A tourist flow data analysis method based on GPS big data

Abstract

The present invention relates to the field of tourism flow technology, specifically a tourism flow data analysis method based on GPS big data, including an invention method and data source and processing. The invention method includes regression analysis, GIS spatial analysis, a tourism flow node-corridor network construction method and an optimal parameter-based geographic detector (OPGD) analysis, the invention data source and processing include data source and data processing, the invention data source includes GPS trajectory data and other data, and the data processing includes data validity verification and data preprocessing. Compared with the prior art, the present invention relies on tourists' GPS trajectory data to characterize the spatial pattern of tourism flow and GIS tools, and can systematically analyze the concentration and sparseness characteristics and hot and cold spots of the spatial distribution of tourism flow, cut in from the perspective of nature and humanity, extract the key influencing factors of the dimensions of natural geography, social economy, tourism resources, and tourism services, and use the geographic detector model based on optimal parameters to deeply explore the influence mechanism of each factor on the spatial heterogeneity of tourism flow, and provide a scientific basis for the high-quality development of tourism.

Description

Technical Field

[0001] The present invention relates to the field of tourist flow technology, in particular to a tourist flow data analysis method based on GPS big data. Background Art

[0002] As a spatial organization form that reflects the dynamic spatial and temporal distribution characteristics of tourists, tourist flows play an important role in deconstructing tourism phenomena and tourist behavior patterns. The spatial layout of tourist flows can reflect the spatial organization form and correlation of tourist elements in tourist destinations, and reveal the flow and changes of information, materials and energy in the process of tourist flow. The invention of factors affecting tourist flows helps to identify the driving factors for the formation of tourist flow spatial patterns, clarify the spatial relationship and action mechanism between the source and destination, and help to identify the driving factors for the formation of tourist flow spatial patterns, and clarify the spatial relationship and action mechanism between the source and destination.

[0003] The existing technology has made certain progress in the spatiotemporal pattern and impact of tourism flows. However, in terms of data selection, there is a relative lack of data capture directly based on tourists' mobile GPS trajectories. At the same time, there are few applications for large-scale destinations (such as inter-provincial scales). A set of systematic method paths for relevant data capture, spatial visualization and spatial statistical analysis with reasonable matching and good connection has not been found, and the combination of other human and natural factors has been neglected. Overall, there are certain bottlenecks in the analysis of tourism flows for large-scale destinations such as climate conditions, natural resources, and tourism development.

[0004] Therefore, in order to solve the above problems, this application proposes a tourist flow data analysis method based on GPS big data, relying on tourists' GPS big data trajectories as the data basis, GIS analysis of spatial distribution characteristics, hot and cold spot analysis and multivariate clustering methods to construct a tourist flow node-corridor network, combining key influencing factors in natural, socio-economic, resource services and other dimensions, and forming a mechanism through methods such as geographic detector models, which provides a scientific basis for tourist flow data analysis and provides a scientific basis for regional tourism management and the development of tourist routes. Summary of the invention

[0005] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a tourist flow data analysis method based on GPS big data. It relies on tourists' GPS big data trajectories as the data basis, GIS analyzes the spatial distribution characteristics, combines the key influencing factors of natural, social economy, resource services and other dimensions, and forms a mechanism through methods such as geographic detector models, which provides a scientific basis for the data analysis of tourist flows.

[0006] In order to achieve the above-mentioned purpose, the present invention provides a tourist flow data analysis method based on GPS big data, including research methods and data sources and processing, as well as a systematic technical path for spatial visualization application;

[0007] The research methods include regression analysis, GIS spatial analysis, tourist flow node-corridor network construction method and geographic detector OPGD analysis based on optimal parameters;

[0008] Regression analysis: Zipf regression is used to measure the balance of the scale of urban tourism flow. The formula is:

[0009] P i =AR i -α (1)

[0010] Taking the natural logarithm of both sides of formula 1 yields:

[0011] LqCy i =-αlnR i +A (2)

[0012] Ri is the rank of the ith city, invention Pi is the scale of tourist flow of the ith city, invention A and invention α are constants;

[0013] α is the Pareto index of the scale distribution of urban tourism flows, and its size can be used to measure the degree of balance of scale distribution; with α = 1 as the benchmark, α < 1 indicates that the scale distribution of urban tourism flows is relatively dispersed, that is, the development of low-rank cities is more prominent, and the development of large cities is insufficient; on the contrary, α > 1 indicates that the scale distribution of urban tourism flows is more concentrated, that is, the development of high-rank cities is more prominent, and the development of small and medium-sized cities is insufficient;

[0014] GIS spatial analysis is as follows: using overlay display, standard deviation ellipse, and buffer analysis to discover the spatial distribution characteristics of different elements and tourist flows; using kernel density analysis to explore the concentration and sparseness characteristics of the spatial structure of tourist flows; using hot and cold spot analysis to extract the spatial hot spots of tourist flows; using spatial autocorrelation analysis, namely the Moran index, to discover the correlation characteristics of tourist flows;

[0015] The kernel density analysis is as follows:

[0016] Describe the spatial distribution characteristics of regional GPS tourism trajectories and preliminarily judge the concentration or dispersion of the spatial distribution of tourism flows. The calculation formula is as follows:

[0017]

[0018] f(j) is the density of point j, invention r represents the search radius, invention dij is the distance from point i to point j, invention k is the weight of dij;

[0019] The spatial autocorrelation analysis is:

[0020] The global Moran's I and local Moran's I are used to measure the global and local autocorrelation of tourist flows, determine the overall spatial agglomeration characteristics of tourist flows, and the specific spatial locations of tourist flow agglomeration and dispersion. The calculation formula is:

[0021]

[0022]

[0023] n is the number of spatial units, invention xi, xj is the mean of the observed values ​​of GPS trajectories of tourist flows in spatial units i and j, i≠j, invention is the mean of the tourist flow intensity of all spatial units, wij is the spatial weight matrix, S2 is the variance of the tourist flow intensity; the value range of Moran's I is [-1,1]. If the value is greater than 0 and significant, it indicates that there is positive spatial autocorrelation;

[0024] Hot and cold spot analysis is:

[0025] The Getis-Ord Gi* method is used to identify the local spatial agglomeration areas of GPS tourist trajectories. The hotspot area represents the high-value agglomeration area of ​​tourist flow, and the cold spot area represents the low-value agglomeration area of ​​tourist flow. The calculation formula is:

[0026]

[0027]

[0028]

[0029] x j is the mean value of the observed value of GPS trajectory of tourist flow in spatial unit j, w i,j is the spatial weight between spatial units i and j, n is the number of spatial units; the final calculated The value is the z score; when the confidence level is high, that is, 90%, 95%, or 99%, when the z value is positive and large, it is a high-value cluster area, that is, a hot spot, otherwise it is a low-value cluster area, that is, a cold spot;

[0030] The method for constructing the tourist flow node-corridor network is as follows:

[0031] Referring to the point-axis theory, the spatial pattern of tourism flow is divided into two parts: nodes and corridors. Nodes represent points with rich tourism resources and greater tourism appeal in the tourism flow; corridors represent the main spatial channels between different nodes.

[0032] For the construction of nodes, hot and cold spot analysis was used to identify the global GPS trajectory data, and areas with a confidence level of more than 95% were selected. The main tourist nodes were selected in combination with the popular tourist cities (prefecture-level and county-level) and scenic spots in the region. The total number of path points within the 30km buffer zone of the node was taken as the tourism intensity of the node.

