learning algorithms are logistic regression, decision tree, Random Forest, support vector machines, k-nearest neighbors, and naive bayes. These algorithms demonstrate strong potential in cardiovascular disease prediction due to their efficiency and ease of deployment. Literature reveals that one of the most widely used methods is Logistic Regression (LR), suitable for binary classification tasks, such as predicting the presence or absence of cardiovascular diseases. Furthermore, its simplicity and interpretability make LR algorithms an apt choice in clinical environments. B Patil (2020) put forward in a study that logistic regression outperformed other algorithms in terms of predictive accuracy when applied to heart disease datasets, offering a reliable method to detect patients at risk.²⁵ However, its performance may suffer in cases of non-linear relationships among variables.

Decision tree, and their aggregated version Random Forest, are another type of commonly used algorithms. These models are preferred because of their ability to handle both categorical and numerical data. They also provide useful insights into feature importance which is a valuable characteristic for clinicians assessing risk factors. Although decision tree can be prone to overfitting, Random Forest algorithms mitigate this by aggregating predictions from multiple trees. Studies suggest that Random Forest are more accurate than individual decision tree and maintain robustness even with noisy or incomplete data.²⁶ Additionally, Khan et al. estimated the performance of the proposed ML algorithm using numerous conditions to recognize the best suitable machine learning algorithm in the class of models. Their findings revealed that the RF algorithm had the highest accuracy of prediction, sensitivity, and recursive operative characteristic curve of 85.01%, 92.11%, and 87.73%, respectively, for CVD. It also had the least specificity and misclassification errors of 43.48% and 8.70%, respectively, for CVD. These results indicated that the RF algorithm is the most appropriate algorithm for CVD classification and prediction.²⁷

Support Vector Machine (SVM) technique provides an alternative approach through boundary-based classification. SVM performs well in high-dimensional spaces and have been successfully applied in cardiovascular disease prediction, especially when kernel functions are used to manage non-linearly separable data. SVM algorithms are more effective when a clear margin of separation exists between classes. However, they require careful tuning and are demographically intensive with large datasets.²⁶ In a comparative study, Support Vector Machine approach achieved higher accuracy than other prediction algorithms such as Naive Bayes in classifying heart-related outcomes, using the UCI dataset.²⁸ The K-Nearest Neighbors (KNN) algorithm operates on distance metrics to classify new data based on closeness to existing labeled samples. It is a simple algorithm, but performance degrades with large or high-dimensional datasets due to increased computation and sensitivity to irrelevant features. B Patil (2020) observed in his comparative study that KNN models outperformed decision tree and Naive Bayes in cardiovascular disease detection accuracy but fell short compared to neural networks.²⁵

Adding further, Naive Bayes (NB) is a simple algorithm that makes predictions based on probabilities and assumes that each feature works independently from the others. Despite its simplicity, NB can yield strong results, particularly when the independence assumption is valid. It performs well with small datasets and is relatively robust to noisy data. However, it may struggle with continuous variables unless the data is properly preprocessed. In his comparative analysis of classical ML algorithms, B Patil proposed that Gaussian NB achieved an accuracy of 58.3%, which was lower than other models such as neural networks and KNN, but it remained competitive due to its computational efficiency.²⁵ Figure 3 provides a graphical representation of common classical machine learning algorithms used in disease prediction, including both linear and non-linear models, highlighting how each algorithms approaches data classification or clustering.

Deep Learning Models

Deep learning approaches have increasingly gained attention among healthcare practitioners, especially when dealing with large datasets and complex patterns. The concept of deep learning was first introduced in 1943, by Walter Pitts and Warren McCullough, which is an advanced form of the neural network method in artificial intelligence. Artificial Neural Networks (ANNs) are commonly used for tabular medical data, comprising of connected units called nodes, that function like brain cells.¹⁴ Each node takes in information, processes it, and then passes the results forward. Such ANNs enable the modeling of non-linear relationships that classical machine learning algorithms can’t handle efficiently. A pioneer study conducted by Das et al. (2008) employed neural networks and reported a classification accuracy of 89.01% when tested on data from the Cleveland Heart Disease database. The findings further revealed that the model achieved a sensitivity of 80.95% and a specificity of 95.91%, indicating strong performance in identifying and diagnosing heart disease.²⁹

