With the rapid development and application of high-speed railway EMU, its safety and reliability have attracted wide attention. In order to estimate the health status of axle box bearing of high-speed railway EMU accurately (hereinafter referred to as axle box bearing), this study proposes a classification algorithm based on Decision Tree and Support Vector Machine, and utilizes Principal Component Analysis (PCA) to reduce the feature dimension simultaneously. In addition, the performance of the classification can be further improved by collecting the temperature data of axle box bearing and various components on the drive side and non-drive side of which either in the same bogie or in different bogie. And the Analytic Hierarchy Process (AHP) has been used to distribute the weight of vectors. Extensive experiments demonstrate the effectiveness of the classification model, and the accuracy has been increased by about 5%. Furthermore, the judgment ability of health status of axle box bearing and the precision of the operation and maintenance policy can be enhanced if we establish a health assessment model through dividing the health status of axle box bearing into four parts including health, temperature rising, strong temperature, and irritative temperature.