Binary geographic flow space-time interaction anomaly detection method based on space-time autocorrelation

Abstract

The invention discloses a binary geographic flow space-time interaction anomaly detection method based on space-time autocorrelation, which measures global space-time autocorrelation of population flow and information flow based on binary geographic flow Baidu migration and Baidu index OD data by taking a municipal administration unit as a research unit and based on a space-time autocorrelation method. And selecting a reasonable time threshold value according to a global space-time self-correlation result and applying the reasonable time threshold value to local space-time self-correlation, thereby identifying two abnormal interaction modes of high-value abnormal interaction and low-value abnormal interaction. According to the method, the defects of binary variable global autocorrelation research are overcome, the spatiotemporal weight is more generalized through the change rate of the attribute value of the stream along with time introduced into the autocorrelation spatiotemporal weight, a local binary G statistic expression with generality is deduced, the algorithm rationality and applicability of the binary stream are higher, and the algorithm efficiency is improved. And a powerful tool is provided for research of binary streams in cities and between cities, maximization of urban multi-element resource utilization efficiency and optimization of element resource allocation.

Description

Technical Field

[0001] The invention belongs to the field of geographic information science and technology, and specifically relates to a method for detecting spatiotemporal interaction anomalies of binary geographic flows based on spatiotemporal autocorrelation. Background Art

[0002] In recent years, with the rapid development of urbanization and the advancement of regional integration, the interaction of various elements between cities has become more frequent and close. In the process of exchanging various resource elements between cities, adjacent cities have gradually developed into urban agglomerations with important status. The spatiotemporal interaction of diverse resources between cities has formed a complex spatiotemporal interaction pattern. The flow of people, logistics and information has also increasingly become the three basic element flows to measure the degree of connection between cities and the development of urban agglomerations. However, the high spatiotemporal agglomeration level of various elements in some regions has caused a waste of resources, while the low agglomeration level has caused inefficient resource utilization and made it difficult to exert the synergistic effect of the overall development of urban agglomerations. Therefore, it is of great significance to associate the geographical flow elements with each other and detect the abnormalities of the spatiotemporal interaction of the two geographical flows based on the spatiotemporal autocorrelation method to achieve fair and reasonable resource allocation and sustainable urban development.

[0003] Autocorrelation analysis is a means of detecting abnormal data distribution. Previous studies have focused on the spatial autocorrelation of geographic flows, using one or more autocorrelation statistics to study them. Autocorrelation statistics include Moran's I, Geary's C, and G statistics (Getis-Ord General and Getis-Ord Gi). However, temporal autocorrelation is also common in geographic flows. The study of spatiotemporal autocorrelation of geographic flows is an extension of the study of spatial autocorrelation to the time dimension. Previous studies on spatiotemporal autocorrelation of geographic flows have mostly focused on univariate flows, and the Moran's index is often used to modify spatiotemporal weights, thereby revealing the spatiotemporal dynamic geographical distribution of flows. However, the Moran's index can only determine whether there is spatial agglomeration of geographic flows with similar attributes, and cannot distinguish whether such agglomeration is high-value agglomeration or low-value agglomeration. The spatiotemporal weight values ​​in the autocorrelation statistics are mostly in binary form. Using only the time interval as a variable to construct the time weight is too simplistic, which may overestimate the impact of spatiotemporal neighboring flows on the central flow and ignore the change of flow values ​​over time. The spatiotemporal autocorrelation of multi-type geographic flows (such as binary flows) has not been fully explored. The spatial autocorrelation of multi-type flows has mostly been studied by local autocorrelation, and there is a lack of research on global autocorrelation.

[0004] Therefore, it is urgent to construct a method that can perform spatiotemporal autocorrelation analysis on binary geographic flow data at both global and local scales and detect abnormal interactions of spatiotemporal flow data between regions, so as to further rationally allocate and optimize resource elements and achieve maximum resource utilization efficiency and sustainable development. Summary of the invention

[0005] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides a method for detecting spatiotemporal interaction anomalies of binary geographic flows based on spatiotemporal autocorrelation. The method takes the city-level area within a single urban agglomeration as the research unit, constructs a binary global G statistic using the univariate global G statistic, and selects a reasonable time threshold through the global spatiotemporal autocorrelation result; then, the improved BiFlowAMOEBA (binary flow multi-directional optimal ecological area) algorithm is used to perform spatiotemporal clustering of four modes, HH, HL, LH and LL, under the time threshold, and the HH and LL clusters are used as the interaction anomaly areas of the binary flow.