On the other hand, Convolutional Neural Networks (CNNs) are particularly effective for analyzing image-based or spatial data and are often referred as a type of ANN approach that learns by different methods such as back propagation.³⁰ In cardiovascular diagnostics, CNNs have been applied to electrocardiogram images to detect anomalies such as arrhythmias or myocardial infarctions. This method outperforms traditional methods in tasks such as ECG classification.³¹

Furthermore, Recurrent Neural Networks (RNNs) and Long Short-Term Memory networks (LSTMs) are well-suited for time-series data, monitoring continuous heart rate or ECG signals and giving results based on this data. These models are capable of retaining temporal dependencies, which is crucial for understanding trends in patient vitals. Applications of LSTMs in heart disease prediction have shown success in capturing signals that point towards heart-related issues, improving both the precision and recall of prediction systems. The work of Alrashdi and Taloba (2025) introduced a novel approach to predicting heart failure (HF), integrating graph neural networks (GNNs) with long short-term memory (LSTM) networks for better prediction accuracy. This hybrid model, GNN-LSTM, applied the advantages of both networks i.e. the complex interdependencies between clinical variables capture clinical relationships and LSTMs can better manage temporal dependencies. The model was tested on a large, highly representative dataset containing diversified clinical variables from HF patients, with 98.9 % predictive accuracy, which outperforms the single models as well as their respective performances by conventional methods like CNN, SMOTE, LSTM-RNN, CNN-LSTM, CNN-GRU, and traditional GNN approaches.³²

Hybrid Models

Hybrid models integrate the strengths of multiple learning algorithms to improve predictive accuracy and robustness. Among these, boosting techniques like XGBoost and AdaBoost have demonstrated an exceptional performance due to their ability to handle imbalanced datasets. These methods focus on correcting the mistakes of weak classifiers by sequentially adjusting weights, resulting in improved accuracy and generalization. A study carried out by Ogunpola et al. (2024) emphasized the effectiveness of meticulously fine-tuning an XGBoost model for cardiovascular diseases by demonstrating that this optimization yields remarkable results with 98.50% accuracy, 99.14% precision, 98.29% recall, and a 98.71% F1 score. The literature findings clearly indicate that this optimization to hybrid approaches significantly enhances the model’s diagnostic accuracy for heart disease.⁵

Recently many researchers have explored that hybrid models that integrate neural networks with classical algorithms (e.g. ANN + Decision tree or ANN + Naive Bayes) leverage the strengths of both paradigms. A latest finding of Navita et al. showed that a hybrid stacking model combining Random Forest, K-Nearest Neighbors, AdaBoost, and Logistic Regression as a meta-learner outperformed traditional methods in detecting cardiovascular disease. The model proposed by Navita et al. (2025) was evaluated using tenfold cross-validation, and key metrics such as accuracy, sensitivity, specificity, F1 score, and ROC-AUC were reported. Without SMOTE–ENN resampling and Chi-square feature selection, the accuracy was 94.74% but after applying both techniques, it escalated to 97.8%, with 96.15% sensitivity, 96.75% specificity, and a ROC-AUC of 98.6%, highlighting the significant role of hybrid sampling and feature selection in boosting model performance.³³

Overall, while classical algorithms remain vital due to their interpretability and efficiency, deep learning and hybrid models result in predictive accuracy, especially for complex and high-dimensional cardiovascular datasets. The efficacy and preference of algorithm should therefore align with data characteristics, clinical requirements, and resource constraints to maximize both performance and trustworthiness.

