Research On Airport Noise Prediction With Time Series Analysis

The rapid development of China’s civil aviation has attracted worldwide attention and theproportion of civil aviation in the national comprehensive transportation system is on the increasewhich contributes a lot more to the national economy, social development, reform and opening up tothe outside world. However, the continuous new expansion of the airport and the continuing growth inair traffic, at the same time, problems about China’s civil aviation airport noise is becoming worse. Inorder to build the green modern civil aviation system, to achieve the sustained and healthydevelopment of the civil aviation industry, there is an urgent need to take advantage of thestate-of-the-art IT monitoring data for further scientific prediction, thereby providing decision supportfor civil aviation sector. Therefore, the prediction technology applied to the actual noise predictionproblem has important theoretical and practical significance.Support vector regression, with the superiority which the algorithm based on empirical riskminimization principle difficult to compare, at the same time, due to its algorithm is simple, there isno local minimum and the curse of dimensionality and generalization ability, etc., suitable for timeseries prediction, therefore, so this article use support vector regression as time series forecastingmethods.In each class, consider the idea of classification based on support vector regression, regressionafter classification can effectively reduce the differences between samples, and be able to achievebetter results. Consider the criteria of airport noise classification is not clear, and the number of classcannot be determined at some time, for this case, combined with the characteristics of cluster analysis,this article proposed a method of regression after clustering in time series prediction.Single clustering vulnerable initialization state and parameters affect the clustering effect has agreater impact for the predicted effect, when clustering works ineffectively, the class internal sampledifferences will increase instead, so that the regression forecast deviation will increase. In order tosolve the problem of unstable exists in single clustering algorithm, this article propose clusteringensemble regression algorithm using probabilistic model of the relationship matrix stack to beintegrated on a single cluster effect. In order to verify the effectiveness of the proposed method, weexperiments on a common data set, and apply it to the airport noise time series prediction.