[0006] According to one aspect of the present invention, a method for detecting spatiotemporal interaction anomalies of binary geographic flows based on spatiotemporal autocorrelation is provided, comprising:

[0007] Step 1: Obtain the study area dataset, including the group flow OD spatiotemporal data of population flow and information flow within the preset time period represented by Baidu Migration Index and Baidu Index respectively at the city level of the study area;

[0008] Step 2: Construct a binary global G statistic by multiplying two univariate global G statistics;

[0009] Step 3: Use the regional spatial adjacency relationship and the rate of change of the flow value over time to construct the spatiotemporal weight, replace the weight in the statistic in step 2, and obtain the global spatiotemporal autocorrelation results of the binary flow under different time thresholds;

[0010] Step 4: Use the L statistic and the univariate local G statistic to derive the formula of the weighted generalized bivariate local G statistic;

[0011] Step 5: Combined with the global spatiotemporal autocorrelation results of the binary flow in step 3, the set time threshold is selected and applied to the improved BiFlowAMOEBA algorithm. The algorithm selects seed flows under four clustering modes, and uses the hierarchical clustering strategy in each mode. Under the assumption that there is no in-situ correlation, the region is merged with the goal of increasing the binary local G statistic;

[0012] Step 6: Obtain the spatiotemporal clustering results of the four modes of HH, HL, LH and LL respectively, and take the spatiotemporal clustering results of HH and LL as the interaction anomaly results of the binary flow.

[0013] As a further technical solution, obtaining the study area dataset also includes:

[0014] Perform unified projection transformation on spatial vector data to form a vector data set with consistent spatial coordinate system;

[0015] The Baidu Migration and Baidu Index OD data are cleaned, and data from continuous time periods with more valid data are selected. The two types of geographic flow data are merged into the same city and time point, and the city ID and time ID fields are established, where the minimum time unit of the OD data is day.

[0016] As a further technical solution, a binary global G statistic is constructed by multiplying two univariate global G statistics:

[0017]

[0018] Among them, w ij is the spatiotemporal weight constructed using spatial adjacency and temporal correlation; x i With y i are two attribute values ​​of geographic flow standardized by z-score; x j and j are the two attribute values ​​of the z-score-normalized geographic flow within the spatiotemporal neighborhood of the i-th region.

[0019] As a further technical solution, the spatiotemporal weight is constructed by using the regional spatial adjacency relationship and the rate of change of the flow value over time:

[0020] w ij =sw ij ×tw ij

[0021] Among them, w ij is the spatiotemporal weight; sw ij is the spatial weight defined by the spatial adjacency; tw ij is the time weight defined by the time distance between flows; the spatial weight and time weight are defined as follows:

[0022]

[0023] in, and are the time weights corresponding to the binary attribute values ​​of the flow; x i(t) With y i(t) They represent two attribute values ​​of the flow at time t; x j(t-Δt) and j(t-Δt) are the two attribute values ​​of adjacent flows at time t-Δt; Δt is the time interval; t θ is the time threshold; tw ij is the time weight, which is the average value of the time weight corresponding to the binary attribute value of the flow.

[0024] As a further technical solution, the formula of the binary local G statistic in the generalized form of weight is derived using the L statistic and the univariate local G statistic:

[0025]

[0026] in, and are the binary attribute values ​​after Z-score standardization; w ij is the spatiotemporal weight of the i-th flow and the j-th flow in the spatiotemporal neighborhood; N * is the number of all flows; n * is the number of adjacent flows in the spatiotemporal neighborhood of the i-th flow.

[0027] As a further technical solution, step 5 further includes:

[0028] The time threshold selected in step 3 and the binary local G statistic in step 4 are combined Applied to the binary flow multi-directional optimal ecological zone algorithm, the hierarchical clustering strategy is used to screen seed flows in four clustering modes: HH, HL, LH and L. If both attribute values ​​of the binary flow after z-score standardization are positive, it is the HH mode; if both attribute values ​​are negative, it is the LL mode; if one attribute value is positive and the other is negative, it is the HL or LH mode.

[0029] The initialization phase sets the seed stream of each mode The value is 0, and the region is merged with the goal of increasing the binary local G statistic. The merging rules are as follows:

[0030]

[0031] in, It is in the merge area BFC i ∪BFC j The number of binary flows in is calculated by calculating the For two clusters BFC i and BFC j ,if is the maximum value, and Gain(BFC i ∪BFC j )>0, it is considered that the two clusters can be merged into a new cluster.

[0032] As a further technical solution, the method also includes: outputting abnormal interaction recognition results and displaying their spatiotemporal distribution.

