A method for detecting a section of imbalance between supply and demand of taxis

Abstract

The application is suitable for the field of spatio-temporal data mining technology, and provides a taxi supply and demand imbalance road section detection method, which comprises the following steps: extracting a GPS data entry sequence corresponding to each license plate from a taxi GPS original data set in chronological order; determining a taxi supply point and a taxi demand point according to all GPS data entry sequences; determining a correction point of the taxi supply point and the taxi demand point in a city road network; finely dividing the city road network to obtain a final city road network; taking a road section with a taxi demand correction point in the final city road network as a seed road section, and expanding the seed road section along the final city road network through an iterative optimization method to obtain a plurality of candidate taxi supply and demand imbalance road sections; performing significance test by using a Monte Carlo simulation method, and taking a candidate taxi supply and demand imbalance road section that passes the significance test as a taxi supply and demand imbalance road section. The application can improve the accuracy of taxi supply and demand imbalance road section detection.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of spatio-temporal data mining, and particularly relates to a method for detecting a taxi supply-demand imbalance section. BACKGROUND

[0002] As a major public transportation auxiliary method, taxis play a vital role in meeting the travel needs of citizens. Therefore, the research on the spatial distribution of taxi supply and demand involving multiple factors such as the scale of taxi operation, price mechanism, market conditions, government policy and industry development trend has become an important research field across multiple disciplines, and the detection of taxi supply-demand imbalance sections has attracted great attention. Taxi supply-demand imbalance not only affects the efficiency and quality of taxi service, but also directly affects the travel experience of passengers and the income level of drivers. Taxi supply-demand mismatch leads to the occurrence of events such as prolonged waiting time for passengers and unstable income for drivers. Therefore, accurately identifying these sections and developing appropriate policies and strategies are crucial for improving the overall efficiency of the transportation system.

[0003] Traditional survey data has the characteristics of low spatio-temporal resolution, and faces great challenges in accurately perceiving the degree of taxi supply-demand imbalance. The widespread use of various sensors provides new impetus for exploring urban phenomena and urban dynamics, and provides a way to effectively obtain key data of urban conditions, enabling decision-makers to have a deeper understanding of the operation of the city, so as to develop effective management and development strategies. Taxi trajectory data obtained from Global Positioning System (GPS) positioning provides a reliable data source for accurately identifying the spatio-temporal dimension of abnormal taxi supply-demand conditions.

[0004] At present, the study of taxi imbalance can be divided into two categories: macro-scale and micro-scale. Macro-scale studies evaluate the overall taxi imbalance of the city, for example: in 2015, Liu used statistical data to construct the supply-demand coordination coefficient, supply-demand elasticity, passenger load rate, mileage utilization rate, and 10,000 people indicators to evaluate the supply-demand matching of major cities in China; Li used BP neural network algorithm to analyze the changes and trends of taxi supply-demand matching rate and matching index in Wuhan based on city statistical yearbook data; Kamga et al. explored the dynamic changes of taxi supply and demand in New York City based on taxi GPS trip records, and found that the number of passengers, taxi availability, trip distance and pick-up frequency fluctuate significantly in different time periods and weather conditions; in 2018, Yang et al. developed an evaluation system and passenger satisfaction optimization model using taxi density and passenger demand density obtained from GPS data. The framework aims to analyze the supply-demand matching of taxi resources. Micro-scale studies focus more on the supply-demand imbalance within the city, which can further locate the specific location of the taxi supply-demand. For example, in 2012, Huang et al. proposed a research framework to detect taxi supply shortage and oversupply areas in a probabilistic manner, using Bayesian spatial scanning statistical methods and Poisson-based hypothesis testing methods to extract areas with supply-demand imbalance by setting thresholds; in 2017, Zhang et al. proposed a DEGBDT model that fully utilizes the advantages of various features to minimize the loss caused by feature sparsity when predicting taxi supply-demand gaps. In 2020, Cai defined taxi supply-demand spatial distribution imbalance as a spatial anomaly based on the definition of local spatial patterns in spatial data mining, and proposed an anomaly spatial mining method that does not limit the shape of the distribution to detect areas of taxi supply-demand imbalance in the city.

