A method for identifying urban land vitality patterns and influencing factors
Through non-negative matrix decomposition and asymmetry gradient regression model and other technologies, the accuracy and efficiency problems of traditional methods in urban vitality pattern recognition and influencing factor analysis are solved, and high-precision and rapid urban vitality pattern recognition and influencing factor analysis are achieved. in complex urban vitality system.
Patent Information
- Application Number
- CN202010575606.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-06-22
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2040-06-22
AI Technical Summary
When traditional methods identify urban land vitality patterns and influencing factors, there is low data accuracy and low processing efficiency, making it difficult to quickly analyze high-dimensional spatiotemporal data. The existing correlation analysis method has a narrow scope of application, so it is impossible to effectively capture the complexity of urban vitality system. Nonlinear relationship.
The dynamic features of the dynamic intensity of urban space-time vitality are extracted by non-negative matrix decomposition technology, and the vitality pattern is identified through K-means clustering, and a progressive gradient regression model is constructed to analyze influencing factors. Combined with tools such as XGBoost algorithm and local dependency graph, a number of indicators are discussed. relationship between.
It realizes high-precision and rapid urban vitality pattern recognition and influencing factor analysis, can handle nonlinear relationships, improves analysis efficiency and accuracy, and provides a more detailed urban vitality characteristics and impact mechanism.
Smart Images

Figure CN111898858B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of data processing, and in particular relates to a method for identifying urban land vitality patterns and influencing factors. Background Art
[0002] As China's urbanization process enters a new stage of quality improvement, urban vitality is considered to be one of the important qualities to measure urban quality. Beijing, Shanghai, Shenzhen and many other places have issued corresponding guidelines to promote the development of night economy to enhance urban vitality. The most direct representation of urban vitality is the spatial distribution of crowd activities, which is also affected by various material elements of the urban built environment.
[0003] Therefore, for urban planners, understanding the current status of urban vitality and studying the impact mechanism of planning elements in the urban built environment on urban vitality are of great significance for subsequent planning adjustment measures. The widespread application of urban-related big data also provides new possibilities for urban vitality research. On the one hand, urban planners can use big data to analyze the spatial distribution characteristics and temporal variation characteristics of urban vitality, and have a detailed understanding of the current status of vitality in different areas of the city, from which they can study the distribution phenomenon of urban vitality and locate areas where problems may exist. On the other hand, studying the impact mechanism between planning elements and vitality in the urban built environment can provide a theoretical basis with reference value for the adjustment of subsequent planning measures in different regions, such as adjustments in land use nature, development intensity, road network density, etc.
[0004] The research on urban vitality mechanism can be roughly divided into three parts: static analysis of urban environment, dynamic analysis of urban vitality pattern, and the influence mechanism between the two.
[0005] Traditional methods of obtaining data for urban vitality research include consulting official statistical yearbooks, questionnaires, random interviews, field surveys, etc., which require relatively high time and labor costs. Big data has reduced these costs to a certain extent. Through crawler software, public online maps, mobile phone signaling and other data can be obtained as direct or indirect indicators of large-scale urban built environment or urban vitality. The main defects of urban vitality analysis using big data include: ① The accuracy of the analysis of static characteristics of the built environment in the planning field is relatively low, the research unit is large, and it is mainly used to analyze the overall spatial structure of the city at a large scale, lacking relatively fine static feature distinction. ② In terms of the dynamic characteristics of vitality patterns, traditional urban data processing methods cannot handle such high-dimensional big data, and it is difficult to quickly identify and classify the characteristics of spatiotemporal data. In addition, classification is usually based on prior knowledge of planning for manual scoring and screening, which is time-consuming and inefficient. ③When analyzing the influence mechanism between the two, the traditional method mainly uses Pearson correlation analysis. This correlation analysis method is only applicable to testing whether there is a linear relationship between two sets of data with normal distribution. It has a narrow scope of application. The urban vitality system is complex, and the impact of built environment elements on vitality is not necessarily linear.