[0033] According to one aspect of the present invention, a device for detecting anomalies of binary geographic streams in time and space interaction based on time and space autocorrelation is provided, comprising:

[0034] The first main module is used to obtain the dataset of the study area, including the group flow OD spatiotemporal data of population flow and information flow within the preset time period represented by Baidu Migration Index and Baidu Index respectively at the city level of the study area;

[0035] The second main module is used to construct a binary global G statistic by multiplying two univariate global G statistics;

[0036] The third main module is used to construct spatiotemporal weights using the regional spatial adjacency relationship and the rate of change of the flow value over time, replace the weights in the statistics of the second main module, and obtain the global spatiotemporal autocorrelation results of binary flows under different time thresholds;

[0037] The fourth main module is used to derive a weight generalized form of a binary local G statistic formula using the L statistic and the univariate local G statistic;

[0038] The fifth main module is used to combine the binary flow global spatiotemporal autocorrelation results in the third main module, select the set time threshold and apply it to the improved BiFlowAMOEBA algorithm. The algorithm selects seed flows under four clustering modes, uses the hierarchical clustering strategy in each mode, and performs regional merging with the goal of increasing the binary local G statistic under the assumption that there is no in-situ correlation;

[0039] The sixth main module is used to obtain the spatiotemporal clustering results of the four modes of HH, HL, LH and LL respectively, and take the spatiotemporal clustering results of HH and LL as the interaction anomaly results of the binary flow.

[0040] According to one aspect of the present invention specification, there is provided a system for detecting spatiotemporal interaction anomalies of binary geographic streams based on spatiotemporal autocorrelation, comprising a memory and a processor; the memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the steps of the method for detecting spatiotemporal interaction anomalies of binary geographic streams based on spatiotemporal autocorrelation.

[0041] According to one aspect of the present specification, a non-transitory computer-readable storage medium is provided, wherein the non-transitory computer-readable storage medium stores computer instructions, wherein the computer instructions enable the computer to execute the steps of a method for detecting spatiotemporal interaction anomalies of binary geographic flows based on spatiotemporal autocorrelation.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] 1. In the aspect of spatiotemporal interaction anomaly detection of binary geographic flow, the present invention is based on binary geographic flow OD data, takes municipal administrative units as research units, and measures the global spatiotemporal autocorrelation of population flow and information flow based on the spatiotemporal autocorrelation method. According to the global spatiotemporal autocorrelation results, a reasonable time threshold is selected and applied to the local spatiotemporal autocorrelation, thereby identifying the two abnormal interaction modes of HH and LL.

[0044] 2. The present invention incorporates the time dimension into the study of autocorrelation of binary flows, and expands the research method of spatiotemporal autocorrelation of univariate flows; a global binary G statistic is constructed through the development of a univariate global G statistic, which makes up for the shortcomings of global autocorrelation research on binary variables; the concept of the rate of change of the attribute value of the flow over time is introduced into the autocorrelation spatiotemporal weight, so that the spatiotemporal weight is more generalized, thereby deriving a general local binary G statistic expression, making the algorithm of binary flows more reasonable and applicable, and providing a powerful tool for the study of binary flows within and between cities and for maximizing the efficiency of urban multi-factor resource utilization and optimizing factor resource allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, a brief introduction is given below to the drawings used in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0046] Figure 1 A flowchart of a method for detecting spatiotemporal interaction anomalies in binary geographic streams based on spatiotemporal autocorrelation is provided in an embodiment of the present invention.

[0047] Figure 2 A flow chart of the BiFlowAMOEBA algorithm based on global autocorrelation provided in an embodiment of the present invention.

[0048] Figure 3 and Figure 4 These are all spatiotemporal distribution diagrams of abnormal interaction traffic with high population flow and information flow provided by the embodiments of the present invention.

[0049] Figure 5 A schematic diagram of the structure of a binary geographic stream spatiotemporal interaction anomaly detection device based on spatiotemporal autocorrelation provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. In addition, the technical features in the various embodiments or single embodiments provided by the present invention are arbitrarily combined with each other to form a new technical solution. This combination is not restricted by the sequence of steps and / or the structural composition mode, but must be based on the ability of ordinary technicians in this field to achieve. When the combination of technical solutions is contradictory or cannot be achieved, it should be considered that this combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0051] The present invention provides a method for detecting spatiotemporal interaction anomalies of binary geographic flows based on spatiotemporal autocorrelation. The method takes the city-level area within a single urban agglomeration as the research unit, constructs a binary global G statistic using a univariate global G statistic, and selects a reasonable time threshold through the global spatiotemporal autocorrelation result; then, the improved BiFlowAMOEBA (binary flow multi-directional optimal ecological area) algorithm is used to perform spatiotemporal clustering of four modes, HH, HL, LH and LL, under the time threshold, and clusters HH and LL as interaction anomaly areas of binary flows. Finally, on this basis, optimization measures for reasonable resource allocation can be proposed. Specifically, the following steps are included:

[0052] Step 1: Collect and process the study area data set, including the group flow OD spatiotemporal data of population flow and information flow within one week at the city level in the study area, represented by Baidu Migration Index and Baidu Index respectively. The minimum time unit of the OD data is day.