[0005] Through analysis, it can be found that the existing macro-scale method can only provide a perception of the overall taxi supply-demand imbalance of the city, and cannot provide fine guidance for the spatial partial regulation of the city's taxi. The micro-scale detection method is based on the Euclidean plane space assumption, and directly expresses the spatial distance of taxi supply-demand points through straight-line distance, without considering the problem that taxi activity space is constrained by road network, thus easily causing false estimation of taxi supply-demand balance state, and further leading to low accuracy of taxi supply-demand imbalance road segment detection. SUMMARY

[0006] The embodiments of the present application provide a taxi supply-demand imbalance road segment detection method, which can solve the problem of low accuracy of taxi supply-demand imbalance road segment detection.

[0007] The embodiments of the present application provide a taxi supply-demand imbalance road segment detection method, which can solve the problem of low accuracy of taxi supply-demand imbalance road segment detection.

[0008] Step 1, extract all GPS data entries corresponding to each license plate number from the taxi GPS raw data set, and respectively for each license plate number, sort all GPS data entries corresponding to the license plate number in chronological order to obtain the GPS data entry sequence of the license plate number; the GPS raw data set includes GPS data entries of all taxis detected in the city;

[0009] Step 2, determine the taxi supply point and the taxi demand point according to all GPS data entry sequences;

[0010] Step 3, match the taxi supply point and the taxi demand point to the city road network of the detected city to determine the taxi supply correction point of the taxi supply point and the taxi demand correction point of the taxi demand point in the city road network;

[0011] Step 4, divide at least part of the roads in the city road network based on the lengths of all roads between different intersections in the city road network to obtain the final city road network;

[0012] Step 5, take the road segment with the taxi demand correction point in the final city road network as a seed road segment, and based on the taxi supply correction point, expand each seed road segment along the final city road network by iterative optimization to obtain multiple candidate taxi supply and demand imbalance road segments;

[0013] Step 6, use the Monte Carlo simulation method to perform significance test on each candidate taxi supply and demand imbalance road segment, and take the candidate taxi supply and demand imbalance road segment that passes the significance test as the taxi supply and demand imbalance road segment.

[0014] Optionally, step 2 includes:

[0015] Respectively for each GPS data entry sequence, the following steps are performed:

[0016] Step 2.1, for a target GPS data entry in the GPS data entry sequence whose state is empty, if the GPS data entry at the next time point in the GPS data entry sequence is empty, the target GPS data entry is taken as a taxi supply point GPS data entry;

[0017] Step 2.2, for a target GPS data entry in the GPS data entry sequence whose state is passenger, if the GPS data entry at the previous time point in the GPS data entry sequence is empty, the target GPS data entry is taken as a taxi demand point GPS data entry;

[0018] Step 2.3, project the latitude and longitude coordinate field in the taxi supply point GPS data entry into a two-dimensional space to form a taxi supply point;

[0019] Step 2.4, project the longitude and latitude coordinate field in the taxi demand point GPS data entry into two-dimensional space to form a taxi demand point.

[0020] Optionally, step 3 includes:

[0021] The determined taxi supply point and taxi demand point are both regarded as a target point, and the following steps are performed for each target point respectively:

[0022] Step 3.1, determine the GPS data entry corresponding to the target point from the sequence of GPS data entries to which the target point belongs, and determine the GPS data entry at the previous time and the GPS data entry at the next time of the GPS data entry in the sequence of GPS data entries;

[0023] Step 3.2, project the longitude and latitude coordinate field in the determined GPS data entry at the previous time into two-dimensional space to form a previous time position point;

[0024] Step 3.3, project the longitude and latitude coordinate field in the determined GPS data entry at the next time into two-dimensional space to form a next time position point;

[0025] Step 3.4, regard the road segment in the urban road network with a shortest straight line distance less than r to the target point as a first candidate road segment, regard the projection point of the target point on each first candidate road segment as a candidate correction point of the target point, and regard the set of all candidate correction points of the target point as a first candidate correction point set;

