Based on the Chi-Square Statistics and Test, this paper proposes a method named ABSA (application behavior significance assessment) to analyze the traffic behavior characteristics of applications. The ABSA method does not focus on any certain applications; in contrast, it aims at providing a quantitative standard for describing the behavior distribution differences among applications, so that the traffic behavior characteristics and their corresponding significances can be determined. The theoretical analysis and experiments results show that 1) ABSA can present the information about characteristics more precisely and copiously to improve the accuracy of application identification; 2) the significance of characteristic is independent of its proportion in sample totals; 3) ABSA can keep the relative significance sequence of behavior characteristics unchanged in a packet sampling environment, which is often used by NetFlow and many other flow information collecting systems to simplify the characteristic re-selecting process when sampling ratio is changed.