[0053] Preferably, step 1 mainly processes the data including: (1) performing unified projection transformation on the spatial vector data to form a vector data set with a consistent spatial coordinate system; and (2) performing statistics on the daily binary flow data (population flow and information flow) between the city-level regions within the study area within 7 days.

[0054] Step 2: Construct a binary global G statistic by multiplying two univariate global G statistics.

[0055] Preferably, the construction of the binary global G statistic in step 2 can be obtained by the following formula:

[0056]

[0057] Among them, w ij is the spatiotemporal weight constructed using spatial adjacency and temporal correlation; x iWith y i are two attribute values ​​of geographic flow standardized by z-score; x j and j are the two attribute values ​​of the z-score-normalized geographic flow within the spatiotemporal neighborhood of the i-th region.

[0058] Step 3: Use the regional spatial adjacency relationship and the rate of change of the flow value over time to construct the spatiotemporal weights, replace the weights in the original statistics, and obtain the global spatiotemporal autocorrelation results of the binary flow under different time thresholds.

[0059] Preferably, the spatiotemporal weight in step 3 is constructed by multiplying the spatial weight by the temporal weight, and the formula is as follows:

[0060] w ij =sw ij ×tw ij

[0061] Among them, w ij is the spatiotemporal weight; sw ij is the spatial weight defined by the spatial adjacency; tw ij is the temporal weight defined by the temporal distance between flows.

[0062] (1) For the origin and destination (OD) areas of the flow, the spatial weight is based on the Queen adjacency rule and is defined as follows:

[0063]

[0064] (2) For the construction of time weight, it is assumed that the flow at time t is affected by both the current time t and the adjacent flow at time t-Δt within a certain time threshold. Because each flow has a timestamp and a binary attribute value, the attribute values ​​of flows at different time points are different. Therefore, the time distance between flows can be calculated using the time difference and the rate of change of the flow over time. The formula is as follows:

[0065]

[0066]

[0067] in, and are the time weights corresponding to the binary attribute values ​​of the flow; x i(t) With y i(t) They represent two attribute values ​​of the flow at time t; x j(t-Δt) and j(t-Δt) are the two attribute values ​​of adjacent flows at time t-Δt; Δt is the time interval; t θ is the time threshold; tw ijis the time weight, which is the average value of the time weight corresponding to the binary attribute value of the flow.

[0068] Step 4: Use the L statistic and the univariate local G statistic to derive the formula for the generalized weighted bivariate local G statistic.

[0069] Preferably, the binary local G statistic formula in step 4 is derived as follows:

[0070] (1) The bivariate local L statistic and the univariate local G statistic are used to construct the bivariate local G statistic. The bivariate local L statistic is used to measure the bivariate spatial dependence, and its formula is as follows:

[0071]

[0072] When the weight matrix is ​​row-normalized, its formula can be simplified as follows:

[0073]

[0074] in, and Represents the local weighted z-scores of variables x and y respectively; S X and S Y is the standard deviation of variables x and y.

[0075] (2) After the weight matrix is ​​row-normalized, the formula for the univariate local G statistic is as follows:

[0076]

[0077] (3) Inspired by G*, the binary local G statistic should have both the spatial correlation of the two variables and the number of local units to represent the local spatial correlation, and then expand to the time dimension to show the spatiotemporal correlation. When the spatiotemporal weight matrix is ​​row-normalized, its formula is as follows:

[0078]

[0079] in, and are the binary attribute values ​​after Z-score standardization; w ij is the spatiotemporal weight of the i-th flow and the j-th flow in the spatiotemporal neighborhood; N * is the number of all flows; n * is the number of adjacent flows in the spatiotemporal neighborhood of the i-th flow.

[0080] Step 5: Combined with the global spatiotemporal autocorrelation results in step 3, select a reasonable time threshold and apply it to the improved BiFlowAMOEBA algorithm, such as Figure 2 As shown in the figure, the algorithm selects seed streams under four clustering modes, applies the hierarchical clustering strategy in each mode, and merges regions with the goal of increasing the binary local G statistic under the assumption that there is no in-situ correlation.

[0081] As a preference, in step 5, a hierarchical clustering strategy is applied in each clustering mode, and in the initialization phase, the seed streams in each mode are The value is 0, and the region is merged with the goal of increasing the binary local G statistic. The merging rules are as follows:

[0082]

[0083] in, It is in the merge area BFC i ∪BFC j The number of binary flows in is calculated by calculating the For two clusters BFC i and BFC j ,if is the maximum value, and Gain(BFC i ∪BFC j )>0, it is considered that the two clusters can be merged into a new cluster.