[0033] For the construction of corridors, a multivariate clustering method is used, and an unsupervised machine learning method is used to find non-spatial natural clusters of features based on feature attribute values. The selected clustering attribute is the number of trajectory points represented by each point, and the clustering method is K-means. After many attempts, the appropriate number of clusters is determined, and the clustered data is classified into 1-n levels according to the number of trajectory points represented from small to large, and the levels with fewer trajectory points are removed until a clearer tourism flow network grid is obtained. According to the extracted tourism flow nodes, the network generated by the clustering results is converted into vector lines and corrected to make them smooth and connected.

[0034] For the corridor connecting nodes A and B, the calculation method of its tourist flow intensity TAB is:

[0035]

[0036] Among them, R1, R2, ... Rn are the clustering levels represented by each grid intersecting the corridor space. The higher the level, the more trajectory points it represents and the higher the intensity of tourist flow.

[0037] Optimal Parameters Geographic Detector (OPGD) analysis:

[0038] First, the factor detection module of the geographic detector based on the optimal parameters is used to analyze the spatial relationship between the explained variable and the explanatory variable, and to determine the explanatory power of the explanatory variable on the explained variable, that is, the degree of influence. The calculation formula is:

[0039]

[0040] L is the stratification or partition of the independent variable and the dependent variable, and Nh and N represent the number of units in the hth layer and the whole area respectively. and σ 2 are the variances of the Y values ​​of the hth layer and the whole area respectively. Invention q is a statistic with a value range of [0,1]. The larger the q value, the stronger the explanatory power of the factor.

[0041] Secondly, the interaction detection module of the geographic detector is used to identify the interactions between different factors affecting tourism flows and to determine whether different factors, when acting together, will affect the explanatory power of the dependent variable.

[0042] Based on the results of interaction detection, the interaction effect variation ratio index IEVR is further used for quantitative analysis, and its formula is:

[0043]

[0044] xi, xj represent the i-th and j-th influencing factors; q(xi) and q(xj) represent the q values ​​of the i-th and j-th influencing factors; n represents the total number of influencing factors; the value of IEVR is used to measure the average variation ratio of the interaction effect and the individual effect of a certain independent variable. IEVR>0 means that the interaction effect of the independent variable is stronger than the individual effect. The larger the value, the more obvious it is, and vice versa.

[0045] Data source and processing include data source and data processing;

[0046] The data sources include GPS track data and other data. The GPS track data is obtained from the track data uploaded by users of various websites. Other data include spatial geographic data and statistical data. Spatial geographic data includes raster geographic data and vector geographic data. Statistical data comes from official websites, specifically including four dimensions: natural geographic conditions, population and economic conditions, infrastructure construction and tourism development potential.

[0047] Data processing includes:

[0048] Data validity verification: Kernel density analysis is performed on the data obtained from the invention method with the invention GPS trajectory data and other data, and the statistics of the kernel density and grid correlation of the two sets of data are calculated;

[0049] Data preprocessing: Two scales are used for spatial analysis: grid scale and administrative district scale. The grid scale can reflect the spatial differentiation of tourism flow more finely, which is conducive to the identification of hot and cold spots and node-corridors. The administrative district scale can be combined with socio-economic indicators to reflect the intensity of tourism flow in each city and its influencing factors. For these two scales, the GPS trajectory raster data is first preprocessed:

[0050] A. Calculation of tourist flow intensity based on uniform grids:

[0051] The fishing net tool of ArcGIS was used to divide the tourist area into 5km×5km uniform grids. The number of path points represented by the GPS track grid in each grid was counted as the tourist flow intensity of the grid, and the uniform tourist flow grid with a resolution of 5km in the whole tourist area was obtained.

[0052] B. Calculation of tourist flow intensity based on administrative districts:

[0053] For administrative regions with a specific shape, the following formula is used to calculate the intensity of tourist flows in the administrative region, i.e., city, state, or county:

[0054]

[0055] Invention I is the tourism flow intensity of the administrative area, invention j is the number of GPS track grids within the administrative area, invention nj is the value of each GPS track grid within the administrative area, that is, the number of path points within the grid, and invention S is the unit area of ​​the GPS track grid.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] Relying on tourists' GPS trajectory data to depict the spatial pattern of tourism flow and GIS tools, we can systematically analyze the concentration and sparseness characteristics of the spatial distribution of tourism flow and hot and cold areas. From the perspective of nature and humanity, we extract the key influencing factors of the dimensions of natural geography, social economy, tourism resources, and tourism services, and use the geographic detector model to deeply explore the impact mechanism of each factor on the spatial heterogeneity of tourism flow, providing a scientific basis for the high-quality development of tourism. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 Schematic diagram of administrative divisions, topography and land coverage of Gansu Province and Qinghai Province according to an embodiment of the present invention.

[0059] Figure 2 It is a schematic diagram of calculating the tourist flow intensity of the tourist corridor of the present invention.

[0060] Figure 3 This is a schematic diagram of the calculation of tourist flow intensity based on administrative districts in the present invention.

[0061] Figure 4 This is a schematic diagram of the spatial distribution of the tourism flow network of the present invention.

[0062] Figure 5 It is a schematic diagram of the regression analysis fitting results and the Moran index calculation results of the present invention.

[0063] Figure 6 It is a schematic diagram of the tourist flow kernel density and standard deviation ellipse of the present invention.

[0064] Figure 7 This is a schematic diagram of the hot and cold spot analysis results of the 5km grid of tourist flow in the present invention.

[0065] Figure 8 This is a schematic diagram of the main tourism network in Gansu and Qinghai regions of the present invention.

[0066] Fig. 9 This is a schematic diagram of the tourism service facilities, road network distribution and GPS track distribution of the present invention.

[0067] Fig.10 This is a schematic diagram of the distribution of A-level scenic spots, nature reserves and GPS tracks of the present invention.

[0068] Fig.11This is a schematic diagram of the road network distribution and GPS track distribution of the present invention.

[0069] Fig.12 Schematic diagram of the relationship between the number distribution of trajectory points and main roads in the present invention.

[0070] Fig.13 Schematic diagram of the interaction effects of the main factors of the present invention and their types.

[0071] Fig.14 This is a schematic diagram of the IEVR calculation results of the present invention. DETAILED DESCRIPTION

[0072] The present invention will now be further described with reference to the accompanying drawings.

[0073] See also Figures 1 to 13 The present invention provides a tourist flow data analysis method based on GPS big data. In view of the fact that large-scale destinations are less used, this embodiment uses GIS analysis methods to describe the spatial distribution characteristics of tourist flows in large-scale tourist areas based on rasterized GPS tourist flow data, and uses geographic detectors and other methods to reveal the formation mechanism of the spatial structure of tourist flows, and puts forward policy recommendations for optimizing the spatial structure of tourist flows and resource allocation.

[0074] Gansu Province and Qinghai Province are located in the northwest of my country, between 89°35′~108°46′ east longitude and 31°36′~42°57′ north latitude. They are located in the upper reaches of the Yangtze River, Yellow River and Lancang River. They are the key place and golden section of the ancient Silk Road.

[0075] The Gansu-Qinghai region has a vast territory, spanning two major topographic units, the Loess Plateau and the Qinghai-Tibet Plateau. It has beautiful and diverse natural landscapes, profound cultural and historical heritage, and distinct ethnic cultural characteristics. It has formed tourist areas dominated by distinctive cultures such as the ancestral culture of southeastern Longdong, the Dunhuang culture of the Hexi Corridor, and the Yellow River urban culture, as well as tourist corridors based on axes such as the Silk Road, the Ancient Tang-Tibet Road, and the Qaidam Cultural Corridor, as well as domestically and internationally renowned tourist destinations such as the Mogao Grottoes in Dunhuang, the Danxia Mountain in Zhangye, the Qinghai Lake, and the Sanjiangyuan.

