Performance Analysis of PCA, Sparse PCA, Kernel PCA and Incremental PCA Algorithms for Heart Failure Prediction

Heart failure (HF) prediction is a challenging issue in medical informatics and is considered a deadliest disease worldwide. Recent research has been concentrated on features transformation and selection for improved HF prediction. In this study, we search optimal feature extraction algorithm by evaluating the performance of different feature extraction algorithms namely Principle Component Analysis (PCA), Sparse PCA, Kernel PCA and Incremental PCA. These algorithms are integrated with machine learning models to improve HF prediction. The performance of all these integrated models are evaluated by analyzing Cleveland heart failure database. Experimental results pointed out that Kernel PCA algorithm integrated with linear discriminant analysis model and Sparse PCA integrated with Gaussian Naive Bayes (GNB) model offers 91.11% of HF classification accuracy. Hence, based on the experimental results it is shown that Kernel PCA and Sparse PCA are suitable feature extraction methods for HF data.