[0006] Therefore, an efficient and rapid method to identify the vitality patterns of urban land is needed. Summary of the invention
[0007] To solve the above problems, the present invention discloses a method for identifying the vitality pattern and influencing factors of urban land. First, the static characteristics of high-precision units in a large scale are quickly analyzed, the dynamic characteristics of urban vitality data in the time dimension are mined and clustered according to these characteristics, and an asymptotic gradient regression model of the urban vitality influencing mechanism is constructed to analyze the factors affecting urban vitality.
[0008] To achieve the above object, the technical solution of the present invention is as follows:
[0009] A method for identifying urban land vitality patterns and influencing factors comprises the following steps:
[0010] (1) First define the spatial scope, time scope, and basic spatial unit of the study area;
[0011] (2) Then we will carry out two aspects of work:
[0012] a: Obtain the vector data of urban spatiotemporal vitality intensity, extract the features of urban spatiotemporal vitality intensity data, and perform clustering based on the features;
[0013] b: Get the vector data of urban land use.
[0014] (3) In a unified coordinate system, the data on built environment indicators, activity intensity, and vitality patterns are spatially integrated and merged into the same research unit.
[0015] (4) Construct an asymptotic gradient regression model of the influencing mechanism of urban vitality and analyze the main influencing factors and combination patterns of different vitality modes.
[0016] The method for identifying the vitality pattern and influencing factors of urban land described in the present invention is based on the premise of first defining the spatial scope, time scope, basic spatial unit, etc. of the study area.
[0017] Furthermore, the method for extracting features of urban spatiotemporal vitality data described in step (2) a comprises the following steps: selecting feature days (such as working days, rest days, holidays), and calculating the vitality intensity data of the corresponding feature days, and extracting features of the activity intensity of the vitality data related to the population distribution at multiple times in multiple regions, in this process adding the restriction that all elements are non-negative numbers, and selecting an appropriate number of decompositions through a non-negative matrix algorithm, thereby decomposing the original time data into a feature vector matrix and a feature coefficient matrix, wherein the feature vector matrix is the dynamic feature of urban vitality based on the time dimension, and the extracted feature vector has a practical interpretation meaning according to the distribution characteristics of its vitality intensity, and the feature coefficient matrix is a dimensionality reduction matrix representing the original matrix, and the feature coefficient matrix can be restored to the original matrix by multiplying the feature vector matrix.
[0018] Furthermore, the clustering according to the features described in step (2) a is performed using the K-means unsupervised clustering method, with the feature coefficient matrix as the object, and a suitable k value is selected for relative clustering. The different types of vitality obtained by clustering are used to identify typical urban vitality patterns and their characteristics.
[0019] Compared with the traditional method, step (2) a can quickly extract dynamic features and has realistic interpretation. On the other hand, it can obtain better clustering effect when applying traditional clustering methods through the data after dimensionality reduction.
[0020] Furthermore, the feature extraction of urban spatiotemporal vitality intensity data described in step (2) a refers to relevant data that can directly characterize the absolute number or relative value of urban population distribution, including but not limited to the following methods: mobile phone signaling data at multiple times obtained from mobile phone operators; relative or absolute values of human population distribution maps at different times provided by online maps.
[0021] Furthermore, step (2) b is also carried out in two steps based on the definition of the spatial scope, time scope, and basic spatial unit of the study area:
[0022] 1. If there is direct vector data of urban land use, then combined with step (2) a, the data of built environment indicators, activity intensity, and vitality patterns are spatially fused and integrated into the same research unit under a unified coordinate system.
[0023] 2. In the absence of direct vector data on urban land use patterns, first obtain data related to the spatial form and functional density of urban land or related built environment, including POI, building outlines, road network, land use patterns, etc., and use spatial statistical techniques to calculate relevant numerical indicators such as road network density, building density, floor area ratio, business office function density, functional mix, etc.; then, combined with step (2) a, in the case of a unified coordinate system, spatially fuse the data on built environment indicators, activity intensity, and vitality patterns and integrate them into the same research unit.