[0084] Step 6: Get the spatiotemporal clustering results of the four modes HH, HL, LH and LL respectively, and use the spatiotemporal clustering results of HH and LL as the interaction anomaly results of the binary flow. Output the abnormal interaction recognition results and display their spatiotemporal distribution.

[0085] As a preferred embodiment, see Figure 1 The present invention provides a method for detecting anomalies of binary geographic flow space-time interaction based on space-time autocorrelation, comprising the following steps:

[0086] Step 1: Using the Guangdong-Hong Kong-Macao Greater Bay Area as the research area, Baidu Migration and Baidu Index (PC plus mobile) OD data are used as representations of population flow and information flow. The data set contains city names (due to data availability, the two special administrative regions of Hong Kong and Macao are excluded), flow values ​​represented by index values, and time. Data processing mainly includes: (1) uniform projection transformation of spatial vector data to form a vector data set with consistent spatial coordinate system; (2) cleaning Baidu Migration and Baidu Index OD data, selecting data from a continuous week (October 1 to October 7, 2021) with more valid data (not 0), merging the two geographic flow data into the same city and time point, and establishing city ID and time ID fields.

[0087] Step 2: Using the spatial adjacency and temporal distance between binary flows, the spatiotemporal weights of each flow and its neighboring flows in the spatiotemporal neighborhood are calculated at different time thresholds without considering the in-situ correlation. The temporal weights are applied to the constructed global autocorrelation expression, which is defined as follows:

[0088]

[0089] Among them, d represents a certain time and space distance range, w ij is the spatiotemporal weight constructed using spatial adjacency and temporal correlation; x i With y i are two attribute values ​​of geographic flow standardized by z-score; x j and j are the two attribute values ​​of the z-score-normalized geographic flow within the spatiotemporal neighborhood of the i-th region.

[0090] In this embodiment, the global spatiotemporal autocorrelation results are shown in Table 1.

[0091] Table 1 Global spatiotemporal autocorrelation results under different time thresholds

[0092]

[0093] In Table 1, *, **, and *** indicate that BG(d) is significant at the 0.05, 0.01, and 0.005 levels, respectively.

[0094] Step 3: Based on the spatiotemporal autocorrelation results and time thresholds obtained in step 2, if the purpose of the analysis is to identify short-term anomalies or fluctuations, a smaller time threshold (such as days) may be more appropriate; if the purpose is to capture long-term trends, a larger time threshold (such as weeks) should be selected. Here, t θ =2 to capture short-term abnormal interactions.

[0095] Step 4: Based on With G * The derived local autocorrelation G statistic has the following formula:

[0096]

[0097] in, and are the binary attribute values ​​after Z-score standardization; w ij is the spatiotemporal weight of the i-th flow and the j-th flow in the spatiotemporal neighborhood; N * is the number of all flows; n * is the number of adjacent flows in the spatiotemporal neighborhood of the i-th flow.

[0098] Step 5: Combine the time threshold selected in step 3 and the BG selected in step 4 * Applied to the BiFlowAMOEBA algorithm, the hierarchical clustering strategy is used to screen seed flows under four clustering modes (HH, HL, LH and LL). If both attribute values ​​of the binary flow after z-score standardization are positive, it is the HH mode; if both attribute values ​​are negative, it is the LL mode; if one attribute value is positive and the other is negative, it is the HL or LH mode.

[0099] The initialization phase sets the seed stream of each mode The value is 0, and the region is merged with the goal of increasing the binary local G statistic. The merging rules are as follows:

[0100]

[0101] in, It is in the merge area BFC i ∪BFC j The number of binary flows in is calculated by calculating the For two clusters BFC i and BFC j ,if is the maximum value, and Gain(BFC i ∪BFC j )>0, it is considered that the two clusters can be merged into a new cluster.

[0102] Step 6: Output the abnormal interaction recognition results and display their spatiotemporal distribution.

[0103] Based on BiFlowAMOEBA algorithm, Figure 3 It shows an abnormal interaction pattern with high population and information flows. In this pattern, only one cluster was finally detected. It can be found that the areas with high population and information flows are Guangzhou, Foshan, Dongguan, Shenzhen and Huizhou. Shenzhen-Dongguan and Shenzhen-Huizhou on October 2nd, and Shenzhen-Huizhou, Shenzhen-Guangzhou, Dongguan-Guangzhou, Dongguan-Huizhou, Guangzhou-Foshan on October 1st. This shows that there is a strong flow of population and information in the first two days of the National Day holiday. The synergy effect of Guangzhou and Foshan in the same city is obvious. The population of Shenzhen is mostly outflow, and it pays more attention to the information of neighboring cities. Compared with other dates, the first two days of the National Day holiday are a period of abnormally high interaction between population and information flows, and the outflow effect in Shenzhen is more obvious.

