Network traffic anomaly detection is a difficult problem in network management. This paper presents a Wavelet Generalized Likelihood Ratio (WGLR) algorithm and an Error Performance Detection (EPD) algorithm to solve this problem. WGLR algorithm combines Generalized Likelihood Ratio (GLR) algorithm and wavelet transform method, and captures the failure point in real time. Error performance Detection (EPD) algorithm is based on the prediction error of the traffic model and regards the error as the statistical variable, comparing with the threshold, which will detect the anomalous change in the signal without test window delay. Simulation and network traffic experiment has demonstrated that the algorithm has better performance in fault detection.