Geographic Spatiotemporal Data Interpolation Model Based On Gaussian Process And Its Application

Geospatial spatiotemporal data is an important data that can be used to describe the evolution of geographic phenomena.However,data missing is unavoidable in the collection and storage of geographic spatiotemporal data,which will cause some interference to geographic analysis.In order to ensure the integrity and continuity of data,some researchers have interpolated these data.However,conventional geographic data interpolation methods focus on the research of static spatial location and attributes,and these methods ignore the spatial-temporal correlation and anisotropy of data,resulting in the loss of a large amount of valuable information.Moreover,geographical attributes do not exist alone,and the changes of various attributes often have a certain correlation.Using the correlation information between these spatio-temporal variables and sharing information in different data sets through multi output learning can usually improve the interpolation effect and better solve the problem of insufficient data.Gaussian process(GP)is a machine learning algorithm that has developed rapidly in recent years.It can fuse the prior information of data for statistical analysis,and has good adaptability to deal with complex problems such as high dimension and nonlinearity.Based on the Gaussian process,this paper analyzes the potential laws of time series data,and then constructs the covariance function to simulate the hidden features in the data set,which can obtain more reliable interpolation results;According to the directional characteristics in the spatial distribution of geographic data,the spatial anisotropy of geographic attributes is studied,and an anisotropic interpolation model with correlation coefficient instead of Euclidean distance is further proposed,which has higher interpolation accuracy than the conventional algorithm;This paper discusses the spatiotemporal integration mode of geographic data,constructs the spatiotemporal covariance function,realizes the intelligent integration of spatiotemporal information,introduces the concept of multi output learning,studies the correlation characteristics of various geographic attributes,and constructs the multi output Gaussian process model of geographic spatiotemporal data set,which can improve the prediction performance of the model by using the relationship between multiple variables in training;At the same time,the hyperparametric optimization method of Gaussian process model is studied,and a complete set of training sample selection method and dynamic prediction process are designed for the model.Finally,the geographic spatiotemporal data interpolation model based on Gaussian process is introduced into air quality monitoring data set,deformation monitoring data set and water quality monitoring data set for example research.The temporal characteristics,spatial characteristics,spatio-temporal characteristics and multi-attribute correlation characteristics of spatio-temporal data in different spatio-temporal sequence data sets are analyzed in detail.The experimental results show that the geographic spatiotemporal data interpolation model based on Gaussian process can achieve excellent interpolation effect in many cases,and improve the calculation efficiency on the premise of ensuring the accuracy.This model is practical and reliable in practical application.