[0026] Step 3.5, regard the road segment in the urban road network with a shortest straight line distance less than r to the previous time position point as a second candidate road segment, regard the projection point of the previous time position point on each second candidate road segment as a candidate correction point of the previous time position point, and regard the set of all candidate correction points of the previous time position point as a second candidate correction point set;

[0027] Step 3.6, regard the road segment in the urban road network with a shortest straight line distance less than r to the next time position point as a third candidate road segment, regard the projection point of the next time position point on each third candidate road segment as a candidate correction point of the next time position point, and regard the set of all candidate correction points of the next time position point as a third candidate correction point set;

[0028] Step 3.7, combine the candidate correction points in the first candidate correction point set, the second candidate correction point set and the third candidate correction point set to obtain a plurality of correction point combinations; each correction point combination contains one candidate correction point in the first candidate correction point set, one candidate correction point in the second candidate correction point set and one candidate correction point in the third candidate correction point set;

[0029] Step 3.8: Calculate the matching degree of each correction point combination, and take the candidate correction point belonging to the first candidate correction point set in the correction point combination with the largest matching degree as the correction point of the target point; where the correction point corresponding to the taxi supply point is the car rental supply correction point, and the correction point corresponding to the taxi demand point is the car rental demand correction point.

[0030] Optionally, calculate the matching degree for each combination of correction points, including:

[0031] Through the formula F(L) ca )=p(C j )*T(C 1n →C j )+p(C 2m )*T(C j →C 2m ) Calculate the matching degree F(L) of the correction point combination ca );

[0032] Wherein, p(C j ) represents a candidate correction point C that belongs to the first candidate correction point set in the combination of correction points. j The probability of being selected as a correction point. σ c Dist(P,C) represents the standard deviation of a normal distribution. j ) represents the target point P and the candidate correction point C. j The straight-line distance between them, μ c T(C) represents the mean of a normal distribution. 1n →C j ) represents a candidate correction point C that belongs to the first candidate correction point set in the combination of correction points. j Candidate correction point C belonging to the second candidate correction point set 1n The transition probability between them dist(P1,P) represents the straight-line distance between the previous position point P1 and the target point P. net_dist(C 1n C j ) indicates candidate correction point C 1n With candidate correction point C j The shortest path distance between them along the urban road network, p(C 2m ) represents the candidate correction point C that belongs to the third candidate correction point set in the combination of correction points. 2m The probability of being selected as a correction point. dist(P,C 2m ) represents the target point P and the candidate correction point C. 2m The straight-line distance between them, T(C) j →C 2m) represents a candidate correction point C belonging to the first candidate correction point set in the combination of correction points j and a candidate correction point C belonging to the third candidate correction point set 2m , the transition probability between dist(P, P2) represents the straight-line distance between the target point P and the position point P2 at the next time, net_dist(C j → C 2m ) represents the shortest path distance along the urban road network between the candidate correction point C j and the candidate correction point C 2m .

[0033] Optionally, step 4 comprises:

[0034] Step 4.1, calculate the lengths of all roads between different intersections in the urban road network, and calculate the average value of the lengths of all roads;

[0035] Step 4.2, divide the road segments in the urban road network with lengths greater than the average value into N equal parts, and take each equal part as a road segment to obtain the final urban road network; the length of each equal part is less than or equal to the average value.

[0036] Optionally, based on the car rental supply correction points, each seed road segment is expanded along the final urban road network by an iterative optimization method to obtain a plurality of candidate car rental supply and demand imbalance road segments, including:

[0037] The following steps are performed respectively for each seed road segment:

[0038] Step 5.1, for each car rental demand correction point on the seed road segment, the number of car rental supply correction points with a shortest path distance less than a threshold value D from the car rental demand correction point is taken as the network parity strength of the car rental demand correction point;

[0039] Step 5.2, calculate the log-likelihood ratio of the seed road segment according to the network parity strength, and calculate the abnormality degree of the seed road segment according to the log-likelihood ratio;

[0040] Step 5.3, respectively combine the seed road segment with each road segment connected to the seed road segment to obtain a plurality of new road segments, and calculate the abnormality degree of each new road segment;

[0041] Step 5.4, the new road segment corresponding to the largest abnormality degree in the abnormality degrees of all new road segments is taken as the seed road segment;

[0042] Step 5.5: Determine whether the degree of abnormality of the seed road segment in step 5.4 is greater than the degree of abnormality of the seed road segment in step 5.2. If it is greater, then the seed road segment in step 5.4 is used as the seed road segment in step 5.3, and the process returns to step 5.3. If the degree of abnormality of the seed road segment in step 5.4 is equal to the degree of abnormality of the seed road segment in step 5.2, then the seed road segment in step 5.4 is used as the candidate taxi supply and demand imbalance road segment.