Evaluation Metrics

Evaluating a machine learning model is mandatory to determine the model’s suitability for practical deployment.  It determines how well the model performs not just on training data, but also on new or unseen cases. In healthcare, where decisions impact lives, this evaluation must be thorough and meaningful. A useful starting point is the confusion matrix that breaks predictions into four categories, known as true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN). Using these categories, accuracy, which measures the proportion of correct predictions, is calculated. However, in clinical prediction tasks that involve imbalanced data, precision and recall are more commonly used. Precision basically shows how many predicted positives are actually correct. On the other hand, recall tells how well the model identifies real cases. Then comes the F1-score that combines both precision and recall into a single value and is useful when there’s a trade-off between the two. To make the results more reliable, cross-validation is used to test performance across different data splits. To refine model behavior, hyperparameter tuning can be used as an evaluation metric. Together, these metrics offer a balanced and clinically meaningful evaluation framework. The table below shows some common evaluation metrics and how they are calculated.⁵

Table 2 presents key metrics used to evaluate classification models. Accuracy reflects the overall correctness by measuring the proportion of true predictions (both positive and negative) out of all predictions. Precision shows how many of the predicted positive cases are actually positive, while recall (or sensitivity) measures the model’s ability to correctly identify actual positive cases. The F1-score provides a balanced measure by combining precision and recall. The variables used are: TP, TN, FP, and FN which represent the outcomes of a model’s predictions compared to actual results.

Table 2: Common evaluation metrics and their formulas.
Metric	Purpose	Formula
Accuracy	Overall correctness of the model	(TP + TN) / (TP + FP + TN + FN)
Precision	How many predicted positives are truly positive	TP / (TP + FP)
Recall (Sensitivity)	Ability to detect actual positives	TP / (TP + FN)
F1-Score	Balance between precision and recall	2 × (Precision × Recall) / (Precision + Recall)

Limitations

Although ML models have demonstrated some success in predicting cardiovascular disease, there is still room for improvement and further development is required before they can be applied in clinical setting without any reservations. A major challenge is the lack of consistency in approaches adopted by researchers as there is no agreed method for how to split data, choose features, or tune models, making it hard to compare studies or draw clear conclusions. The inadequate documentation of hyperparameters and other technicalities also leads to inconsistency.³⁴ Furthermore, many models are trained on small or uneven datasets, a problem that is particularly evident in certain patient cohorts, leading to unfair or unreliable predictions.

In many cases, important performance measures like balanced accuracy are not mentioned which misleads the readers, resulting in biased or inaccurate predictions. The pooled results across studies could also potentially be biased due to small sample size for CVDs. Overfitting is another undesirable behavior of ML models which limits their reliability. It refers to models that perform great during training, but often struggle with new data. And beyond the technical stuff, there are bigger concerns, such as patient privacy, legal uncertainties, and the risk of bias. These concerns not only raise doubts about the readiness of these tools for predicting cardiovascular diseases, but also challenge the integration of AI into real-world healthcare settings.

Future Recommendations

Application of machine learning in cardiovascular care holds a lot of potential, however ensuring its effectiveness in real-world clinical environment demands careful planning and strategic efforts. A critical focus area is the development of explainable AI tools that clarify how models generate their predictions. When clinicians can comprehend the rationale behind an AI-driven outcome, they are more inclined to trust and incorporate these tools into patient care.³⁵ Scientists are working on the growing use of wearable devices, like smartwatches, which can track heart rate or rhythm in real time, even outside clinical settings. This continuous data collection can give early warnings for potential risks and further developments must be made in its application.  Privacy, of course, remains a huge concern, but researchers emphasize integrating multimodal data and federated learning as essential next steps for personalized and secure cardiovascular AI systems.³⁶

Federated learning allows hospitals to train models collaboratively without ever sharing raw patient data, whereas transfer learning is a technique that makes it easier to apply one hospital’s model to another setting without needing to start from scratch. Additionally, integrating various types of data such as genetics, medical scans, and routine lab results, has the potential to significantly enhance the accuracy of predictions and early diagnosis and treatment of CVDs. Researchers, practitioners and healthcare departments must collaborate and adopt strategic approaches to transform these promising ideas into everyday tools that physicians can apply in clinical settings.