[0076] Regression Analysis:

[0077] The distribution of route points in cities (prefectures) in Gansu and Qinghai can reflect the scale of tourism flows in each city to a certain extent. The rank-scale theory can be used to analyze differences in tourism scale. This paper uses Zipf regression to measure the balance of tourism flow scale in cities in Gansu and Qinghai. The formula is:

[0078] P i =AR i - α Formula 1

[0079] Taking the natural logarithm of both sides of the formula yields:

[0080] LqCy i =-αlnR i +A Formula 2

[0081] Among them, Ri is the rank of the i-th city, Pi is the scale of tourism flow of the i-th city, and A and α are constants. α is the Pareto index of the scale distribution of urban tourism flow, and its size can be used to measure the degree of balance of scale distribution. Taking α=1 as the benchmark, α<1 indicates that the scale distribution of urban tourism flow is relatively scattered, that is, the development of low-rank cities is more prominent, and the development of large cities is insufficient; on the contrary, α>1 indicates that the scale distribution of urban tourism flow is more concentrated, that is, the development of high-rank cities is more prominent, and the development of small and medium-sized cities is insufficient.

[0082] GIS Spatial Analysis:

[0083] Spatial analysis is the process of obtaining knowledge by explicitly using the spatial location and spatial relationship of each object's observation value through a series of techniques and models. GIS spatial analysis aims to establish an understanding of the distribution of spatial data, quantitatively describe and statistically test the data distribution pattern, thereby showing the relationship between different elements in space and revealing the distribution law of data.

[0084] The present invention mainly adopts overlay display, standard deviation ellipse and buffer zone analysis to invent the spatial distribution characteristics of different elements and tourist flow, adopts kernel density analysis to explore the collection and sparse characteristics of the spatial structure of tourist flow, adopts hot and cold spot analysis to extract the spatial hot spots of tourist flow, and adopts spatial autocorrelation analysis (Moran's index) to invent the correlation characteristics of tourist flow.

[0085] Kernel density estimation:

[0086] The kernel density method is a non-parametric estimation method for the probability density function of random variables in the measurement space, which reflects the density characteristics and agglomeration differences of the spatial distribution of the invention object. The present invention uses the kernel density method to characterize the spatial distribution characteristics of GPS tourist trajectories and preliminarily judge the agglomeration or dispersion trend of the spatial distribution of tourist flows. The kernel density calculation formula is as follows:

[0087]

[0088] Where f(j) is the density of point j, r represents the search radius, dij is the distance from point i to point j, and k is the weight of dij.

[0089] Spatial Autocorrelation:

[0090] Global spatial autocorrelation can measure the clustering, dispersion or random changes in the distribution of regional attribute values ​​and determine its overall spatial correlation characteristics. Local spatial autocorrelation reflects the local clustering or dispersion effects of attribute values ​​by measuring the spatial correlation characteristics of regional attribute values ​​in local areas [i]. The present invention uses the Global Moran's I and Local Moran's I to measure the global and local autocorrelation of tourist flows, respectively, to determine the overall spatial agglomeration characteristics of tourist flows, as well as the specific spatial locations of tourist flow aggregation and dispersion. The calculation formula is:

[0091]

[0092]

[0093] Where n is the number of spatial units, xi and xj are the mean values ​​of the GPS trajectories of tourist flows in spatial units i and j, i≠j, is the mean of the tourist flow intensity of all spatial units, wij is the spatial weight matrix, and S2 is the variance of the tourist flow intensity. The value range of Moran's I is [-1,1]. If the value is greater than 0 and significant, it means that there is positive spatial autocorrelation.

[0094] Getis-Ord Gin (hot spot analysis) method:

[0095] The Getis-Ord Gi* (hot spot analysis) method is a spatial statistical method for identifying local spatial associations and heterogeneous units, and reflects the local spatial agglomeration or discrete state by evaluating the z-score of the attribute value. The present invention uses the Getis-Ord Gi* method to identify the local spatial agglomeration area of ​​the GPS travel trajectory. The hot spot area represents the high-value agglomeration area of ​​the tourist flow, and the cold spot area represents the low-value agglomeration area of ​​the tourist flow. The calculation formula is:

[0096]

[0097]

[0098]

[0099] In the formula, x j is the mean value of the observed value of GPS trajectory of tourist flow in spatial unit j, w i,j is the spatial weight between spatial units i and j, and * is the number of spatial units. The value is the z score. Generally speaking, under the condition of high confidence level (90%, 95%, 99%), when the z value is positive and large, it is a high-value cluster area (hot spot), otherwise it is a low-value cluster area (cold spot) (Baruah et al., 2022). With reference to the existing invention, the area of ​​the present invention, and the visualization effect of the data, the grid unit size is set to 10km.

[0100] Tourism flow node-corridor network construction:

[0101] Kernel density analysis and hot and cold spot analysis indicate that the tourism flow in Gansu and Qinghai has obvious local hot spots and network connectivity structure, and has the basic form of point-axis development mode. Therefore, the spatial pattern of tourism flow in Gansu and Qinghai can be divided into two parts: nodes and corridors. Among them, nodes represent points with rich tourism resources and greater tourism attraction in the tourism flow; corridors represent the main spatial channels between different nodes.

[0102] For the construction of nodes, hot and cold spot analysis was used to identify the global GPS trajectory data, and areas with a confidence level of more than 95% were selected. The main tourist nodes were selected in combination with the popular tourist cities (prefecture-level and county-level) and scenic spots in the region. The total number of path points within the 30km buffer zone of the node was taken as the tourism intensity of the node.

[0103] For the construction of corridors, a multivariate clustering method was used, and the number of appropriate clustering methods was determined to be 7. The 7 types of clustered data were classified into 1-7 levels according to the number of trajectory points represented from small to large, and the first level (the number of trajectory points was less than 824) was removed to obtain the tourism flow network grid in Gansu and Qinghai. According to the extracted tourism flow nodes, the network generated by the clustering results was converted into vector lines and corrected to make them smooth and connected.

[0104] The natural breakpoint method is used to divide the tourism flow intensity of tourism nodes and routes into five levels, and the main tourism network in Gansu and Qinghai is obtained. From level one to level five, the tourism intensity of nodes or the connectivity of corridors gradually decreases.

[0105] GeoDetector:

[0106] Geographic detectors are statistical methods for detecting spatial heterogeneity and revealing the driving forces behind it. They are widely used in analyzing the evolution of geographic element patterns and regional spatial heterogeneity.

[0107] The present invention first uses the factor detection module of the geographic detector to analyze the spatial relationship between the explained variable (tourism flow intensity in 74 districts and counties) and the explanatory variable (influencing factor of tourism flow), and determines the explanatory power of the explanatory variable on the explained variable, that is, the degree of influence. Among them, the tourism flow intensity of each district and county is expressed by the total number of GPS track points divided by the regional area. Secondly, the present invention uses the interaction detection module of the geographic detector to identify the interaction between different influencing factors of tourism flow, and determines whether different factors such as X1 and X2 will affect the explanatory power of the dependent variable when they work together. The principle of the geographic detector method is to detect the spatial distribution trend of independent variables (such as X1, X2) and dependent variables (Y) through spatial heterogeneity, and then measure the explanatory power of independent variables on the spatial differentiation of dependent variables. The detection result is expressed by the q statistic, and its calculation formula is:

[0108]

[0109] Where L is the stratification or partition of the independent and dependent variables, Nh and N represent the number of units in the hth layer and the whole area, respectively. and σ 2 are the variances of the Y values ​​of the hth layer and the entire area respectively. The value range of q is [0,1]. The larger the q value is, the stronger the explanatory power of the factor is.