[0043] Optionally, step 5.2 includes:

[0044] Through formula Calculate the log-likelihood ratio log of seed segment z. * (z); N represents the number of car rental demand correction points on the final urban road network, μ and σ represent the mean and standard deviation of the network isotope strength of all car rental demand correction points on the final urban road network, respectively, x i This represents the network co-location strength at the i-th car rental demand correction point. The variance of the network isotope strength for each car rental correction point within seed segment z. x z μ represents the sum of network isotropic strengths of the car rental demand correction points within the seed segment z. z and λ z Let n represent the average network isotope strength of the car rental demand correction points inside and outside the seed segment z, respectively. z X represents the number of car rental demand correction points within seed segment z, and X represents the sum of the network isotropic strength of all car rental demand correction points.

[0045] If the target is a road segment with excessive supply, then the log-likelihood of the seed road segment z is compared to log... * (z) and indicator function I(μ) z >λ z The product of ) is used as the degree of anomaly of the seed segment z;

[0046] If the target segment is a road with insufficient supply, then the log-likelihood of the seed segment z is compared to log... * (z) and indicator function I(μ) z <λ z The product of ) is used as the anomaly degree of seed segment z.

[0047] Optionally, the detection methods also include:

[0048] If multiple candidate taxi supply and demand imbalance sections overlap, then only the candidate taxi supply and demand imbalance section with the highest degree of abnormality among these multiple candidate taxi supply and demand imbalance sections will be retained.

[0049] Optionally, step 6 includes:

[0050] Steps 6.1 and 6.2 are performed for each seed road segment respectively:

[0051] Step 6.1, a new network parity strength is randomly re-assigned to each rental demand correction point on the seed road segment, and the process is repeated K times to obtain K sets of new network parity strengths;

[0052] Step 6.2, for each set of new network parity strengths, the set of new network parity strengths is used as the network parity strength in step 5.2 to perform steps 5.2 to 5.5 in step 5.2, and the maximum abnormality degree is retained;

[0053] Step 6.3, the significance p-value(Z m ) of the mth candidate rental supply and demand imbalance road segment Zm is calculated by the following formula:

[0054]

[0055] wherein, represents the maximum abnormality degree corresponding to the kth set of new network parity strengths, AL obs represents the abnormality degree of the mth candidate rental supply and demand imbalance road segment, I represents an indicator function, and I takes 1 when is 1, otherwise 0;

[0056] Step 6.4, the significances of all candidate rental supply and demand imbalance road segments are sorted in ascending order to obtain p-value(Z1)≤p-value(Z2)≤……≤p-value(Z m )≤……≤p-value(Z M ); M represents the number of candidate rental supply and demand imbalance road segments, and Z M represents the Mth candidate rental supply and demand imbalance road segment;

[0057] Step 6.5, if the significance p-value(Z m ) of the mth candidate rental supply and demand imbalance road segment Zm is less than or equal to the significance level a adj , the mth candidate rental supply and demand imbalance road segment Zm is taken as the rental supply and demand imbalance road segment;

[0058] wherein, a adj =p-value(Z q ), q is the maximum subscript value of significance that satisfies the condition , and a is a preset significance level.

[0059] The above scheme of the present application has the following beneficial effects:

[0060] In the embodiment of the present application, by matching the taxi supply points and taxi demand points determined based on taxi GPS raw data to the urban road network, the correction of the taxi supply and demand location points is realized, and then the candidate taxi supply and demand imbalance road segments in the urban road network are identified based on the corrected correction points, and the Monte Carlo simulation method is used to perform significance test on the candidate taxi supply and demand imbalance road segments, and the candidate taxi supply and demand imbalance road segments that pass the significance test are taken as the taxi supply and demand imbalance road segments. Wherein, since the spatial correlation between the taxi and the road network is considered when detecting the taxi supply and demand imbalance road segments, the road segments with the taxi supply and demand imbalance can be accurately located, the accuracy and interpretability of the detection of the taxi supply and demand imbalance road segments are greatly improved, and the practicability of the present application is enhanced.

