The Application Of Spatial-temporal Data Model In The Population Movement

Population migration had produced a large amount of historical data, how these data canbe accurately used in the policy-oriented study is important. Given theory has shown that thecloser two areas distance is, the more correlated their population are. However, previous stud-ies only focused on the redistribution of population in time dimension or just in the field ofspatial statistical analysis. This paper took both space and time into consideration. Regardingtwo types of spatial-temporal data: continuous data and area data, we respectively apply spa-tial-temporal error model and kriging model.In theory, this paper has done the following things: introducing the spatial-temporal datatypes, the assumptions and premises in modeling, their form and basic parameters estimation,iterative method, likelihood ratio test and forecast methods. In application, it used the abovemethod, by using the R software to realize the entire calculation process. In the empirical part,the initial description of data is good for master its distribution in space and time dimensions.According to the characteristics of the first order differential of the data, it is reasonable to usespatial error model. During the model fitting and testing, we find the space and time-dependent coefficients are significant. Therefore, these results back up the spatial dependencetheory by providing the mathematical aspects, and then it interpreted the actual meaning. Bycomparing the predictive accuracy and computational efficiency, we found that regarding thisproblem, kriging method is more preferable. Above content provided some ideas for analyzingthe spatial-temporal data model.This empirical data is Sweden’s population data due to its high spatial resolution. It ismore accurate than provincial data. Data quality contributes to fit the model. On the otherhand, it has consistent integrity and segmentation in the time dimension. But the majority ofdomestic demographic data is based on province and city. In view of the availability of datacollecting, the limited spatial resolution and the hard work of spacing our domestic data, weselect Sweden population which can be directly used in analyzing spatial and temporal data.As long as the data quality is high enough, or the completed preprocessing of spatial data, ourresearch is worthwhile in the analysis of spatial-temporal data modeling.