A Study On Two Classes Of Spatio-temporal Data Models And Their Application

In recent years, with the rapid development of sensor technology and mobile network technology, spatio-temporal data has increased dramatically. The analysis and processing of spatio-temporal data has increasingly become the important topic and was paid close attention by IT industry, business and environmental meteorological and national security departments.Modeling and prediction for spatio-temporal data is the key to the spatio-temporal data mining problems. This dissertation studies spatio-temporal autoregressive moving average model, spatio-temporal kriging model, and discusses their modeling process. The performance of these two kinds of model is verified by real-world problem. Furthermore, the advantages and disadvantages of those models are discussed respectively.In this paper, the main work includes the following aspects:(1) The property of the spatio-temporal data is studied and analyzed. Spatio-temporal data generally has many properties, including dynamic, mass, high dimension, multi-scale,spatio-temporal autocorrelation, variability, heterogeneity and nonlinearity. Then, the spatio-temporal autocorrelation and spatio-temporal heterogeneity are emphatically discussed.The study of the spatio-temporal data property is the important basis for the spatio-temporal data modeling.(2) The spatio-temporal autoregressive moving average model was studied and analyzed.Firstly, we deeply discusses the regression model, the spatial autoregressive model, the spatio-temporal autoregressive moving average model, the time moving average model, the space model moving average model and the spatio-temporal of moving average model,respectively. Then the modeling process of spatio-temporal autoregressive moving average model is given. Finally, the effectiveness of the spatio-temporal autoregressive moving average model is verified based on the real data of the gross domestic product(GDP) from2006 to 2014 of 13 districts and counties of Xi’an city.(3) The modeling process of the spatio-temporal kriging model is studied and analyzed.Firstly, the definition of spatio-temporal covariance function is introduced, and then the spatio-temporal fully symmetry and separability is discussed. Several classes of nonseparable stationary spatio-temporal covariance function model are introduced. Then, the modeling process of simple spatio-temporal kriging model is studied. Finally, based on the real spatio-temporal data- Irish wind speed data, the effectiveness of simple spatio-temporal kriging prediction model is verified by establishing a general stationarycorrelation function model.