[0061] In addition, since the taxi imbalance degree is determined by the non-human determination threshold in the present application, the subjectivity of the evaluation result is reduced, and the reliability of the present application is enhanced.

[0062] Other beneficial effects of the present application will be described in detail in the subsequent specific embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiments or the prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creating laborious work.

[0064] Figure 1 The flow chart of the taxi supply and demand imbalance road segment detection method provided by an embodiment of the present application;

[0065] Figure 2 The spatial distribution diagram of the taxi supply points and demand points in an example of the present application;

[0066] Figure 3 The schematic diagram of the supply and demand imbalance road segment identification result in an example of the present application. DETAILED DESCRIPTION

[0067] In the following description, specific details such as specific system structures, techniques, etc. are presented in order to thoroughly understand the embodiments of the present application. However, it should be clear to those skilled in the art that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits and methods are omitted to avoid unnecessary details that hinder the description of the present application.

[0068] It will be understood that the terms “comprises”, “comprising”, “includes”, “including”, “has”, “having” and variants thereof when used in this specification and in the following claims are not to be interpreted in an excluding sense but in an inclusive sense. That is, they will be understood to allow for elements, features, steps, or acts that would otherwise be excluded.

[0069] It will be further understood that the terms “comprises”, “comprising”, “includes”, “including”, “has”, “having” and variants thereof when used in this specification and in the following claims are not to be interpreted in an excluding sense but in an inclusive sense. That is, they will be understood to allow for elements, features, steps, or acts that would otherwise be excluded.

[0070] As used in this specification and claims, the terms “if’ and “when” can be interpreted to mean “upon” or “in response to determining,” or “in response to detecting,” depending on the context. Similarly, the phrase “if it is determined” or “if [a described condition or event] is detected” can be interpreted to mean “upon determining” or “in response to determining,” or “upon detecting [the described condition or event]” or “in response to detecting [the described condition or event],” depending on the context.

[0071] In addition, the description in the specification of this application and the appended claims, the terms “first”, “second”, “third”, etc. are used merely as labels for convenience, and are not intended to imply or create a relative importance of the described elements.

[0072] Reference throughout this specification to “one embodiment” or “an embodiment” or “some embodiments” means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. Thus, appearances of the phrases “in one embodiment”, “in some embodiments”, “in other embodiments”, “in additional embodiments”, and so on, in various places throughout this specification are not necessarily all referring to the same embodiment, unless otherwise specified. The terms “comprise”, “comprising”, “has”, “having”, and variants thereof, are meant to be interpreted broadly in an inclusive and not an exclusive sense, unless otherwise specifically indicated.

[0073] In view of the low accuracy of the current taxi supply and demand imbalance section detection, the embodiment of the present application provides a taxi supply and demand imbalance section detection method. The method matches the taxi supply points and taxi demand points determined based on taxi GPS original data to the urban road network, corrects the taxi supply and demand position points, and then identifies candidate taxi supply and demand imbalance sections in the urban road network based on the corrected points. The Monte Carlo simulation method is used to perform significance test on the candidate taxi supply and demand imbalance sections, and the candidate taxi supply and demand imbalance sections that pass the significance test are used as the taxi supply and demand imbalance sections. Since the spatial correlation between the taxi and the road network is considered during the detection of the taxi supply and demand imbalance section, the section with the taxi supply and demand imbalance can be accurately located, the accuracy and interpretability of the taxi supply and demand imbalance section detection are greatly improved, and the practicability of the present application is enhanced.

[0074] In addition, since the present application determines the taxi imbalance degree in a non-human manner, the subjectivity of the evaluation result is reduced, thereby enhancing the reliability of the present application.