[0024] Furthermore, the step (2) b of obtaining the vector data of urban land use patterns also includes the following steps: according to the data cleaning rules, the obtained vitality intensity data is cleaned, including data duplication checking, missing value interpolation or deletion, spatial coordinate system one, deleting data that does not meet the non-negative matrix decomposition conditions (all values are 0), and fusing and associating with the data analysis unit of step (2) a.
[0025] Furthermore, step (4) is based on step (2) a. Under different types of vitality modes, the XGBoost algorithm is used to construct asymptotic gradient regression models under different types of vitality modes, with vitality intensity as the dependent variable and multiple indicators of the built environment as independent variables. Based on the obtained model, an importance ranking diagram of influencing factors, a partial dependency plot and an individual conditional expectation diagram are drawn. By interpreting the result drawings, the spatial characteristics and influencing factors of different urban vitality modes are described.
[0026] Steps (3) and (4) of the present invention are based on the clustering results of step (2). According to different clustering results, the progressive gradient regression model, partial dependency plots and individual conditional expectation plots constructed using the XBGoost algorithm in the field of machine learning are all regression models constructed under the influence of multiple indicators, rather than a single relationship between the two.
[0027] The beneficial effects of the present invention are:
[0028] The non-negative matrix decomposition technology used in the first half of the present invention can quickly and accurately extract fine dynamic features of small scales under a large range, and provide an equivalent matrix after dimensionality reduction for later effective clustering. The extracted features have realistic interpretation due to the non-negative number constraints imposed in the method; the progressive gradient regression model proposed in the second half can be used to test multiple linear or nonlinear correlation relationships between multiple elements, and can analyze the main influencing factors of different vitality modes with higher accuracy. And the use of partial dependency plots can obtain the marginal effect of this indicator on urban vitality when considering the average level of multiple indicators, while the Pearson analysis in the existing correlation technology is only applicable to the test of linear relationship between two elements. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION
[0030] The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention.
[0031] As shown in the figure, the method for identifying the vitality pattern and influencing factors of urban land described in the present invention is divided into two parts based on the definition of the spatial scope, time scope and basic spatial unit of the study area:
[0032] The first part discloses a method for feature extraction of urban spatiotemporal vitality intensity data and clustering based on the features.
[0033] The second part obtains the vector data of urban land use patterns.
[0034] Then, in a unified coordinate system, various indicators are integrated to construct asymptotic gradient regression models under different types of vitality modes. Based on the obtained model, factor importance ranking diagrams, partial dependency plots and individual conditional expectation diagrams are drawn. By interpreting the results drawings, the spatial characteristics and influencing factors of different urban vitality modes are described.
[0035] The first part of the present invention is a method for extracting and clustering vitality data features, comprising the following steps:
[0036] Obtaining vitality intensity data related to the city's crowd distribution, such as mobile phone signaling data and online map vector heat maps; the crowd distribution includes the relative intensity or absolute value of the crowd distribution at different times;
[0037] Select characteristic days (such as working days, rest days, holidays), and calculate the vitality intensity data of the corresponding characteristic days.
[0038] The feature extraction method of vitality intensity data disclosed in the present invention is to extract features from the activity intensity of vitality data related to the distribution of people at multiple times in multiple regions, add the restriction that all elements are non-negative numbers in this process, and select the appropriate number of decompositions through the non-negative matrix algorithm, so as to decompose the original time data into a feature vector matrix and a feature coefficient matrix, wherein the feature vector matrix is the dynamic feature of urban vitality based on the time dimension, the extracted feature vector has a realistic interpretation meaning according to the distribution characteristics of its vitality intensity, and the feature coefficient matrix is a dimension reduction matrix representing the original matrix, and the feature coefficient matrix can be restored to the original matrix by multiplying the feature vector matrix. Using unsupervised clustering methods such as K-means, taking the feature coefficient matrix as the object, selecting an appropriate k value for relative clustering, the different types of vitality obtained by clustering are to identify the typical urban vitality mode and its characteristics, and the characteristics of the vitality mode can be further described through the subsequent process of the present invention; compared with the traditional method, this step can be quickly extracted and has realistic interpretation on the one hand, and on the other hand, better clustering effect can be obtained when applying the traditional clustering method through the dimension-reduced data.
