Next-Gen Cardio Care: Engineering Precision with Hyperdimensional K in KNN Cardiac Disease Prognosis

Background: Cardiovascular-Diseases (CVD’s) are acknowledged as the primary source of death and disability worldwide. The global mortality toll from cardiovascular disease raised from 12.4 million in 1990 to 19.8 million in 2022, according to the 2023 study on the Global Burden of Disease (GBD), which was documented in the Journal of the American College of Cardiology. India has a higher age-standardized death rate from cardiovascular disease (CVD) than the global average, according to the study’s findings. India has a death rate of two hundred seventy-two per one lakh people, compared to the global average of two hundred thirty-five per lakh people. Objective: Our primary objective is to find out the optimal K value in the KNN algorithm by evaluating several approaches for selecting the best K value. Through these various methodologies, we aim to refine the KNN algorithm’s performance and establish a robust, reliable approach for optimal K value selection. This optimization is critical for ameliorate the accuracy and efficiency of KNN in prognostic tasks. Method : We use the KNN machine learning classification method to predict heart disease. For comparison, we have proposed five methods for K nearest neighbors to calculate the k optimal value and compared them to each other. Lastly, we evaluate which optimal value of K to choose based on the specific characteristics of our heart failure prediction dataset and the performance of the KNN algorithm. The confusion matrix alongside various performance metrics and ROC-AUC have been attuned to assess our model. The entire collection of data has been split into two halves under this architecture: the training set makes up 70% of the data. In order to test and assess the model’s performance, thirty percent of the whole dataset is set aside. Results: The developed KNN model has demonstrated exceptional performance using domain knowledge, achieving 89.49% accuracy, with a precision of 0.90%, sensitivity of 0.89%, and an F1 score of 0.89% on the test dataset. Additionally, the model attained a ROC-AUC value of 0.89, reflecting its strong classification ability and robustness in distinguishing between positive and negative cases. Conclusion: This paper aims to determine which KNN method gives the best optimal value and best accuracy for classification in heart disease prediction. The resulting proposed model has shown significant accuracy in the method. According to the research, the suggested architecture is appropriate for real-time medical applications and can improve the KNN algorithm for our cardiac prediction dataset.