An improved principal component analysis method for railway accident prediction using uncertain data

摘要

Uncertain data is an important issue in research of big data and railway accident prediction. In this paper, we introduce categories of uncertain data, and propose a method to deal with uncertain data in railway accident prediction. This method assumes that uncertain data obeys a special probability distribution, and then extracts features from the whole data using fuzzy principal component analysis (FPCA). Finally, employing total support vector classification (TSVC), railway accident prediction result is obtained, and the experiment shows that the prediction result has been risen by 5.02% compared to classical principal component analysis and support vector machine method.