[0039] The characteristics and categories of the vitality patterns of different characteristic days are associated and integrated with the basic spatial unit data.
[0040] In certain embodiments of the first part of the present invention, the acquisition of vitality intensity data is achieved by at least one of the following methods: mobile phone signaling data at multiple times obtained from a mobile phone operator; relative or absolute values of a crowd distribution map at different times provided by an online map.
[0041] In certain embodiments of the first part of the present invention, the vitality analysis method also includes the following steps: performing data cleaning on the acquired vitality intensity data according to data cleaning rules, including data duplication checking, missing value interpolation or deletion, spatial coordinate system one, and deleting data that does not meet the non-negative matrix decomposition conditions (all values are 0).
[0042] The second part of the present invention:
[0043] In the absence of direct vector data on urban land use patterns, data related to the spatial form and functional density of urban land or related built environment are obtained, including POI, building outlines, road network, land use patterns and other data, and spatial statistical techniques are used to calculate relevant numerical indicators such as road network density, building density, floor area ratio, business office function density, functional mix, etc.; in the case of direct vector data on urban land use patterns, the next step of spatial fusion of data can be carried out directly.
[0044] The second part of the present invention further includes the following steps of obtaining vector data of urban land use patterns: performing data cleaning on the obtained vitality intensity data according to data cleaning rules, including data duplication checking, missing value interpolation or deletion, and spatial coordinate system one, deleting data that does not meet the non-negative matrix decomposition conditions (all values are 0), and fusing and associating with the data analysis unit of the first part.
[0045] In the case of a unified coordinate system, the data of built environment indicators, activity intensity, and vitality patterns in the first and second parts are spatially integrated and unified into the same research unit. Taking vitality intensity as the dependent variable and multiple indicators of the built environment as independent variables, an asymptotic gradient regression model is constructed under different types of vitality patterns. Based on the obtained model, a factor importance ranking diagram, a partial dependency plot, and an individual conditional expectation diagram are drawn. Through the interpretation of the results drawings, the spatial characteristics and influencing factors of different urban vitality patterns are described.
[0046] This invention proves the feasibility of feature extraction, clustering and correlation analysis by studying the influence mechanism between urban vitality and built environment indicators within the central ring of Shanghai. In terms of feature extraction, the difference in vitality characteristics between weekdays and weekends is successfully distinguished from the time dimension, and the morning peak and evening peak periods of weekdays are distinguished within the time period of a day; in terms of clustering results, the results distinguish between land use areas dominated by commercial vitality patterns and areas dominated by residential vitality patterns; in terms of correlation analysis, the correlation obtained by the progressive gradient regression model of machine learning has certain practical explanatory significance; based on different clustering results, it is applied to the inspection of Shanghai commercial outlets, and the spatial distribution and composition ratio of the vitality pattern of commercial centers are tested, identifying the development stages of different commercial centers and the differences in typical functions of commercial centers.
