Anomaly Detection Of Traffic

One region’s traffic is a way to show the performance of the number of activities in this area. The traffic’s abnormal diversification is often accompanied by the changes in the number of activities. Anomaly detection system can be used for emergencies, providing the effective basis for the government timely to prevent the events. Now, the anomaly detection methods can not be warning timely or need a large number of learning sample size, which is the inadequacies to those methods.This paper mainly introduces the wavelet decompose algorithm and the ARMA algorithm, using Hurst exponent to search signal abnormal points. This paper proposes using DB wave to decompose signal function, extracting the detail coefficients and the approximation coefficients. Detail coefficient is used to find out abnormal signals, approximation coefficient is used to rebuild signal function.Experiments show, select the appropriate DB wave and wavelet-decomposed can be more accurate and timely analysis of anomaly points. According to the characteristics of wavelet decomposition of the traffic data, this paper combines with the ARMA algorithm, can analysis the direction of data changing, thus more accurate analysis of normal inflection points and abnormal inflection points. All this can provide reliable for the project.