[0110] At the same time, the present invention mainly adopts two models of factor detection and interactive detection in geographic detectors. Among them, the q value of factor detection can be used as a standard for evaluating the partition effect. The larger the q value, the better the effect. In previous inventions based on geographic detectors, partitioning is often empirical and lacks accurate quantitative evaluation. The detection results obtained may not be consistent with the actual situation. The present invention uses a geographic detector (OPGD) model based on optimal parameters for classification and detection, thereby obtaining accurate quantitative evaluation results.

[0111] Based on the results of interaction detection, the present invention further uses the interaction variation ratio index (IEVR) for quantitative analysis. The formula is:

[0112]

[0113] Among them, xi and xj represent the i-th and j-th influencing factors; q(xi) and q(xj) represent the q values ​​of the i-th and j-th influencing factors; n represents the total number of influencing factors. The value of IEVR is used to measure the average variation ratio of the interaction effect and the individual effect of a certain independent variable. IEVR>0 means that the interaction effect of the independent variable is stronger than the individual effect. The larger the value, the more obvious it is, and vice versa.

[0114] Data source and processing

[0115] Data Source

[0116] (1) GPS track data:

[0117] GPS big data has been proven to be useful for inventing tourist flow network structures. Compared with data sources such as place names, scenic spot information, or tourist reservation data in online travel notes, GPS big data can truly and accurately reflect the spatial movement of tourists and record the real tourist flow network. Compared with social network analysis, gravity model and other methods commonly used in tourist flow invention, GIS spatial analysis can spatially visualize the aggregation status of GPS tourist trajectories and major tourist corridors, and facilitate geographic spatial statistical analysis.

[0118] The GPS trajectory data of the present invention comes from the two-step road outdoor network. The website was launched in 2008 and provides users with trajectory recording, uploading and downloading functions. It is one of the platforms commonly used by domestic outdoor travelers. First, this paper uses the Selenium package of Python language to crawl the trajectory data of the website. Considering that the tourism behavior after 2020 is affected by force majeure, the invention data time is limited to 2018-2019; in order to avoid interference of non-land mobile trajectories such as flying on the spatial pattern, the trajectory types are selected as "walking", "driving", "cycling", "water", "land sliding", "water sliding" and "other"; the selected regions are "Gansu" and "Qinghai". After the crawling is completed, the trajectory file is rasterized using the Transbigdata package of Python language. TransBigData is a Python package developed by Yu et al. for traffic spatiotemporal data processing, analysis and visualization. It provides a fast and concise method for processing common traffic spatiotemporal big data such as taxi GPS data, shared bicycle data and bus GPS data, and maps GPS data to the generated geographic grid.

[0119] According to the working principle of Transbigdata package, the boundary of Gansu and Qinghai provinces is used to filter the tracks using DataV Alibaba Cloud data visualization platform (http: / / datav.aliyun.com / portal / school / atlas / area_selector), remove the data outside the boundary, generate raster geographic graphics, convert them into GeoDataFrame, and save them as json files. The final GPS track raster data resolution is 2km, and the grid value represents the number of path points within the area. The number of crawled tracks and path points is shown in Table 1:

[0120]

[0121] Table 1. Number of crawled trajectories and path points

[0122] (2) Other data:

[0123] In addition to GPS big data, this paper uses spatial geographic data and statistical data to supplement the description of regional characteristics and explore the relationship between tourist flows and other factors. The classification, name, time and source of the spatial geographic data used are shown in Table 2:

[0124]

[0125]

[0126] Table 2 Spatial geographic data used in the invention

[0127] The statistical data used in this paper mainly come from the official website, among which the data on GDP, total population at the end of the year, and the proportion of the tertiary industry in 2019 are from the Gansu Provincial Bureau of Statistics (http: / / tjj.gansu.gov.cn / ) and the Qinghai Provincial Bureau of Statistics (http: / / tjj.qinghai.gov.cn / ); the number of traditional villages and national historical and cultural towns and villages is from the Ministry of Housing and Urban-Rural Development of the People's Republic of China (https: / / www.mohurd.gov.cn / ); the number of museums is from the State Administration of Cultural Heritage (http: / / www.ncha.gov.cn / ). The above statistical data are all from 2019.

[0128] The formation of the tourism flow network structure in Gansu and Qinghai depends on local tourism resources, and is also closely related to the regional economic development level, infrastructure, etc. Through the investigation of regional characteristics and related inventions, and comprehensive consideration of the representativeness, scientificity, and data availability of indicators, this paper divides the influencing factors into four dimensions: natural geographical conditions, population and economic conditions, infrastructure construction, and tourism development potential. A total of 16 influencing factors are selected to invent their contribution to the spatial structure of tourism flows in Gansu and Qinghai. The specific influencing factors are shown in Table 3:

[0129]

[0130]

[0131] Table 3 Selection of influencing factors

[0132] Data processing:

[0133] Data validity verification:

[0134] The aim is to explore the spatial structure of tourism flows in Gansu and Qinghai, and to explore the relative differences in the spatial agglomeration of tourism flows in different regions, rather than focusing on the absolute number of path points in each region. To ensure that the data used can truly and effectively reflect the spatial pattern of tourism flows in Gansu and Qinghai, the same method as 2.3.1 (1) was used, and the crawling conditions were modified to the "Liangbu Road" website, with the keywords "Gansu" and "Qinghai". The first 7,000 trajectory data were comprehensively sorted, and a total of 14,000 trajectory data were obtained. The GPS trajectory grids obtained by the data used in the present invention (data A) and the data crawled under the above conditions (data B) were subjected to kernel density analysis with a resolution of 5 km, and the statistics of the kernel density of the two sets of data and the grid correlation were calculated. The calculation results show that the kernel density means and standard deviations of data A and B are similar, and they have a spatial similarity of 77.39%. Therefore, it can be considered that the spatial pattern reflected by the trajectory data is basically stable and can more realistically reflect the spatial characteristics of tourism flows in Gansu and Qinghai.

[0135] Data preprocessing:

[0136] Two scales are used for spatial analysis: grid scale and administrative district scale. The grid can reflect the spatial differentiation of tourist flows in a more detailed manner, which is conducive to the identification of hot and cold spots and node-corridors; the administrative district scale can be combined with socio-economic indicators to reflect the intensity of tourist flows in each city and its influencing factors. For these two scales, the GPS trajectory grid data is first preprocessed: A. Calculation of tourist flow intensity in uniform grid units:

[0137] The fishing net tool of ArcGIS was used to divide the Gansu-Qinghai region into 5km×5km uniform grids. The number of path points represented by the GPS trajectory grids in each grid was counted as the tourist flow intensity of the grid, and a uniform tourist flow grid with a resolution of 5km in the entire Gansu-Qinghai region was obtained.