[0075] The taxi supply and demand imbalance section detection method provided by the embodiment of the present application will be exemplarily described below in combination with specific embodiments.

[0076] As shown in Figure 1 The taxi supply and demand imbalance section detection method provided by the embodiment of the present application includes the following steps:

[0077] Step 1: Extract all GPS data entries corresponding to each license plate number from the taxi GPS original data set, and sort all GPS data entries corresponding to each license plate number in chronological order to obtain the GPS data entry sequence of the license plate number.

[0078] The above-mentioned GPS original data set includes GPS data entries of all taxis in the city. The GPS data entries can be GPS positioning data carrying time and taxi passenger state (such as empty car and passenger carrying). In actual application, the GPS data entries in a specific time period (such as morning peak period and evening peak period) can be used for taxi supply and demand imbalance section detection.

[0079] In some embodiments of the present application, after obtaining the above-mentioned taxi GPS original data set at a certain sampling frequency, the GPS data entries with the same license plate number can be sorted in ascending order of time to obtain the GPS data entry sequence corresponding to each license plate number.

[0080] Step 2: Determine the taxi supply points and taxi demand points according to all GPS data entry sequences.

[0081] In some embodiments of the present application, the following steps can be performed to extract the taxi supply points and taxi demand points respectively for each sequence of GPS data entries:

[0082] Step 2.1, for a target GPS data entry in the sequence of GPS data entries whose status is empty, if the status of the GPS data entry at the next time point of the target GPS data entry in the sequence of GPS data entries is empty, the target GPS data entry is taken as a taxi supply point GPS data entry.

[0083] Step 2.2, for a target GPS data entry in the sequence of GPS data entries whose status is passenger, if the status of the GPS data entry at the previous time point of the target GPS data entry in the sequence of GPS data entries is empty, the target GPS data entry is taken as a taxi demand point GPS data entry.

[0084] Step 2.3, the longitude and latitude coordinate field in the taxi supply point GPS data entry is projected into a two-dimensional space to form a taxi supply point.

[0085] Step 2.4, the longitude and latitude coordinate field in the taxi demand point GPS data entry is projected into a two-dimensional space to form a taxi demand point.

[0086] It should be noted that the above two-dimensional space can be understood as a two-dimensional space in which the urban road network of the surveyed city is located, so as to establish the spatial correlation between the taxi and the road network. Specifically, the longitude and latitude coordinate field can be projected into a two-dimensional space by using a common projection method (such as equidistant cylindrical projection) to obtain the corresponding coordinate position of the GPS data entry in the two-dimensional space (i.e. the above-mentioned taxi supply point and taxi demand point).

[0087] Step 3, by matching the taxi supply point and the taxi demand point to the urban road network of the surveyed city, the taxi supply correction point of the taxi supply point and the taxi demand correction point of the taxi demand point are determined in the urban road network.

[0088] After the taxi supply point and the taxi demand point are determined, the positions of the two points need to be corrected in the urban road network to accurately establish the spatial correlation between the taxi and the road network. In the correction process, the correction processes of the taxi supply point and the taxi demand point are the same, to avoid too much repetition, the correction process is exemplarily described by taking a target point (the target point represents the taxi supply point or the taxi demand point) as an example.

[0089] Specifically, the determined taxi supply point and taxi demand point can be taken as a target point, and the following steps can be performed for each target point:

[0090] Step 3.1: Determine the GPS data entry corresponding to the target point P from the GPS data entry sequence, and determine the GPS data entry of the previous time and the GPS data entry of the next time in the GPS data entry sequence.

[0091] Step 3.2: Project the latitude and longitude coordinate fields from the determined GPS data entry of the previous moment into a two-dimensional space to form the location point P1 of the previous moment.

[0092] Step 3.3: Project the latitude and longitude coordinate fields in the determined GPS data entry for the next time moment into a two-dimensional space to form the location point P2 for the next time moment.

[0093] In some embodiments of this application, common projection methods (such as equidistant cylindrical projection) can be used to project latitude and longitude coordinate fields onto a two-dimensional space to obtain the coordinate positions of the previous position point P1 and the next position point P2.

