Visualizing the Full Spectrum Optimization of K-Nearest Neighbors From Data Preprocessing to Hyperparameter Tuning and K-Fold Validation for Cardiovascular Disease Prediction

Cardiovascular disease (CVD) is a prominent cause of death worldwide. This alarming need requires an accurate prediction model using machine learning that can detect and help prevent or mitigate the risk. This study focuses on this issue and has come up with new dimensional capabilities to enhance the KNearest Neighbors (KNN) algorithm to predict cardiovascular diseases at an early stage by incorporating various techniques for data preprocessing and feature selection thereby improving the efficiency of the model. The proposed model identifies the most relevant features using Principal Component Analysis. The main innovation revolves around fine tuning the hyperparameter of K-Nearest Neighbors, specifically the choice of neighbors (K), using a data driven approach to ensure accuracy across different datasets. The performance of the optimized K-Nearest Neighbors algorithm is evaluated using the Framingham heart disease dataset. This model achieved an impressive prediction accuracy of 92.46% and outperformed methods that solely rely on traditional K-Nearest Neighbors. As machine learning techniques plays an important role in the development of prediction models for early detection and prevention of cardiovascular disease, this model can be considered as a valuable tool for healthcare professionals and researchers. The core contribution of this study lies in offering a comprehensive optimization of the traditional K-Nearest Neighbors (KNN) algorithm. This includes robust data preprocessing using the Hampel filter for outlier removal, feature selection through Principal Component Analysis (PCA), and performance enhancement using grid search for hyperparameter tuning combined with 10-fold cross-validation. Unlike prior studies that apply KNN with minimal adjustments, this research emphasizes the importance of an end-to-end machine learning pipeline. This holistic refinement significantly improves the predictive performance and reliability of KNN for cardiovascular disease prediction, achieving 92.46% accuracy on the Framingham dataset.