[0104] Figure 4The abnormal interaction pattern with high population flow and information flow is shown. Three clusters are finally detected under this pattern. The lower abnormal interaction has a larger time span within a week. The first cluster includes Foshan-Dongguan, Zhaoqing-Dongguan, Zhaoqing-Huizhou, Jiangmen-Shenzhen, Jiangmen-Huizhou, Zhaoqing-Shenzhen and Jiangmen-Dongguan on October 2, and Zhaoqing-Huizhou, Jiangmen-Huizhou and Zhuhai-Huizhou on October 1. The starting and ending points of these binary flows are mostly not neighboring counties and cities, and the distance is far. The second cluster includes Foshan-Zhongshan, Zhaoqing-Jiangmen, Zhaoqing-Zhongshan, Zhaoqing-Zhuhai, Jiangmen-Zhuhai and Zhuhai-Jiangmen on October 7, and Zhaoqing-Zhuhai, Zhaoqing-Jiangmen and Zhaoqing-Zhongshan on October 6. This shows that compared with other cities, Zhaoqing City had a lower population outflow and a lower level of attention to other cities during the last two days of the National Day holiday. The third cluster includes Huizhou-Zhongshan, Huizhou-Jiangmen, Huizhou-Zhuhai and Huizhou-Foshan on October 5, and Huizhou-Jiangmen, Huizhou-Zhuhai and Huizhou-Zhaoqing on October 4. This shows that compared with other cities, Huizhou has less population outflow and less attention to other cities during the National Day holiday. The spatiotemporal proximity of the two abnormal interaction patterns can help urban planning departments improve the configuration and optimization of transportation infrastructure and improve the efficiency of resource utilization.

[0105] The implementation basis of each embodiment of the present invention is to implement programmed processing through a device with a processor function. Therefore, in engineering practice, the technical solutions and functions of each embodiment of the present invention are encapsulated into various modules. Based on this reality, on the basis of the above embodiments, an embodiment of the present invention provides a binary geographic flow spatiotemporal interaction anomaly detection device based on spatiotemporal autocorrelation, which is used to execute a binary geographic flow spatiotemporal interaction anomaly detection method based on spatiotemporal autocorrelation in the above method embodiment.

[0106] See also Figure 5The device includes: a first main module, which is used to obtain the study area data set, including the population flow and information flow OD spatiotemporal data of the city-level scale of the study area within a preset period of time represented by the Baidu Migration Index and the Baidu Index respectively; a second main module, which is used to construct a binary global G statistic in the form of multiplication of two univariate global G statistics; a third main module, which is used to construct a spatiotemporal weight using the regional spatial adjacency relationship and the rate of change of the flow value over time, replace the weight in the statistics of the second main module, and obtain the global spatiotemporal autocorrelation results of the binary flow under different time thresholds; a fourth main module, which is used to derive the global spatiotemporal autocorrelation results of the binary flow under different time thresholds using the L statistic and the univariate local G statistic. The formula of binary local G statistics in the generalized form of weights is derived; the fifth main module is used to combine the global spatiotemporal autocorrelation results of binary flows in the third main module, select the set time threshold and apply it to the improved BiFlowAMOEBA algorithm. The algorithm selects seed flows under four clustering modes, and uses the hierarchical clustering strategy in each mode. Under the assumption that there is no in-situ correlation, regional merging is performed with the goal of increasing the binary local G statistics; the sixth main module is used to obtain the spatiotemporal clustering results of the four modes of HH, HL, LH and LL, respectively, and use the spatiotemporal clustering results of HH and LL as the interactive anomaly results of binary flows.

[0107] The embodiment of the present invention provides a binary geographic flow spatiotemporal interaction anomaly detection device based on spatiotemporal autocorrelation. The spatiotemporal autocorrelation of multi-type geographic flows has not been fully explored. The spatial autocorrelation of multi-type flows is mostly studied on local autocorrelation, and there is a lack of research on global autocorrelation. Figure 5 Several modules in the paper take the city-level areas within a single urban agglomeration as the research unit, construct the binary global G statistic using the univariate global G statistic, and select a reasonable time threshold through the global spatiotemporal autocorrelation results; then use the improved binary flow multi-directional optimal ecological zone algorithm to perform spatiotemporal clustering of four modes, HH, HL, LH and LL, under the time threshold, and take the HH and LL clusters as the interactive abnormal areas of the binary flow, so as to further rationally allocate and optimize resource elements and achieve maximum resource utilization efficiency and sustainable development.

