Kernel Principal Component Analysis for Uncertain Data Objects and Its Application in Classification

ABSTRACT Uncertain data mining has been a growing field of research in recent years. Numerous data mining techniques for performing tasks such as clustering, classification, anomaly detection, and so on have been developed for uncertain data. Principal component analysis (PCA) and its extension, kernel principal component analysis (KPCA), are two well‐known techniques that are widely used for dimensionality reduction and feature extraction for traditional certain data. However, for uncertain data, these techniques have not been developed to the best of our knowledge. In this paper, uncertain principal component analysis (UPCA) and uncertain kernel principal component analysis (UKPCA) are developed. The proposed techniques consider the inherent uncertainty of the uncertain data, unlike the traditional techniques that ignore such uncertainty. In addition, in this paper, we propose the decision tree algorithm classification model combined with the developed UKPCA technique. The proposed model is capable of achieving high classification accuracy for both real‐world and synthetic data, especially for cases that involve classes having nonlinear/arbitrary shapes.