Hypertext transfer protocol (HTTP) video streaming has occupied more than half of mobile Internet traffic today. Quality of experience (QoE) estimation for HTTP video streaming is significantly meaningful for network operators, as they can utilize valuable network resources more efficiently based on the QoE estimated. But, it is very hard to estimate QoE accurately with theoretical methods. In this paper, we adopt three buffering-related application metrics, which have been verified to be the main factors that influence the QoE of HTTP videos, together with three fundamental quality of service (QoS) metrics in network layer, to estimate QoE. We suggest a two-layered mapping model to capture the correlation between network QoS and application QoS and that between application QoS and QoE. Then, we exploit four regression methods, that is, linear regression, back propagation neural network, general regression neural network and support vector regression (SVR), to investigate the mapping relations of the network metrics, the application buffering-related metrics and QoE. Experimental results demonstrate that general regression neural network and SVR perform almost equally well in fitting ability, and SVR performs best among the four methods in terms of generalization ability. Finally, we analyze the influences of different network QoS parameters on QoE based on the experimental results.KEY WORDS: quality of experience (QoE); HTTP video streaming; mobile Internet; support vector regression (SVR); general regression neural network (GRNN); back propagation neural network (BPNN); quality of service (QoS); linear regression (LR)