[0108] It should be noted that the device embodiments provided by the present invention are not only used to implement the methods in the above-mentioned method embodiments, but also used to implement the methods in other method embodiments provided by the present invention. The only difference lies in the setting of corresponding functional modules, and the principles thereof are basically the same as the principles of the above-mentioned device embodiments provided by the present invention. As long as technical personnel in this field refer to the specific technical solutions in other method embodiments on the basis of the above-mentioned device embodiments, obtain corresponding technical means and technical solutions composed of these technical means by combining technical features, and on the premise of ensuring the practicality of the technical solutions, they will improve the devices in the above-mentioned device embodiments to obtain corresponding device class embodiments for implementing the methods in other method class embodiments.

[0109] Based on the same inventive concept as the above-mentioned embodiment, an embodiment of the present invention also provides a binary geographic flow spatiotemporal interaction anomaly detection system based on spatiotemporal autocorrelation, comprising a memory and a processor; the memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the steps of the binary geographic flow spatiotemporal interaction anomaly detection method based on spatiotemporal autocorrelation.

[0110] Based on the same inventive concept as the above-mentioned embodiment, an embodiment of the present invention also provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions enable the computer to execute the steps of the method for detecting spatiotemporal interaction anomalies of binary geographic flows based on spatiotemporal autocorrelation.

[0111] In summary, the present invention is based on the Baidu migration and Baidu index OD data of binary geographic flow, takes the municipal administrative unit as the research unit, and measures the global spatiotemporal autocorrelation of population flow and information flow based on the method of spatiotemporal autocorrelation. According to the global spatiotemporal autocorrelation result, a reasonable time threshold is selected and applied to the local spatiotemporal autocorrelation, thereby identifying two abnormal interaction modes of high-value abnormal interaction and low-value abnormal interaction. The present invention develops and constructs a global binary G statistic based on the univariate global G statistic, which makes up for the shortcomings of the global autocorrelation research of binary variables; the concept of the rate of change of the attribute value of the flow over time is introduced into the autocorrelation spatiotemporal weight, so as to make the spatiotemporal weight more generalized, thereby deriving a general local binary G statistic expression, making the algorithm of binary flow more reasonable and applicable, and providing a powerful tool for the study of binary flow within and between cities and the maximization of the efficiency of urban multi-factor resource utilization and the optimization of factor resource allocation.

[0112] It should be understood that parts not described in detail in this specification may be considered as prior art.

[0113] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting spatiotemporal interaction anomalies in binary geographic flows based on spatiotemporal autocorrelation, characterized in that: include: Step 1: Obtain the study area dataset, including the group flow OD spatiotemporal data of population flow and information flow within the preset time period represented by Baidu Migration Index and Baidu Index respectively at the city level of the study area; Step 2: Construct a binary global G statistic by multiplying two univariate global G statistics; Step 3: Use the regional spatial adjacency relationship and the rate of change of the flow value over time to construct the spatiotemporal weight, replace the weight in the statistic in step 2, and obtain the global spatiotemporal autocorrelation results of the binary flow under different time thresholds; Step 4: Use the L statistic and the univariate local G statistic to derive the formula of the weighted generalized bivariate local G statistic; Step 5: Combined with the global spatiotemporal autocorrelation results of the binary flow in step 3, the set time threshold is selected and applied to the improved BiFlowAMOEBA algorithm. The algorithm selects seed flows under four clustering modes, and uses the hierarchical clustering strategy in each mode. Under the assumption that there is no in-situ correlation, the region is merged with the goal of increasing the binary local G statistic; Step 6: Obtain the spatiotemporal clustering results of the four modes of HH, HL, LH and LL respectively, and take the spatiotemporal clustering results of HH and LL as the interaction anomaly results of the binary flow.

2. According to claim 1, a method for detecting spatiotemporal interaction anomalies of binary geographic flows based on spatiotemporal autocorrelation is characterized in that: Get the study area dataset, including: Perform unified projection transformation on spatial vector data to form a vector data set with consistent spatial coordinate system; The Baidu Migration and Baidu Index OD data are cleaned, and data from continuous time periods with more valid data are selected. The two types of geographic flow data are merged into the same city and time point, and the city ID and time ID fields are established, where the minimum time unit of the OD data is day.

3. According to claim 1, a method for detecting spatiotemporal interaction anomalies of binary geographic flows based on spatiotemporal autocorrelation is characterized in that: The binary global G statistic is constructed by multiplying two univariate global G statistics: Among them, w ij is the spatiotemporal weight constructed using spatial adjacency and temporal correlation; x i With y i are two attribute values ​​of geographic flow standardized by z-score; x j and j are the two attribute values ​​of the z-score-normalized geographic flow within the spatiotemporal neighborhood of the i-th region.

