| Different functional areas are the carriers for the effective performance of various functions of the city,which are of great significance for regulating the urban land policy and urban land development,and detecting the problems faced in the process of urban development.Therefore,the mining of urban functional areas has become an important content in urban calculations.The various economic activities of humans in cities are related to different types of functions.This localization of economic activities contributes to the functionality of urban areas.At the same time,the function of an area is not single.For example,there are still restaurants,entertainment facilities,and commercial facilities in the cultural and educational area.Therefore,an area can be expressed by the distribution of functions.The use of people’s daily activity data,that is,crowd-sourced geographic data,can mine function distribution at a fine scale and provide factual and theoretical support for the study of urban spatial structure.Existing researches on function division based on urban crowdsourced spatiotemporal data have the following problems: 1)Due to the modifiable areal unit problem,it is difficult to determine a suitable spatial analysis unit.Spatial units of different sizes will affect the scale of function recognition;2)Most studies divide urban functional areas into a single functional type,and lack quantitative analysis of the strength of mixed functions in the area;3)The functional semantic information implicit in spatiotemporal data is not fully explored,and most studies only analyze its frequency distribution pattern from the perspective of data volume.Moreover,it is difficult to mine the topic information in multidimensional feature data in previous research.These problems reduce the reliability of the results of urban function division.Therefore,in response to the above-mentioned problems,this article conducts research on the division of urban functional areas from the establishment of spatial analysis units,the quantitative mining of functional topic information,and the calculation of multidimensional feature data,and provides new ideas and methods for the calculation of spatiotemporal data and the analysis of urban functional structure.The main research contents and conclusions of this article are as follows:(1)Optimization framework of spatial analysis unitThe establishment method of spatial analysis unit for spatiotemporal data is studied.Through the analysis of the distribution pattern of spatiotemporal data,an optimal spatial analysis unit optimization framework that can evaluate candidate analysis units based on multiple criteria is proposed.This framework relies on Pareto optimality to select spatial analysis units,thereby overcoming the subjectivity and randomness of traditional spatial analysis unit settings,and also alleviating the influence of MAUP to a certain extent.The framework is used to analyze the floating car trajectory data,and the analysis unit that optimizes the spatial autocorrelation mode is determined by combining the global spatial autocorrelation index and the local spatial autocorrelation coefficient of variation.The results show that 700 m and 800 m are the best spatial analysis units in the experimental area.The multistandard spatial analysis unit optimization framework can provide spatial analysis scales for the study of geographic phenomena such as the division of urban functions.(2)Function topic information extraction modelThis paper studies the method of functional information mining based on semantic topic,and applies the traditional topic extraction model to spatiotemporal data computing to get the specific topic of the data.For the spatiotemporal data with diverse topic information and complex structure,a topic probability generation model based on histogram(H-LDA)is proposed.By comprehensively considering the relationship between urban residents’ travel behavior and urban function,the large amount of floating car data is applied to the H-LDA model,and the potential functional semantic information of the study area is extracted from the travel trajectory data.Experimental results show that the accuracy of function distribution based on the H-LDA model is 88.1%.At the same time,the correlation coefficient between the functional information entropy(FIE)based on POI data and the topic information entropy(TIE)based on H-LDA model is0.612,which indicates that they are highly correlated.All these explain the accuracy and rationality of the distribution results of urban function and mixed use based on H-LDA model.This model lays a foundation for the measurement of similarity functions.(3)Data clustering method considering multidimensional featuresThis paper studies a data clustering method which can deal with multidimensional features.By introducing Gauss kernel function to calculate the similarity of multidimensional features,the problem of linear indivisibility of multidimensional attributes is solved.The adaptive parameter setting strategy is applied to improve the convergence speed of the algorithm.Finally,a multidimensional feature affinity propagation clustering algorithm(MDF-AP)which can analyze spatiotemporal big data quickly and accurately is proposed.The MDF-AP algorithm is applied to public data sets to verify the effectiveness of the algorithm.At the same time,clustering the spatiotemporal data of the floating car,it is found that the MDF-AP algorithm has an average matching point accuracy(ARP)of 81.47% within a buffer radius of 1000 m,which is better than the clustering results of the comparison algorithms.It shows that MDF-AP algorithm is effective in clustering multidimensional spatiotemporal data.Through the clustering algorithm of multidimensional spatiotemporal data,the comprehensive analysis of multidimensional features such as spatial,temporal,attribute,etc.is completed,so as to more accurately extract the information implicit in the multidimensional feature data,which is of great significance for realizing the division of urban functional areas based on multidimensionality.(4)Multidimensional urban functional area division experimentTaking the 3rd Ring Road of Wuhan City as the research area,the optimal spatial analysis unit is established by using the floating car trajectory data and the global and local correlation standards.Then the urban spatial function and the degree of mixed use are calculated by using the function topic information extraction model and multidimensional feature data clustering method.At the same time,a comparison was made with the urban function and mixed use calculated by the POI data.The experimental results show that: 1)The optimal analysis unit of floating car data on weekdays and weekends is 1000 m;2)The probability of functional topics in the study area is divided into 8 cluster categories,and the mixed use of functional topics in the experimental area is high;3)Compared with the street view map,it is found that the urban functions based on floating car data can identify roads and bridge areas,which is difficult for POI data without length and other attribute information.At the same time,comparing FIE with TIE,it is found that the mixed use of urban functions calculated based on floating car data is higher.The above analysis proves that the urban functional areas divided from the dimensions of spatial,temporal and characteristics are in line with the actual situation of Wuhan.The method proposed in this paper can provide accurate and reasonable quantitative results for the identification of urban functions and mixed use. |
PDF Full Text Request