International Journal of Science, Engineering and TechnologyAuthors: Reshoo Devi, Shilpy Sharma, Raj Kumar
Abstract: Heart disease remains a major cause of death worldwide, accounting for a significant portion of annual mortality and posing significant challenges to healthcare systems. Preventing severe cardiac events requires timely and accurate diagnosis, but conventional diagnostic procedures are often time-consuming, costly, and dependent on expert clinical judgment, which may lead to variability in diagnosis. Intelligent systems that can assist clinicians in early disease prediction and risk assessment have been developed thanks to the increasing availability of medical data and advancements in machine learning (ML) and deep learning (DL). This research proposes a comprehensive Heart Disease Prediction System (HDPS) that utilizes machine learning and deep learning techniques to classify patients as either having heart disease or being healthy based on multiple clinical parameters. The system is developed using the UCI Cleveland Heart Disease dataset, from which 14 standard medical attributes were selected, including age, sex, chest pain type, cholesterol level, resting blood pressure, electrocardiographic results, and exercise-induced angina. The impact of data preprocessing strategies like feature selection, normalization, and outlier detection on predictive performance was analyzed using a multi-phase modeling approach. A sequential deep learning model was used along with multiple ML classifiers, including Logistic Regression, K-Nearest Neighbors, Support Vector Machine, Decision Tree, Random Forest, and XGBoost, for implementation and evaluation. Metrics used to assess model performance included accuracy, precision, recall, specificity, sensitivity, F1-score, and confusion matrix analysis. Proper preprocessing can significantly enhance model accuracy and stability, as demonstrated by experimental results. It has been confirmed by the results that predictive systems based on machine learning and deep learning can be reliable, efficient, and cost-effective clinical decision support tools for detecting heart disease early. Healthcare environments with limited resources can benefit from systems that can enhance diagnostic accuracy, reduce manual workload, and support timely medical intervention. The importance of data quality and preprocessing in medical analytics is highlighted in this study, which provides a foundation for future research that focuses on hybrid models, larger datasets, and real-time healthcare applications.
DOI: http://doi.org/10.5281/zenodo.21456108
