Identification And Analysis Of Urban Functional Area Based On Improved Word2Vec And POI Data

The effective identification of urban functional areas is of great significance to understand the spatial structure of cities,guide the allocation of resources,and improve the governance mechanism of urban management and planning.POI is an important carrier of location text information for smart city,which has the characteristics of low cost and fast update,and has been widely used in urban functional area identification.Therefore,this paper combines Word2 Vec model with POI data and natural language processing method to identify urban functional areas according to the potential correlation between POI data and regions.The Word2 Vec model only considers the relationship between intermediate words and surrounding words,and does not consider the relationship between intermediate words and the whole document.However,the magnitude of POI data different categories varies greatly,which will result in bias for the extraction of feature vectors of research units.In this paper,TF-IDF algorithm is introduced to improve the Word2 Vec model.By assigning corresponding weights to different categories of POI data and taking full account of the differences among POI data,the extraction results of feature vectors of research units are more reliable.By improving Word2 Vec model and POI data,functional area identification of Changsha built-up area is carried out.Mainly completed work is as follows:Landsat images and NPP-VIIRS data were used to extract the built-up area of Changsha.The first order Rock adjacency method is adopted to calculate the spatial weight matrix to obtain the optimal research element.By using improved Word2 Vec model to process POI data,the functional zoning of Changsha built-up area is obtained.Several analysis methods including statistical calculation,kernel density analysis,spatial autocorrelation analysis,standard deviational ellipse,the mixing degree analysis for land use and urban spatial dynamic analysis are adopted to analyze the characteristics of function distribution of Changsha from the overall distribution function gathering for the distribution characteristics and space,each function,space,shape characteristic,the urban function of mixing degree.This paper points out the problems existing in the current spatial structure of Changsha city and puts forward some optimization suggestions.The main conclusions are as follows:(1)The improved Word2 Vec model can effectively identify urban functional areas,with an overall accuracy of 77.3% and a Kappa coefficient of 0.73.Comparing the recognition results of LDA model and Word2 Vec model,the improved Word2 Vec model used in this paper has higher recognition accuracy for the functional area identification of Changsha built-up area.(2)By analyzing the functional distribution of Changsha built-up areas,this paper finds out that the functional distribution is relatively perfect and reasonable,but there are still some unbalanced functional distribution problems,such as public service functions,which need further targeted construction.(3)Through the analysis of functional mixing degree in the built-up area of Changsha,it is found that the central urban area has a high degree of land use mixing degree and a relatively concentrated functional distribution,while the area near the urban boundary has a low degree of land use mixing degree and a relatively single functional distribution;And it shows an obvious development situation in which the south is strong and the north is weak,and the west is strong and the east is weak.Through the comparative analysis of urban spatial vitality and functional mixing degree,it is concluded that the functional layout of some regions is relatively perfect,but their social vitality is low.In the future,it is suggested to strengthen the guidance of population and human activities to flow to this region,give full play to the responsibility of relieving the pressure of the urban center,and realize the multi-center development mode of Changsha.