Claims
1. A method for identifying urban land vitality patterns and influencing factors, characterized by: The following steps are involved: (1) First define the spatial scope, time scope, and basic spatial unit of the study area; (2) Then we will carry out two aspects of work: a: Extract the characteristics of urban spatiotemporal vitality intensity data and cluster them according to the characteristics; The method for extracting characteristics of urban spatiotemporal vitality data comprises the following steps: selecting characteristic days and calculating the vitality intensity data of the corresponding characteristic days, extracting characteristics of the activity intensity of the vitality data related to the population distribution at multiple times in multiple regions, adding the restriction that all elements are non-negative numbers in this process, and selecting the appropriate number of decompositions through the non-negative matrix algorithm, thereby decomposing the original time data into a characteristic vector matrix and a characteristic coefficient matrix, wherein the characteristic vector matrix is the dynamic characteristics of urban vitality based on the time dimension, and the extracted characteristic vector has a practical interpretation meaning according to the distribution characteristics of its vitality intensity, and the characteristic coefficient matrix is a dimensionality reduction matrix representing the original matrix, and the characteristic coefficient matrix can be restored to the original matrix by multiplying the characteristic coefficient matrix with the characteristic vector matrix; The said urban spatiotemporal vitality intensity data feature extraction includes obtaining the vitality intensity data of the city, which refers to the relevant data that can directly characterize the absolute number or relative value of the distribution of the urban population, including but not limited to the following methods: mobile phone signaling data at multiple times obtained from mobile phone operators; relative or absolute values of the human distribution map of the population at different times provided by online maps; b: Get the vector data of urban land use; (3) In the case of a unified coordinate system, the data on built environment indicators, activity intensity, and vitality patterns are spatially integrated and unified into the same research unit; (4) Construct a progressive gradient regression model of the influencing mechanism of urban vitality and analyze the main influencing factors and combination patterns of different vitality modes; On the basis of step (2) a, under different types of vitality modes, the XGBoost algorithm is used to construct progressive gradient regression models under different types of vitality modes, with vitality intensity as the dependent variable and multiple indicators of the built environment as independent variables. Based on the obtained model, the importance ranking diagram of influencing factors, the local dependence diagram and the individual conditional expectation diagram are drawn. The spatial characteristics and influencing factors of different urban vitality modes are described by interpreting the importance ranking diagram of influencing factors, the local dependence diagram, the individual conditional expectation diagram, the spatial distribution diagram of vitality modes, the frequency distribution diagram of the built environment and the integration of the results based on these diagrams.
2. A method for identifying urban land vitality patterns and influencing factors according to claim 1, characterized in that: The clustering according to features described in step (2) a is to use the K-means unsupervised clustering method, take the feature coefficient matrix as the object, select the appropriate k value for relative clustering, and the different types of vitality obtained by clustering are used to identify the typical urban vitality pattern and its characteristics.
3. The method for identifying the vitality pattern and influencing factors of urban land according to claim 1, characterized in that: Step (2) b is also based on the definition of the spatial scope, time scope, and basic spatial unit of the study area, and is carried out in two steps: a. If there is direct vector data of urban land use, then combined with step (2) a, the data of built environment indicators, activity intensity, and vitality patterns are spatially fused and integrated into the same research unit under a unified coordinate system; b. In the absence of direct vector data on urban land use patterns, first obtain data related to the spatial form and functional density of urban land or related built environment, including POI, building outlines, road networks, and land use patterns, and use spatial statistical techniques to calculate relevant numerical indicators; then, combined with step (2) a, spatially fuse the data on built environment indicators, activity intensity, and vitality patterns in a unified coordinate system and integrate them into the same research unit.
4. The method for identifying the vitality pattern and influencing factors of urban land according to claim 1, characterized in that: The step (2) b of obtaining the vector data of urban land use patterns also includes the following steps: according to the data cleaning rules, the obtained vitality intensity data is cleaned, including data duplication checking, missing value interpolation or deletion, spatial coordinate system 1, deleting data that does not meet the non-negative matrix decomposition conditions, and fusing and associating with the data analysis unit of step (2) a.
Citation Information
Patent Citations
Method for identifying and predicting bus passenger flow influence factor based on geographically and temporally weighted regression
CN107103392A
Urban land function identification and change detection method
CN109325085A
A non-negative tensor decomposition city dynamic analysis method based on context awareness
CN109684604A
Restaurant heat analysis method for ultra-large city
CN109857983A