Conclusion

While significant developments are still required to overcome the limitations in implementing ML algorithms in clinical practice, their overall performance has demonstrated promising output. Random Forest (RF), Hybrid algorithms, and Artificial Neural Networks (ANNs) are widely used in CVD prediction and have successfully exposed risk patterns that might otherwise go unnoticed by conventional approaches. However, to ensure accurate interpretation within the appropriate clinical context, it is essential to select the most suitable algorithm, comprehensively report evaluation metrics, rigorously validate the models, and compare their performance against that of human experts. Once thoroughly validated, these algorithms can be seamlessly integrated into electronic health record systems and applied in clinical practice, particularly within well-resourced healthcare settings. If implemented responsibly, ML integration could support more personalized and proactive cardiovascular care in the near future.

References

1. World Health Organization: WHO. Cardiovascular diseases [Internet]. 2019. Available from: https://www.who.int/health-topics/cardiovascular-diseases#tab=tab_1
2. Sanchez-Martinez S, Camara O, Piella G, Cikes M, González-Ballester MÁ, Miron M, et al. Machine Learning for Clinical Decision-Making: Challenges and Opportunities in Cardiovascular Imaging. Frontiers in Cardiovascular Medicine [Internet]. 2022;8. Available from: https://www.frontiersin.org/journals/cardiovascular-medicine/articles/10.3389/fcvm.2021.765693/full
3. Neri L, Oberdier MT, Van Abeelen KCJ, Menghini L, Tumarkin E, Tripathi H, et al. Electrocardiogram Monitoring Wearable Devices and Artificial-Intelligence-Enabled Diagnostic Capabilities: A Review. Sensors [Internet]. 2023;23(10):4805. Available from: https://www.mdpi.com/1424-8220/23/10/4805
4. Jurgens CY, Lee CS, Aycock DM, Creber RM, Denfeld QE, DeVon HA, et al. State of the Science: The Relevance of Symptoms in Cardiovascular Disease and Research: A scientific Statement from the American Heart Association. Circulation [Internet]. 2022;146(12). Available from: https://doi.org/10.1161/cir.0000000000001089
5. Ogunpola A, Saeed F, Basurra S, Albarrak AM, Qasem SN. Machine Learning-Based Predictive Models for detection of cardiovascular diseases. Diagnostics [Internet]. 2024 ;14(2):144. Available from: https://doi.org/10.3390/diagnostics14020144
6. Safdar S, Zafar S, Zafar N, Khan NF. Machine learning based decision support systems (DSS) for heart disease diagnosis: a review. Artificial Intelligence Review [Internet]. 2017;50(4):597–623. Available from: https://doi.org/10.1007/s10462-017-9552-8
7. Xi Y, Wang H, Sun N. Machine learning outperforms traditional logistic regression and offers new possibilities for cardiovascular risk prediction: A study involving 143,043 Chinese patients with hypertension. Frontiers in Cardiovascular Medicine [Internet]. 2022;9. Available from: https://doi.org/10.3389/fcvm.2022.1025705
8. Deepa R, Sadu VB, C PG, Sivasamy A. Early prediction of cardiovascular disease using machine learning: Unveiling risk factors from health records. AIP Advances [Internet]. 2024;14(3). Available from: https://doi.org/10.1063/5.0191990
9. Ho KJ. Cardiovascular diseases. In: CRC Press eBooks [Internet]. 2018. p. 75–100. Available from: https://doi.org/10.1201/9781351075145-3
10. Hamer M, O’Donovan G, Stamatakis E. Association between physical activity and sub-types of cardiovascular disease death causes in a general population cohort. European Journal of Epidemiology [Internet]. 2018;34(5):483–7. Available from: https://doi.org/10.1007/s10654-018-0460-2
