Abstract:
This study examines the utilization of machine learning algorithms for the early
identification of cardiovascular illnesses through electrocardiogram (ECG) data. The study
assesses various models, including Support Vector Machines (SVM), Logistic Regression
(LR), NB, Linear Regression (LinReg), and ensemble methods such as XGBoost, RF,
Gradient Boosting (GB, and KNN. The dataset obtained from Kaggle contains ECG
measurements and binary labels denoting normal or pathological cardiac function. The
performance of each algorithm is evaluated according to its proficiency in accurately
classifying ECG patterns and differentiating between normal and pathological cardiac
activity. The findings indicate that SVM attains the best accuracy of 0.994, illustrating its
proficiency in identifying intricate, non-linear correlations within high-dimensional ECG
data. LR and RF closely follow with accuracies of 0.993, showcasing their robustness in
modeling linear and probabilistic trends. LinReg performs admirably with an accuracy of
0.992, while XGBoost and KNN achieve comparable scores of 0.9904, highlighting their
versatility and noise tolerance. GB and NB report slightly lower accuracies of 0.9888 and
0.9688, respectively, yet remain valuable due to their unique strengths in handling diverse
data distributions and probabilistic classification. Ensemble techniques like XGBoost, RF
and GB leverage the power of multiple weak learners to deliver strong predictive
performance. Meanwhile, KNN's adaptability to varying data patterns underscores its
utility in practical applications. These findings highlight the capability of machine learning
algorithms to automate ECG interpretation and aid healthcare professionals in prompt
diagnosis. Future work should focus on optimizing these models further and validating
their performance in real-world clinical settings. By enhancing diagnostic accuracy and
efficiency, this research contributes to advancing cardiovascular health monitoring and
improving patient outcomes.
