Research On Method And Application Of Guangxi Population Spatialization Based On Geographic Big Data And Machine Learning

Population is an important basic data in social economy,human geography and other fields.Most of the traditional population data are based on the statistics of administrative regions,which does not have accurate spatial positioning characteristics,which has a negative impact on the accurate research of regional economy and other related fields.The refined population spatial distribution data can be used as the basic data in the coupling analysis of natural resources and social factors,and applied in urban planning,regional economic development,urban living environment,disaster prevention and relief,etc.The arrival of big data and artificial intelligence era provides method support for population spatialization.Therefore,based on geographic big data and machine learning method,this paper proposes a method of fine simulation of urban and rural population.So as to improve the accuracy of the population data set,in order to be better used as a population factor variable in geographic application analysis.This paper collected and processed the data of 1152 township-level administrative boundaries in Guangxi in 2017,the township population data in the 2017 China County Statistical Yearbook,and the multi-geographic spatial data of Guangxi in 2017,and the statistical density of the township-level in Guangxi was obtained by calculation.Using correlation analysis and geographic detector methods,11 spatial variables that affect the spatial distribution of the population in the study area are extracted.Using meta-linear regression,random forest,and convolutional neural network methods to construct a population spatialization model,compare the accuracy of the three population spatial methods and select the best population spatialization model.The optimal population spatialization data is applied to analyze the population space,and the characteristics of the population spatialization in Guangxi and the accessibility and fairness of medical resources in Nanning are respectively explored.The research results of this paper are as follows:(1)Research shows that the proportion of impervious area,average night light,poi and population density in Guangxi are positively correlated,and proportion of forest area,slope,elevation and population density in Guangxi are negatively correlated.Impervious surfaces,night lights,and POI are the three important human factors that affect the population distribution in Guangxi,and the interaction between the factors strengthens the impact on the population distribution.(2)The accuracy of population spatial data is not only related to the accuracy of the fitting data,but also affected by the method used for modeling.Overall,the average relative error based on the random forest method is the lowest(25.04%),followed by the average relative error based on the convolutional neural network method(27.68%),and finally the average relative error based on the multiple regression method(29%).Compared with the existing population data set,it is found that the accuracy of the population spatialization data based on random forest is better than that of worldpop and Land Scan data.(3)Guangxi’s population space mainly presents the following four characteristics:(1)The population density of Guangxi The average density of the total population is 192 persons/km2,but the population density is unevenly distributed in regions.(2)The high-density population is distributed in the urban center of southeast Guangxi,and the low-density population is distributed in the rural areas of northwest Guangxi.(3)The spatial pattern of urban population is clustered and linear,and rural settlements are scattered.(4)The population at different scales in Guangxi has spatial agglomeration characteristics.(4)There are certain differences and imbalances in medical service facilities in Nanning.The allocation of medical service facilities and the matching degree of population and medical services all present a circle structure."High enjoyment type" and "general type" are mainly distributed in cities and county areas.The "lagging type" is mainly distributed in suburban and rural areas.These studies have successfully provided new ideas for the accessibility and fairness of urban resources and the allocation of government planned medical resources.