[0138] B. Calculation of tourist flow intensity based on administrative districts:

[0139] For administrative regions with a specific shape, the following formula is used to calculate the tourism flow intensity of the administrative region (city, state, county):

[0140]

[0141] Wherein, I is the tourist flow intensity of the administrative area, j is the number of GPS track grids within the administrative area, nj is the value of each GPS track grid within the administrative area (the number of path points within the grid), and S is the unit area of ​​the GPS track grid. In the present invention, the resolution of the GPS track grid is 2km, and S is 4km2. Due to the large difference in the area of ​​each administrative area in Gansu and Qinghai, this method can eliminate the area influence of each administrative area and perform the upscaling calculation of the track data more reasonably.

[0142] The tourism flow network covers most areas of Gansu and Qinghai provinces. Areas with a large number of path points are distributed in key tourist destinations and routes. GPS tourism tracks show obvious linear agglomeration characteristics in spatial distribution. The density of tourism flow tracks in Gansu Province is higher than that in Qinghai Province. Tourism flow tracks are concentrated in the Hexi Corridor from Lanzhou to Dunhuang in Gansu Province, and in Xining City and the surrounding areas around Qinghai Lake in Qinghai Province. Among them, Dunhuang is a world-famous historical and cultural city with rich cultural and tourism resources and a strong tourism brand influence, which is very attractive to domestic and foreign tourists; and Qinghai Lake is a national 5A-level tourist attraction, the largest inland saltwater lake in China, spanning Hainan and Haibei Prefectures, and is one of the most famous tourist destinations in Qinghai Province and even the western region. From the perspective of points and lines, the areas of point-like agglomeration are concentrated in Lanzhou, Xining, Dunhuang and the vicinity of Qinghai Lake. Lanzhou and Xining are provincial capitals, while Dunhuang and Qinghai Lake are traditional tourist destinations. Some emerging tourist destinations such as Delingha, Golmud, Jiayuguan, Zhangye, etc. also have a large number of route points concentrated there. The areas of linear agglomeration are mainly located along the Lanzhou-Haidong-Xining line, the Qaidam Basin and the "Gansu-Qinghai Grand Loop".

[0143] Pi in formula (2) is replaced by the tourism flow intensity of the ith city. Regression analysis and global spatial autocorrelation analysis are performed on the 22 prefecture-level cities (states) in Gansu and Qinghai, and the distribution Pareto index and Moran index of each city (state) are calculated.

[0144] The results show that the Pareto index of the distribution of tourism flow intensity in Gansu and Qinghai is 1.1701, indicating that the distribution of urban tourism scale in Gansu and Qinghai is concentrated, the cities with high ranking are prominently developed, and a large number of small and medium-sized tourism cities are concentrated in the low ranking; the Moran index of all cities is greater than 0 and the results are significant, indicating that the tourism flow in Gansu and Qinghai shows positive spatial correlation, and the Moran index of the provincial central cities Lanzhou and Xining with spatial agglomeration characteristics has reached 0.61 and 0.60 respectively, indicating that the tourism flow in Lanzhou and Xining shows a more significant agglomeration, and is significantly higher than other cities in the region; about half of the cities have Moran indexes higher than the average, including Jiuquan City, Haibei Tibetan Autonomous Prefecture, Gannan Tibetan Autonomous Prefecture in the tourism hotspot area, and Dingxi City, Tianshui City, Baiyin City in the tourism relatively cold spot area. Combined with the results of kernel density analysis, it can be seen that the tourism flow of these cities is concentrated in some popular destinations in the region, with obvious spatial differentiation, while the spatial positive correlation of cities with Moran indexes below the average is not obvious enough. In summary, the spatial distribution of tourism flows in Gansu and Qinghai is uneven, with large differences among regions and obvious agglomeration characteristics, which is in line with the general law of regional tourism development.

[0145] From the perspective of the standard deviation ellipse, the distribution center is located in the Qinghai Lake area of ​​Hainan Tibetan Autonomous Prefecture and Haibei Tibetan Autonomous Prefecture, and the distribution direction of the major axis is roughly the same as that of Qilian Mountain. The overall tourist hotspots are concentrated in the northwest of Gansu Province and the northeast of Qinghai Province on both sides of Qilian Mountain. From the perspective of kernel density, the high-density areas of tourist flow show linear distribution and fragmentation characteristics. Among them, the high-density areas are located in Lanzhou-Xining and its along the line, the area around Qinghai Lake, the northeast of Haixi Prefecture, the northwest of Jiuquan City, and the southern part of Gannan Prefecture. There are also a small number of high-density path points in the urban centers of Jiayuguan City, Zhangye City, Jinchang City, Wuwei City, and Tianshui City, as well as the Qilian Mountain area in the north of Haibei Prefecture. Overall, the spatial differences in tourist flow density indicate that tourists' travel behavior in Gansu and Qinghai is concentrated. The "Gansu-Qinghai Grand Loop" tourist route is the focus of tourists' attention, while the central areas of each city and the linear high-density areas meet the accommodation and living needs of tourists during travel and the behavioral characteristics of moving along transportation routes.

[0146] The hot spots with a confidence level of 99% coincide with the tourist nodes of the "Gansu-Qinghai Grand Loop" and the more famous tourist destinations, including Lanzhou, Xining and the area around Qinghai Lake, the Qaidam Natural and Cultural Corridor (Delingha, Dachaidan, Golmud, etc.), Dunhuang, Guazhou, Jiayuguan, Jiuquan, Zhangye and the Lazikou-Zhagana Scenic Area on the Gannan Plateau. It can be seen that the tourist hot spots in the Gansu-Qinghai region are mainly the node cities of the "Gansu-Qinghai Grand Loop", among which the tourist hot spots in the two provincial central cities of Lanzhou and Xining, the area around Qinghai Lake and Dunhuang are the most concentrated and contiguous; the hot spots such as Jiayuguan and Zhangye in the Hexi Corridor are relatively insignificant; the tourist hot spots outside the "Gansu-Qinghai Grand Loop" present a small-scale and fragmented pattern.

[0147] Kernel density analysis and hot and cold spot analysis indicate that the tourism flow in Gansu and Qinghai has obvious local hot spots and network connectivity structure, and has the basic form of point-axis development model. Referring to the point-axis theory, the spatial pattern of tourism flow in Gansu and Qinghai can be divided into two parts: nodes and corridors:

[0148] (1) Node: A point in the tourist flow that has abundant tourism resources and greater tourism appeal and serves as a source of tourism. Different spatial tourism behaviors converge or spread here. Tourist nodes are usually key areas where tourists stay and visit. High-level tourist nodes represent popular tourist destinations.

[0149] (2) Corridor: The main spatial channel between different nodes. A high-level tourism corridor represents a more complete transportation channel, and usually has more complete infrastructure and service facilities along the route.

[0150] Based on the definition of tourism flow network nodes and corridors, on the one hand, according to the results of kernel density analysis, the areas with kernel density values ​​greater than 1 / 2 standard deviation are extracted as the main tourism corridors, and the extracted results are corrected according to the road network to make them connected; on the other hand, according to the results of hot spot analysis, the main tourism nodes are selected in combination with the popular tourist cities (prefecture-level and county-level) and scenic spots in the region. The total number of path points in the 30km buffer zone of the node and the 10km buffer zone of the corridor are respectively used as the tourism intensity of the node or the connectivity index of the corridor, and the tourism nodes and routes are divided using the quantile method to obtain the main tourism network in Gansu and Qinghai. From level one to level five, the node tourism intensity or corridor connectivity gradually decreases.

