Network Traffic Model Based on Wavelet Decomposition and Arima
|Course||Applied Computer Technology|
|Keywords||Network traffic model Time Series Wavelet decomposition and reconstruction ARIMA|
With the advent of the information age, the Internet rapid development, the endless variety of new network applications, resulting in the emergence of a variety of network problems, enormous challenges to network monitoring. Network detection is an important means to ensure the normal operation of the network, a suitable network traffic model effectively network monitoring has important significance. From the similarity of the current performance of the network traffic, multiple points of pleomorphic other complex characteristics, traditional traffic model has been unable to meet the demand. The paper analyzes the characteristics of the traditional model of network traffic forecast mature model for the smooth flow of network performance. However, current network traffic because of the emergence of a variety of complex features, generally exhibit unstable characteristics of network traffic in order to solve a variety of complex features stationarity, the introduction of wavelet transform technology, using its multi-resolution characteristics, a variety of complex factors of network traffic is decomposed into different scales, easy to separate. This use of the wavelet decomposition of non-stationary time series can be decomposed into the appropriate different bands on multiple stationary time series, relatively convenient and efficient processing instead of unified signal processing on these smooth timing eventually restored to the original scale methods to improve the prediction accuracy of real network traffic. Introducing wavelet technology prediction accuracy at the same time, because of the single time series is decomposed into a plurality of groups of time sequence of the different scales, resulting in serious deterioration of the time complexity of the method, affecting its practical application. By a detailed analysis and study of this method, and put forward a number of improvement measures, under the premise of not seriously affect the predicted effect, as much as possible to shorten the time complexity of the algorithm. First, the wavelet decomposition of the obtained plurality of groups of the sub-sequence of the network traffic, to the analysis of the characteristics will merge under similar spectral characteristics similar flow promoter sequence, to reduce the number of sub-sequences, thus reducing the number of modeling prediction. The premise of the method is similar sequence features a smooth, steady sequence of algebraic operations remain stable, and therefore you can still get a reasonable final results, experiments show that the prediction accuracy is not affected. Then, the combined sub-sequence is divided into high spectral detail items, outline entry of low-frequency spectrum, and the middle of the spectrum of a periodic term of three to take different measures to further reduce the time complexity of three different sequences. The thesis this manner, so that the three components using a small amount of historical data, a lower sampling frequency, to achieve as much as possible without affecting the improvement brought by the wavelet transform on the prediction accuracy. Proved by experiments, taking a series of time complexity measures are improved, compared to the forecast results in the introduction of wavelet decomposition technique simple subsequence complete modeling to predict the effect had not been seriously affected. Therefore, the use of wavelet transform and traditional ARIMA model of network traffic forecast method is feasible.