4. According to claim 1, a method for detecting spatiotemporal interaction anomalies of binary geographic flows based on spatiotemporal autocorrelation is characterized in that: The spatiotemporal weight is constructed by using the regional spatial adjacency relationship and the rate of change of the flow value over time: In ij =sw ij ×tw ij Among them, w ij is the spatiotemporal weight; sw ij is the spatial weight defined by the spatial adjacency; tw ij is the time weight defined by the time distance between flows; the spatial weight and time weight are defined as follows: in, and are the time weights corresponding to the binary attribute values ​​of the flow; x i(t) With y i(t) They represent two attribute values ​​of the flow at time t; x j(t-Δt) and j(t-Δt) are the two attribute values ​​of adjacent flows at time t-Δt; Δt is the time interval; t θ is the time threshold; tw ij is the time weight, which is the average value of the time weight corresponding to the binary attribute value of the flow.

5. According to claim 1, a method for detecting spatiotemporal interaction anomalies of binary geographic flows based on spatiotemporal autocorrelation is characterized in that: The formula of the binary local G statistic in the generalized form of weights is derived using the L statistic and the univariate local G statistic: in, and are the binary attribute values ​​after Z-score standardization; w ij is the spatiotemporal weight of the i-th flow and the j-th flow in the spatiotemporal neighborhood; N * is the number of all flows; n * is the number of adjacent flows in the spatiotemporal neighborhood of the i-th flow.

6. According to claim 1, a method for detecting spatiotemporal interaction anomalies of binary geographic flows based on spatiotemporal autocorrelation is characterized in that: Step 5 also includes: The time threshold selected in step 3 and the binary local G statistic in step 4 are combined Applied to the binary flow multi-directional optimal ecological zone algorithm, the hierarchical clustering strategy is used to screen seed flows in four clustering modes: HH, HL, LH and L. If both attribute values ​​of the binary flow after z-score standardization are positive, it is the HH mode; if both attribute values ​​are negative, it is the LL mode; if one attribute value is positive and the other is negative, it is the HL or LH mode. The initialization phase sets the seed stream of each mode The value is 0, and the region is merged with the goal of increasing the binary local G statistic. The merging rules are as follows: in, It is in the merge area BFC i ∪BFC j The number of binary flows in is calculated by calculating the number of binary flows between each pair of adjacent clusters. For two clusters BFC i and BFC j ,if is the maximum value, and Gain(BFC i ∪BFC j )>0, it is considered that the two clusters can be merged into a new cluster.

7. According to claim 1, a method for detecting spatiotemporal interaction anomalies of binary geographic flows based on spatiotemporal autocorrelation is characterized in that: The method also includes: outputting abnormal interaction recognition results and displaying their spatiotemporal distribution.

8. A device for detecting anomalies of binary geographic flow space-time interaction based on space-time autocorrelation, characterized in that: include: The first main module is used to obtain the dataset of the study area, including the group flow OD spatiotemporal data of population flow and information flow within the preset time period represented by Baidu Migration Index and Baidu Index respectively at the city level of the study area; The second main module is used to construct a binary global G statistic by multiplying two univariate global G statistics; The third main module is used to construct spatiotemporal weights using the regional spatial adjacency relationship and the rate of change of the flow value over time, replace the weights in the statistics of the second main module, and obtain the global spatiotemporal autocorrelation results of binary flows under different time thresholds; The fourth main module is used to derive a weight generalized form of a binary local G statistic formula using the L statistic and the univariate local G statistic; The fifth main module is used to combine the binary flow global spatiotemporal autocorrelation results in the third main module, select the set time threshold and apply it to the improved BiFlowAMOEBA algorithm. The algorithm selects seed flows under four clustering modes, uses the hierarchical clustering strategy in each mode, and performs regional merging with the goal of increasing the binary local G statistic under the assumption that there is no in-situ correlation; The sixth main module is used to obtain the spatiotemporal clustering results of the four modes of HH, HL, LH and LL respectively, and take the spatiotemporal clustering results of HH and LL as the interaction anomaly results of the binary flow.

9. A binary geographic flow spatiotemporal interaction anomaly detection system based on spatiotemporal autocorrelation, characterized in that: It includes a memory and a processor; the memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the steps of a binary geographic flow spatiotemporal interaction anomaly detection method based on spatiotemporal autocorrelation as described in any one of claims 1 to 7.

10. A non-transitory computer readable storage medium, characterized in that: The non-transitory computer-readable storage medium stores computer instructions, which enable the computer to execute the steps of a method for detecting spatiotemporal interaction anomalies of binary geographic flows based on spatiotemporal autocorrelation as described in any one of claims 1 to 7.

Images