Research On Interactive Prediction Of Airport Noise Monitoring Points Based On Time Series Similarity Measure

With the significant developing of civil aviation,the GDP is promoted,as well as the increasingly prominent problem of airport noise pollution,so the prevention work of noise is becoming crucial.To solve the problem,with the help of the IoT(Internet of Things)technique,the noise monitoring nodes are set up in the affected area to collect and analyse the noise data for the decision of the prevention work.However,the monitoring nodes will be aging and broken due to the service life and weather which prevents the nodes collecting correct noise data continuously.Therefore,it is important to know how to find the abnormal nodes by analysing the collected data.More importantly,how to predict the valid data for the abnormal nodes becomes a valuable problem to research,the main works includes:(1)The similarities between noise monitoring series reflect the closeness of the monitoring nodes.In this paper,we provide a comprehensive survey of the typical methods of time series representations and time series similarity measures.(2)Aiming at the defect that most of the existing time series representationss can not accurately extract and represent the trend features of time series,this paper proposes an adaptable Trend Segmentation Representation of time series based on Iterative End Point algorithm(IEPF-TSR).IEPFTSR combines with the IEPF algorithm,can accurately extract the key trend turning points in time series and remove other redundant data points in the series.,The time series which after IEPF-TSR can intuitively show the starting point,the steep trend and the trend average of each key trend in the series,which can describe the trend information more clearly and clearly than the existing time series representations.(3)Aiming at the defect that the existing similarity measures of time series can not accurately measure the similarity of the trend features of time series.So,combining with the proposed method of time-series feature representation based on trend segment,in this paper we proposed a time-series similarity measure based on Adaptive Trend Segmentation(TSR-DIST).TSR-DIST combining with the principle of DTW that can automatically finds the optimal measure path,measures the similarity between sequences from the trend segment average and the segment trend variable respectively,which can solve the existing problems in the measurement of the trend characteristics of the current similarity measure method.(4)From the view of the interaction between monitor nodes,in this paper,an interactive prediction model of monitor nodes was built after analysing the noise data of Beijing Capital International Airport.The similarity of time-series is explored and the performance of Euclidean distance,Dynamic Time Warping distance and TSR-DIST distance are compared.Then,the correlative monitor nodes collection is selected and according to the similarity value of time-series,a time-series similarity based feature weighted support vector regression machine is proposed.Finally,the predicting performance and generalization capability of the model are verified on the real airport noise dataset.