[0151]

[0152] Table 4 Classification of major tourist nodes

[0153]

[0154] Table 5 Classification of major tourist corridors

[0155] According to the results of the classification of tourism nodes (Table 4), the most important tourism nodes in Gansu and Qinghai include the provincial capital cities of Lanzhou and Xining, Qinghai Lake and its nearby Haiyan and Gonghe counties, and Dunhuang and Guazhou in the Dunhuang cultural circle; followed by Delingha, Chaka Salt Lake, Mangya in Haixi Prefecture, Aksai, Jiayuguan, Zhangye in the Hexi Corridor, and Zhagana Scenic Area in the south of Gannan Prefecture. According to the results of the classification of tourism corridors (Table 5), the more mature and well-connected tourism corridors include the inter-provincial tourism corridor connecting Wuwei-Lanzhou-Xining-Qinghai Lake Scenic Area, the Qaidam Basin tourism corridor connecting Gonghe-Chaka Salt Lake-Delingha-Dachaidan, the Dunhuang Cultural Circle Tourism Corridor connecting Aksai-Dunhuang-Guazhou, and the Qinghai-Tibet Highway Tourism Corridor from Golmud.

[0156] The classification results show that the tourism flow in Gansu and Qinghai has basically formed two polar tourism hotspots of Lanzhou-Xining-Qinghai Lake and Dunhuang-Guazhou, and a relatively mature tourism corridor has been formed in the Qaidam Basin and the Hexi Corridor; Longnan, Pingliang, Qingyang in the southeast of Gansu and Yushu, Guoluo, Huangnan and other cities and prefectures in the south of Qinghai are relatively cold tourism areas, and the levels of tourism nodes and corridors are relatively low. The "Gansu-Qinghai Grand Loop" (Lanzhou-Xining-Qinghai Lake-Chaka Salt Lake-Delingha-Dachaidan-Aksai-Dunhuang-Guazhou-Yumen-Jiayuguan-Linze-Zhangye-Qilian-Menyuan-Xining) is generally a popular tourism area, and its connected tourism nodes and tourism corridors have a high level. However, there is spatial differentiation within the loop. The tourism corridor intensity of the Guazhou-Yumen-Jiayuguan-Linze-Zhangye line located in the Hexi Corridor is relatively low, and its tourism connectivity is worse than that of the Qinghai Lake-Qaidam line in the south.

[0157] At the same time, there is no absolute coupling relationship between the levels of tourist nodes and tourist corridors. For example, the Lanzhou-Xining-Qinghai Lake corridor has a high level of tourism intensity and connects higher-level tourist nodes, but although tourist nodes such as Zhagana and Tianshui have relatively high levels, they lack high-level tourist corridors connected to them. Given that most of the tourist corridors in the invention area are formed by tourists' self-driving behavior, the coupling phenomenon of high-level tourist nodes and low-level tourist corridors shows that tourists' choice of tourist corridors depends not only on the attractions or cities they are connected to, but also on the overall construction of the landscape, facilities, etc. along the corridor. When the tourist corridor does not meet the tourists' viewing intentions or self-driving conditions, tourists are more inclined to reach their destinations by public transportation, and this process will not be reflected in the trajectory data.

[0158] The optimal parameter discretization method in was used to divide each influencing factor into 5 to 8 levels for spatial detection, and the explanatory power q of each influencing factor on the intensity of tourist flow in Gansu and Qinghai was calculated (Table 6). When the detection result is significant, the larger the q value, the stronger the explanatory power of the factor on the intensity of tourist flow.

[0159]

[0160] Table 6 Single factor test results

[0161] All 16 influencing factors passed the significance test. Among them, the factors with larger q values ​​are tourism service facility density (X13), A-level scenic spots and nature reserves density (X14), and ecological space area (X6), with q values ​​of 0.830, 0.761, and 0.745, respectively. The agglomeration of tourism service facilities in Gansu and Qinghai is similar to the road network. They are concentrated in Lanzhou, Baiyin, Dingxi, Linxia, ​​Tianshui, Longnan, Pingliang, Qingyang in southeastern Gansu, and Xining and Haidong in northwest Qinghai, while in other areas they are sparsely distributed along the main roads. The construction of tourism service facilities reflects the impact of supporting facilities and transportation convenience of tourist destinations on tourist behavior, and its spatial differentiation reflects the linearization of tourism space in Gansu and Qinghai.

[0162] The density of A-level scenic spots and nature reserves and the area of ​​ecological space reflect the number of natural landscapes that attract tourists. The spatial distribution density of route points shows that tourist flows are basically concentrated around tourist destinations such as scenic spots and nature reserves. From the overall pattern, the scenic spots in the northwest of Gansu and Qinghai are relatively sparse and few in number, but have strong tourist appeal, while the number of A-level scenic spots and nature reserves in the Lanzhou-southeast Longdong area is large, but the attraction is not ideal. This spatial non-coupling phenomenon shows that the agglomeration of tourist flows is not only affected by the spatial layout of the number of scenic spots or nature reserves, but also by non-spatial factors such as the viewing of tourist attractions and the development and management of destinations.

[0163] The factors with greater influence include the proportion of the tertiary industry (X9), road network density (X11) and gross domestic product (X8), with q values ​​of 0.400, 0.389 and 0.372 respectively. The road network density reflects the convenience of transportation facilities in the tourist destination, and the proportion of the tertiary industry and gross domestic product reveal the economic development of the county. Lanzhou, Baiyin, Dingxi and Tianshui in the southeast of Gansu Province connected to the Guanzhong Plain are high-density road network areas, while Haixi Prefecture and Yushu Prefecture in the west of Qinghai Province and Jiuquan City in the northwest of Gansu Province are sparsely distributed. Overall, there is a deviation between the distribution center of the road network and the distribution of path points. However, the tourist flow structure in key areas shows a high degree of consistency with the road network, including the ring line from Delingha to Dachaidan, Aksai, Dunhuang, Guazhou, Yumen and Jiayuguan, the area around Qinghai Lake and the Qinghai-Tibet Highway from Golmud.

[0164] The correlation between the road network and the spatial distribution of tourist flows in Gansu and Qinghai was further calculated, and the main roads (including primary roads, secondary roads, highways and expressways in OSM classification) were extracted for buffer zone analysis. The results showed that the road network and tourist flows had a 62.81% similarity in spatial distribution, and 90% of the path points were distributed within the 20-kilometer buffer zone of the main roads, which was consistent with the spatial visualization results, that is, regardless of the agglomeration of path points, the spatial layout of tourist flows showed a high degree of consistency with the road network. When examining the path points located within the 20-kilometer buffer zone of the main roads, about 58.5% of the path points were within 2 km of the main roads, and 90% of the path points were distributed in the buffer zone 15 km away from the main roads, which also shows that tourist flows have a great dependence on the road network in space.

[0165] Factors with q values ​​between 0.15 and 0.3 include annual precipitation (X2), vegetation coverage (X4), total population at the end of the year (X10), and the number of museums (X6). Annual precipitation affects natural conditions such as regional humidity, vegetation coverage reflects regional ecological conditions, total population at the end of the year represents the potential tourism market size in the region, and the number of museums indicates the amount of cultural and museum resources in the region. The q values ​​of the above factors are relatively low, and their impact on tourists' travel intention is relatively weak.

[0166] First, different natural conditions create diverse tourist landscapes. The Longnan mountainous area with high annual precipitation and high vegetation coverage and the sandy land with low annual precipitation and low vegetation coverage are equally attractive to tourists. Secondly, tourists in Gansu and Qinghai usually travel across regions by car, and the population size in the region has little impact on the distribution of trajectory points. Finally, museums are not the main destination for self-driving tourists in Gansu and Qinghai, and have little impact on tourists' itineraries.

