The Prediction Method of Actual Traffic Based on ARMA and Wavelet Transform

Abstract

In order to improve the prediction accuracy of wireless sensor network, a novel prediction algorithm is proposed by ARMA and wavelet transform. In this algorithm, the characteristic of ARMA is defined, and the Theoretical foundation of is given. Then, the prediction error of actual traffic is decreased by fusion the results of ARMA and wavelet transform. Finally, a simulation was conducted to study the key influence factor of algorithm with OPNET and MATLAB. The results show that, compared to FARIMA model, This paper has better suitability. Introduction Wireless Sensor Network (Wireless Sensor Network, WSN) have unique applications, and obtained the rapid development of [1-5]. Network congestion in WSN is to reduce the performance of the key factors, in recent years, some scholars puts forward some solutions to network congestion in WSN, CODA [6], fusion [7], etc. Through the network nodes and the flow condition of congestion control. How to effectively handle before network congestion, has become a WSN important method to improve the network performance. How to control the actual flow, predict the network traffic status has been becoming the current hot and difficult. A typical network prediction method has AR, ARMA and FARIMA model. At the same time, with the deepening of the research, other methods such as wavelet transform, chaos model have also been introduced in actual traffic prediction. To this, the domestic and foreign scholars have done a lot of research work. Literature [8] with super linear convergence of the variable metric method to improve ARMA prediction model, and based on autocorrelation coefficient and partial autocorrelation coefficient trailing method of actual traffic forecasting. Literature [9] base by extending the start of the iteration and the last stage of the search time, realizes the algorithm of the balance between global search and local search ability, and optimize the model parameters, based on chaos IPSO optimization support vector machine forecasting model. Literature [10] by using the least squares support vector machine (SVM) and fuzzy LSSVM training, set up a kind of optimal sample subset online fuzzy prediction algorithm, and further study of the actual flow of degeneration and long cycles. Literature [11] in view of the traditional forecast model high dependence of training data, the combination of wavelet technology and ant colony algorithm to build the Back Propagation network weights of prediction methods, improve the prediction accuracy. Literature [12] to improve the Logistic model, based on the cosine function and using the method of nonlinear time series analysis and Logistic model to describe the evolution of the state of actual traffic situation and chaotic characteristics. Literature [13] traffic time series wavelet decomposition, the wavelet transform scale coefficients obtained sequence, and the wavelet coefficients by coefficient sequences and the flow of the original time series as input and output of the model respectively, and structure of artificial neural network to train the literature [14] for real-time update of prediction accuracy problems, set up an online fuzzy least squares support vector machine (SVM) method, but need to further consider the effect of time scale. On the basis of the above work, this paper puts forward a new kind of actual traffic prediction algorithm, this algorithm through [fusion ARMA model and the prediction of wavelet transform, in order to reduce the prediction error, through mathematical simulation validates the effectiveness of the algorithm. At the same time the paper structure as follows: section 1 presents the characteristics 5th International Conference on Information Engineering for Mechanics and Materials (ICIMM 2015) © 2015. The authors Published by Atlantis Press 1182 of ARMA model, section 2 presents a judgment based on ARMA model, and to establish a prediction algorithm, section 3 experiments are carried out using OPNET and MATLAB simulation, the fourth quarter to summarize the full text. Design of the Chip Kick Mechanics the ARMA model Regression sliding model has high forecast precision and rapidity, timeliness is good wait for a characteristic. In the wireless sensor normal flow sequence is: W1, W2,... Wt, the model representation is as follows: