Research On Prediction Model Of Single Amonitoring Point Airport Noise Time Series

With the continuous development of the society and the modernization of China, China’s aviationindustry gets sustainable development. Although air transport in the city brings convenience andprosperity, it has also brought a series of environmental problems, especially the noise pollution. Inorder to plan airport layout better, construct schedule more reasonable and prevent noise moreeffectively, we need to study the airport noise time series. Therefore, this paper focuses on the airportnoise time series analysis and prediction methods.This paper first introduces the concept of time series, forecasting model and its application range.Based on the analysis on the characteristics of various models, the paper presents the prediction modelbased on SSA (Singular Spectrum Analysis). This method uses SSA to decompose airport-noise timeseries, and gets the principal component and the experience of political function. We analyze thecharacteristics of trend and vibration, and select the appropriate feature vectors for sequencereconstruction, and then obtain the prediction model by linear repeating formula. At last, we use thestate-transition matrix and contribution ratio to revise the forecast values. The experiment on themeasured data of an airport shows that the accuracy of this model is better than other originalprediction models.In view of the complexity of the practical problems, time series data is usually accompanied withlinear and nonlinear features, linear model simply cannot well capture the composite characteristics.Therefore, it is necessary to combine linear model and nonlinear model together, to constitute a newmodel, which is used to improve the time series prediction performance. Grey prediction is accordingto the small amount of data, then establish the grey differential model, to search for the rule and minitrend, and prediction. Support vector machine is a new machine learning method. It accords to thelimited sample information, and find the best compromise between model complexity and learningability, which is succeeded in solving high dimension problem and the local minimum. We putforward the prediction model of airport noise time series based on GM_LSSVR. This method is basedon the advantages of two algorithms, and takes the characteristics of airport noise time series intoaccount. Namely, it decomposes the noise time sequence into tendency and the residual, thenestablishes both the GM(1,1) model of trend term and the LSSVR model of residual term, which areused for airport-noise prediction model. The experiments on the real data of an airport show that theaccuracy of this model is better than other prediction models.