[0167] The q values ​​of altitude (X3), river system density (X5), slope (X7), built-up area (X12), and the number of traditional villages and national historical and cultural towns and villages (X15) are all lower than 0.15, indicating that they have little or no effect on the spatial differentiation of tourism flows in Gansu and Qinghai. This shows that tourists in Gansu and Qinghai have a variety of tourism types, and their destinations are not limited to a certain altitude, river basin or area with a high level of urban construction and a large number of traditional villages.

[0168] The average q value of the factor under the same dimension represents the influence of the dimension, and the ranking of the influence of each dimension is: tourism development potential (0.455) > population and economic situation (0.313) > infrastructure construction (0.260) > natural geographical conditions (0.206). This shows that tourism development potential is the most important factor affecting the spatial differentiation of tourism flow in Gansu and Qinghai. Perfect service facilities and high-quality tourist attractions are the key driving forces to attract self-driving tourists to visit, which can more effectively enhance tourists' willingness to travel compared with other influencing factors.

[0169] The interaction detector of the geographic detector was used to explore the interaction effects of the main influencing factors, and the average variance ratio (IEVR) of the interaction effect and individual effect of each variable was calculated. The interaction effects of the main influencing factors were mainly double-factor enhancement. The top 5 interaction pairs included tourism service facility density ∩ scenic spot density (0.9229), tourism service facility density ∩ road network density (0.9129), tourism service facility density ∩ ecological space area (0.8969), tourism service facility density ∩ gross domestic product (0.8746), and scenic spot density ∩ road network density (0.8674). The top 5 interaction pairs and individual effects revealed that tourism service facility density is a key factor in explaining the spatial differentiation of tourism flows in Gansu and Qinghai, and the higher interaction detection values ​​of tourism service facility density and scenic spot density indicate that areas with more scenic spots and more complete tourism facilities usually become tourist hotspots.

[0170] From the calculation results of IEVR, it can be seen that except for the IEVR value of the density of A-level scenic spots and nature reserves (X14) which is slightly less than 0, the other IEVRs are all positive, and the overall interaction effect is 21.21% higher than the individual effect on average. The IEVR value of road network density (X11) is 60.58%, which greatly improves the explanatory power when interacting with the ecological space area (X6), the density of tourism service facilities (X13), and the density of A-level scenic spots and nature reserves (X14), indicating that the road network density has a large indirect impact on the intensity of tourism flow, mainly through the ecological space area, service facilities and tourist attractions. In addition, the IEVR values ​​of GDP (X8) and the proportion of the tertiary industry (X9) also reached 32.75% and 27.13% respectively. The interaction effect between GDP and the ecological space area (X6) is strong, while the interaction effect between the proportion of the tertiary industry (X9) and the ecological space area (X6) and GDP (X8) is large. This shows that the level of regional economic development mainly indirectly promotes the agglomeration of tourism flows through ecological factors.

[0171] The present invention spans the scales of scenic spots, towns, and provinces, applies GPS trajectory data to cross-provincial units, and reveals the spatial distribution characteristics of tourism flows in Gansu and Qinghai and their influencing factors. The results of the invention show that the tourism flow in Gansu and Qinghai presents a linear structure, forming the "four poles" of Lanzhou, Xining, Dunhuang, and Qinghai Lake, "multiple points" such as Delingha, Zhangye, and Jiayuguan, and the spatial layout of tourism corridors such as the "Gansu-Qinghai Grand Loop", Qaidam, and Qinghai-Tibet Highway. The regional tourism intensity is uneven, and its center of gravity is located in the northeast of Qinghai Province and the northwest of Gansu Province. The conclusion of the invention basically echoes the actual development and existing invention results. First, the spatial distribution of tourist hotspots in Gansu and Qinghai is consistent with the actual tourism development. Lanzhou, Xining, Qinghai Lake, Dunhuang, etc. are all hot spots for tourists to visit; the "Gansu-Qinghai Grand Loop" tourist corridor also conforms to the organization law of self-driving tour routes in Gansu and Qinghai. Secondly, the self-guided tour network of Gansu, Qinghai, and Ningxia is dense in the north and sparse in the south in space. The tourism flow between nodes in the northern tourism area of ​​Gansu and Qinghai is the closest, and the tourism flow connection between the southeastern Longdong area and other areas is the weakest. The present invention has concluded that the tourism center of gravity in Gansu and Qinghai is biased to the north, and the southeastern Longdong area is a relatively cold tourism area. In addition, the self-driving accessibility and the synergy degree of A-level scenic spots in Qinghai Province have spatial heterogeneity. The eastern counties represented by Xining City have better synergy, while the western and southern counties have serious imbalances. The synergy imbalance is similar to the phenomenon of separation of road network density and tourist flow hotspots at the provincial scale obtained by the present invention. Finally, the present invention points out the separation of the center of gravity of road network density and tourist flow distribution hotspots in Gansu and Qinghai, but the road network distribution and tourist flow distribution are highly similar. These two conclusions seem to contradict each other, but in fact they reveal that the road network is a sufficient but not necessary condition for tourist flow, that is, tourists rely on the road network for spatial movement, but the aggregation of the road network does not necessarily lead to higher tourist visits. The main tourist destinations for tourists in Gansu and Qinghai are the natural landscapes along the Grand Loop, and most of them travel by car. One or several roads connecting the scenic spots can meet the tourism needs, and there is no need to rely on a continuous high-density road network.

[0172] In terms of influencing factors, tourism development potential is the most critical dimension affecting the spatial structure of tourism flow in Gansu and Qinghai. The ecological space area, tourism service facility density, A-level scenic spots and nature reserves density, gross domestic product, the proportion of the tertiary industry and road network density are the main influencing factors. The influencing mechanism of influencing factors is differentiated. The first three factors mainly play a direct role and have a strong independent explanatory power for the spatial differentiation of tourism flow in Gansu and Qinghai; the latter three play an indirect role in the spatial differentiation of tourism flow. Among them, the tourism road network mainly affects the ecological space area, the density of tourism service facilities, and the density of A-level scenic spots and nature reserves, while the gross domestic product and the proportion of the tertiary industry mainly rely on the ecological space area. The indirect influence of the tourism road network once again confirms the separation of the center of gravity of the road network density and the hotspot of tourism flow distribution. The simple road network distribution cannot attract tourists to visit. It must be combined with good ecological conditions, service facilities and scenic spots to promote the agglomeration of tourism flow. At the same time, the regional economic development level has limited ability to predict the intensity of tourism flow, and it must be combined with a good ecological environment to produce a strong attraction.

[0173] The above are only preferred embodiments of the present invention, which are only used to help understand the method and core ideas of the present application. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

[0174] The present invention solves the problems in the prior art as a whole, such as the lack of inventions directly based on tourists' movement trajectories, the lack of applications to large-scale destinations, the failure to find reasonably matched related data acquisition and analysis methods, and the neglect of the combination of other human and natural factors. The present invention characterizes the spatial pattern of tourist flows through tourist GPS trajectory data, and forms a mechanism through methods such as GIS tools and geographic detector models, thus providing a scientific basis for data analysis of tourist flows.