11. Geraghty L, Figtree GA, Schutte AE, Patel S, Woodward M, Arnott C. Cardiovascular disease in Women: From pathophysiology to novel and emerging risk factors. Heart Lung and Circulation [Internet]. 2020;30(1):9–17. Available from: https://doi.org/10.1016/j.hlc.2020.05.108
12. Frąk W, Wojtasińska A, Lisińska W, Młynarska E, Franczyk B, Rysz J. Pathophysiology of Cardiovascular Diseases: New Insights into Molecular Mechanisms of Atherosclerosis, Arterial Hypertension, and Coronary Artery Disease. Biomedicines [Internet]. 2022;10(8):1938. Available from: https://doi.org/10.3390/biomedicines10081938
13. Magnussen C, Ojeda FM, Leong DP, Alegre-Diaz J, Amouyel P, Aviles-Santa L, et al. Global effect of modifiable risk factors on cardiovascular disease and mortality. New England Journal of Medicine [Internet]. 2023;389(14):1273–85. Available from: https://doi.org/10.1056/nejmoa2206916
14. Mathur P, Srivastava S, Xu X, Mehta JL. Artificial intelligence, machine learning, and cardiovascular disease. Clinical Medicine Insights Cardiology [Internet]. 2020;14:117954682092740. Available from: https://doi.org/10.1177/1179546820927404
15. Shin S, Austin PC, Ross HJ, Abdel–Qadir H, Freitas C, Tomlinson G, et al. Machine learning vs. conventional statistical models for predicting heart failure readmission and mortality. ESC Heart Failure [Internet]. 2020;8(1):106–15. Available from: https://doi.org/10.1002/ehf2.13073
16. Johnson KW, Soto JT, Glicksberg BS, Shameer K, Miotto R, Ali M, et al. Artificial intelligence in cardiology. Journal of the American College of Cardiology [Internet]. 2018;71(23):2668–79. Available from: https://doi.org/10.1016/j.jacc.2018.03.521
17. Subramani S, Varshney N, Anand MV, Soudagar MEM, Al-Keridis LA, Upadhyay TK, et al. Cardiovascular diseases prediction by machine learning incorporation with deep learning. Frontiers in Medicine [Internet]. 2023;10. Available from: https://doi.org/10.3389/fmed.2023.1150933
18. Liu W, Laranjo L, Klimis H, Chiang J, Yue J, Marschner S, et al. Machine-learning versus traditional approaches for atherosclerotic cardiovascular risk prognostication in primary prevention cohorts: a systematic review and meta-analysis. European Heart Journal – Quality of Care and Clinical Outcomes [Internet]. 2023; Available from: https://doi.org/10.1093/ehjqcco/qcad017
19. Masih N, Ahuja S. Application of data mining techniques for early detection of heart diseases using Framingham heart study dataset. International Journal of Biomedical Engineering and Technology [Internet]. 2022;38(4):334. Available from: https://doi.org/10.1504/ijbet.2022.123149
20. Alfadli KM, Almagrabi AO. Feature-Limited prediction on the UCI heart disease dataset. Computers, Materials & Continua/Computers, Materials & Continua (Print) [Internet]. 2022;74(3):5871–83. Available from: https://doi.org/10.32604/cmc.2023.033603
21. Ashrafi N, Abdollahi A, Zhang J, Pishgar M. Optimizing Mortality Prediction for ICU Heart Failure Patients: Leveraging XGBoost and Advanced Machine Learning with the MIMIC-III Database [Internet]. arXiv.org. 2024. Available from: https://arxiv.org/abs/2409.01685
22. Sami O, Elsheikh Y, Almasalha F. The role of data pre-processing techniques in improving machine learning accuracy for predicting coronary heart disease. International Journal of Advanced Computer Science and Applications [Internet]. 2021;12(6). Available from: https://doi.org/10.14569/ijacsa.2021.0120695
23. Olamendy JC. Mastering feature scaling and normalization in machine learning. Medium [Internet]. 2025; Available from: https://medium.com/%40juanc.olamendy/mastering-feature-scaling-and-normalization-in-machine-learning-985fdecfab5c
24. Aryuni M, Adiarto S, Miranda E, Madyatmadja ED, Albert VDS, Sestomi E. Imbalanced learning in heart disease Categorization: Improving minority class prediction accuracy using the SMOTE algorithm. International Journal of Fuzzy Logic and Intelligent Systems [Internet]. 2023;23(2):140–51. Available from: https://doi.org/10.5391/ijfis.2023.23.2.140
25. B Patil P. MACHINE LEARNING BASED ALGORITHM FOR RISK PREDICTION OF CARDIO VASCULAR DISEASE (CVD). Journal of Critical Reviews [Internet]. 2020;7(09). Available from: https://doi.org/10.31838/jcr.07.09.157
26. Ray S. A quick review of machine learning algorithms [Internet]. IEEE Conference Publication | IEEE Xplore. 2019. Available from: https://ieeexplore.ieee.org/abstract/document/8862451/
27. Khan A, Qureshi M, Daniyal M, Tawiah K. A novel study on Machine Learning Algorithm-Based Cardiovascular Disease Prediction. Health & Social Care in the Community [Internet]. 2023;2023:1–10. Available from: https://doi.org/10.1155/2023/1406060
28. Asif MdAAR, Nishat MM, Faisal F, Dip RR, Udoy MH, Shikder MdF, et al. Performance evaluation and comparative analysis of different machine learning algorithms in predicting cardiovascular disease. EBSCOhost [Internet]. 2021; Available from:https://openurl.ebsco.com/EPDB%3Agcd%3A2% 3A4098514/detailv2?sid=ebsco%3Aplink%3Ascholar&id=ebsco% 3Agcd%3A150582089&crl=c&link_origin=www.google.com
29. Das R, Turkoglu I, Sengur A. Effective diagnosis of heart disease through neural networks ensembles. Expert Systems With Applications [Internet]. 2008;36(4):7675–80. Available from: https://doi.org/10.1016/j.eswa.2008.09.013
30. Dilsizian ME, Siegel EL. Machine Meets Biology: a Primer on Artificial Intelligence in Cardiology and Cardiac Imaging. Current Cardiology Reports [Internet]. 2018 Oct 18;20(12). Available from: https://doi.org/10.1007/s11886-018-1074-8
31. Kutiame S, Millham R, Adekoya AF, Tettey M, Weyori BA, Appiahene P. Application of Machine Learning Algorithms in Coronary Heart Disease: A Systematic Literature Review and Meta-Analysis. International Journal of Advanced Computer Science and Applications [Internet]. 2022 Jan 1;13(6). Available from: https://doi.org/10.14569/ijacsa.2022.0130620
32. Alrashdi I, Taloba AI. Integration of graph neural networks and long short-term memory models for advancing heart failure prediction. Alexandria Engineering Journal [Internet]. 2025 May 12;127:143–63. Available from: https://doi.org/10.1016/j.aej.2025.05.014
33. Navita N, Mittal P, Sharma YK, Lilhore UK, Simaiya S, Saleem K, et al. Advanced hybrid machine learning model for accurate detection of cardiovascular disease. International Journal of Computational Intelligence Systems [Internet]. 2025 Mar 6;18(1). Available from: https://doi.org/10.1007/s44196-025-00771-1
34. Krittanawong C, Virk HUH, Bangalore S, Wang Z, Johnson KW, Pinotti R, et al. Machine learning prediction in cardiovascular diseases: a meta-analysis. Scientific Reports [Internet]. 2020 Sep 29;10(1). Available from: https://doi.org/10.1038/s41598-020-72685-1
35. Stafie CS, Sufaru IG, Ghiciuc CM, Stafie II, Sufaru EC, Solomon SM, et al. Exploring the intersection of artificial intelligence and clinical healthcare: A multidisciplinary review. Diagnostics [Internet]. 2023 Jun 7;13(12):1995. Available from: https://doi.org/10.3390/diagnostics13121995
36. Alasmari S, AlGhamdi R, Tejani GG, Sharma SK, Mousavirad SJ. Federated learning-based multimodal approach for early detection and personalized care in cardiac disease. Frontiers in Physiology [Internet]. 2025 Apr 23;16. Available from: https://doi.org/10.3389/fphys.2025.1563185

Cite this article as:
Kashif M. A Comprehensive Review of Machine Learning Algorithms in Predicting Cardiovascular Diseases. Premier Journal of Cardiology 2026;6:100015

Export to EndNote Export to Mendeley