Claims

1. A tourist flow data analysis method based on GPS big data, characterized in that: Including research methods, data sources and processing, and systematic technical paths for spatial visualization applications; The research methods include regression analysis, GIS spatial analysis, tourist flow node-corridor network construction method and geographic detector OPGD analysis based on optimal parameters; The regression analysis is: using Zipf regression to measure the balance of the scale of urban tourism flow, the formula is: P i =ON i -α (1) Taking the natural logarithm of both sides of formula 1 yields: lnP i =-αlnR i +A(2) Ri is the rank of the i-th city, Pi is the scale of tourist flow of the i-th city, and A and α are constants; The α is the Pareto index of the scale distribution of urban tourism flow, which is used to measure the degree of balance of scale distribution; with α=1 as the benchmark, α<1 indicates that the scale distribution of urban tourism flow is relatively dispersed, that is, the development of low-rank cities is more prominent, and the development of large cities is insufficient; on the contrary, α>1 indicates that the scale distribution of urban tourism flow is more concentrated, that is, the development of high-rank cities is more prominent, and the development of small and medium-sized cities is insufficient; The GIS spatial analysis is as follows: using overlay display, standard deviation ellipse, and buffer zone analysis to study the spatial distribution characteristics of different elements and tourist flows; Kernel density analysis is used to explore the distribution characteristics of the spatial structure of tourist flows; hot and cold spot analysis is used to extract the spatial hot spots of tourist flows, and spatial autocorrelation analysis, namely the Moran index, is used to study the correlation characteristics of tourist flows; The kernel density analysis is specifically as follows: Describe the spatial distribution characteristics of regional GPS tourism trajectories and preliminarily judge the concentration or dispersion of the spatial distribution of tourism flows. The calculation formula is as follows: f(j) is the density of point j, r represents the search radius, dij is the distance from point i to point j, and k is the weight of dij; The spatial autocorrelation analysis is: The global Moran's I and local Moran's I are used to measure the global and local autocorrelation of tourist flows, determine the overall spatial agglomeration characteristics of tourist flows, and the specific spatial locations of tourist flow agglomeration and dispersion. The calculation formula is: n is the number of spatial units, xi and xj are the mean values ​​of the observed values ​​of GPS trajectories of tourist flows in spatial units i and j, i≠j, and is the mean of the tourist flow intensity of all spatial units, wij is the spatial weight matrix, and σ 2 is the variance of tourist flow intensity; the value range of Moran's I is [-1,1]. If the value is greater than 0, it indicates that there is positive spatial autocorrelation; The hot and cold spot analysis is as follows: The Getis-Ord Gi* method is used to identify the local spatial agglomeration areas of GPS tourist trajectories. The hotspot area represents the high-value agglomeration area of ​​tourist flow, and the cold spot area represents the low-value agglomeration area of ​​tourist flow. The calculation formula is: The x j is the mean value of the observed value of GPS trajectory of tourist flow in spatial unit j, w i,j is the spatial weight between spatial units i and j, and n is the number of spatial units; in formula (6), represents the mean of the observed values ​​of GPS trajectories of tourist flows in all spatial units; Represents the variance value of the GPS trajectory of tourist flows in all spatial units; the final calculated The value is the z score; when the z value>1.28, that is, under the condition of confidence level greater than 90%, it is a high-value cluster area, i.e. a hot spot, and vice versa, it is a low-value cluster area, i.e. a cold spot; The method for constructing the tourist flow node-corridor network is as follows: The spatial pattern of tourist flow is divided into two parts: nodes and corridors. Nodes represent points with richer tourist resources and greater tourist attractions in the tourist flow; corridors represent the main spatial channels between different nodes. For the construction of nodes, we use hot and cold spot analysis to identify the global GPS trajectory data, select areas with a confidence level of more than 95%, and select major tourist nodes in combination with popular tourist cities in the region, namely, prefecture-level and county-level government seats, and scenic spots; The total number of path points in the buffer zone of a node is taken as the tourism flow intensity of the node; For the construction of corridors, a multivariate clustering method is used, which uses unsupervised machine learning methods to find non-spatial natural clusters of features based on feature attribute values; The selected clustering attribute is the number of trajectory points represented by each point, and the clustering method is K-means; Determine the appropriate number of clusters, classify the clustered data into 1-m levels according to the number of trajectory points represented from small to large, and remove the levels with fewer trajectory points until a clearer tourist flow network grid is obtained; According to the extracted tourist flow nodes, the network generated by the clustering results is converted into vector lines and corrected to make them flat. Slide, connect; For the corridor connecting nodes A and B, its tourist flow intensity T AB The calculation method is: Among them, L1, L2, ... Lm are the clustering levels represented by each grid intersecting the corridor space; the higher the level, the more trajectory points it represents and the higher the intensity of tourist flow; The OPGD analysis based on the optimal parameters of the geographic detector: The factor detection module of the geographic detector based on the optimal parameters is used to analyze the spatial relationship between the explained variable and the explanatory variable, and to determine the explanatory power of the explanatory variable on the explained variable, that is, the degree of influence. The calculation formula is: L is the stratification or partition of the independent variable and the dependent variable, and N h and N represent the number of units in the hth layer and the whole area respectively, and σ 2 are the variances of the Y values ​​of the hth layer and the entire region, respectively; q is a statistic whose value range is [0,1]. The larger the q value, the stronger the explanatory power of the factor; Secondly, the interaction detection module of the geographic detector is used to identify the interactions between different factors affecting tourism flows and to determine whether different factors, when acting together, will affect the explanatory power of the dependent variable; Based on the results of interaction detection, the interaction effect variation ratio index IEVR is further used for quantitative analysis, and its formula is: The xi and xj represent the i-th and j-th influencing factors; the q(xi) and q(xj) represent the q values ​​of the i-th and j-th influencing factors; the t represents the total number of influencing factors; the IEVR value is used to measure the average variation ratio of the interaction effect and the individual effect of a certain independent variable. IEVR>0 means that the interaction effect of the independent variable is stronger than the individual effect, and the larger the value, the more obvious it is, and vice versa; The data source and processing include data source and data processing; The data sources include GPS track data and other data. The GPS track data is obtained from crawling track data uploaded by users of various websites. The other data include spatial geographic data and statistical data. The spatial geographic data includes raster geographic data and vector geographic data. The statistical data comes from government websites and specifically includes natural geographic conditions, population and economic conditions, infrastructure construction and tourism development potential. The data processing includes: Data validity verification: kernel density analysis was performed on the data obtained in the research method and the GPS trajectory data and other data, and the statistics of the kernel density and grid correlation of the two sets of data were calculated; data preprocessing: two scales were used for spatial analysis: grid scale and administrative district scale; the grid scale can reflect the spatial differentiation of tourism flow more finely, which is conducive to the identification of hot and cold spots and node-corridors; the administrative district scale can be combined with socio-economic indicators to reflect the intensity of tourism flow in each city and its influencing factors; for these two scales, the GPS trajectory grid data was first preprocessed: A. Calculation of tourist flow intensity based on uniform grids: The fishing net tool of ArcGIS is used to divide the tourist area into uniform grids. The number of path points represented by the GPS track grid in each grid is counted as the tourist flow intensity of the grid, and the uniform tourist flow grid with full-domain resolution of the tourist area is obtained. B. Calculation of tourist flow intensity based on administrative districts: For administrative regions, the following formula is used to calculate the intensity of tourist flows in the administrative region, i.e., city, state, or county: I is the tourism flow intensity of the administrative area, j is the number of GPS track grids within the administrative area, and n j is the value of each GPS track grid within the administrative area, that is, the number of path points within the grid, and D is the unit area of ​​the GPS track grid.

Images