[0094] Step 3.4: Select road segments in the urban road network whose shortest straight-line distance to the target point P is less than r as first candidate road segments, and select the projection points of the target point on each first candidate road segment as candidate correction points of the target point. The set of all candidate correction points of the target point is then used as the first candidate correction point set C_set, where C_set = {C1, C2, ..., C...} j ,…,C J}. Where r can be set according to the actual situation, C j Let J represent the j-th candidate correction point of the target point, where j = 1, 2, ..., J, and J is the number of candidate correction points of the target point.

[0095] Step 3.5: Select road segments in the urban road network whose shortest straight-line distance to the previous time point P1 is less than r as second candidate road segments. Then, select the projection points of the previous time point P1 onto each of these second candidate road segments as candidate correction points for the previous time point. Finally, select the set of all candidate correction points for the previous time point as the second candidate correction point set C_set1, where C_set1 = {C...} 11 C 12 ,…,C 1n ,…,C 1N}, C 1n This represents the nth candidate correction point of the previous time position, where n = 1, 2, ..., N, and N is the number of candidate correction points of the previous time position.

[0096] Step 3.6, taking the road segment in the urban road network with a shortest straight-line distance to the position point P2 at the next time less than r as a third candidate road segment, and taking the projection point of the position point P2 at the next time on each third candidate road segment as a candidate correction point of the position point at the next time, taking the set of all candidate correction points of the position point at the next time as a third candidate correction point set C_set2 = {C 21 , 22 , 2m , 2M , 2m}, C ca represents the mth candidate correction point of the position point at the next time, m = 1, 2, …, M, and M is the number of candidate correction points of the position point at the next time.

[0097] Step 3.7, combining the candidate correction points in the first candidate correction point set, the second candidate correction point set and the third candidate correction point set to obtain a plurality of correction point combinations. Each correction point combination contains one candidate correction point in the first candidate correction point set, one candidate correction point in the second candidate correction point set and one candidate correction point in the third candidate correction point set.

[0098] In some embodiments of the present application, all possible correction point combinations can be obtained by combining one candidate correction point from each of the first candidate correction point set, the second candidate correction point set and the third candidate correction point set.

[0099] Step 3.8, calculating the matching degree of each correction point combination, and taking the candidate correction point belonging to the first candidate correction point set in the correction point combination with the largest matching degree as the correction point of the target point.

[0100] Wherein, the correction point corresponding to the taxi supply point is a taxi supply correction point, and the correction point corresponding to the taxi demand point is a taxi demand correction point.

[0101] In some embodiments of the present application, the matching degree F(L ca ) of the correction point combination can be calculated by the following formula:

[0102] F(L ca ) = p(C j ))*T(C 1n →C j )+p(C 2m )*T(C j →C 2m )

[0103] Wherein, p(C j ) represents the probability of the candidate correction point C j belonging to the first candidate correction point set being selected as the correction point, σ cDist(P,C) represents the standard deviation of a normal distribution. j ) represents the target point P and the candidate correction point C. j The straight-line distance between them, μ c T(C) represents the mean of a normal distribution. 1n →C j ) represents a candidate correction point C that belongs to the first candidate correction point set in the combination of correction points. j Candidate correction point C belonging to the second candidate correction point set 1n The transition probability between them dist(P1,P) represents the straight-line distance between the previous position point P1 and the target point P. net_dist(C 1n C j ) indicates candidate correction point C 1n With candidate correction point C j The shortest path distance between them along the urban road network, p(C 2m ) represents the candidate correction point C that belongs to the third candidate correction point set in the combination of correction points. 2m The probability of being selected as a correction point. dist(P,C 2m ) represents the target point P and the candidate correction point C. 2m The straight-line distance between them, T(C) j →C 2m ) represents a candidate correction point C that belongs to the first candidate correction point set in the combination of correction points. j Candidate correction point C belonging to the third candidate correction point set 2m The transition probability between them dist(P,P2) represents the straight-line distance between the target point P and the next time step position P2. net_dist(C j →C 2m ) indicates candidate correction point C j With candidate correction point C 2m The shortest path distance between them along the city's road network.

[0104] Step 4: Based on the length of all roads between different intersections in the urban road network, divide at least a portion of the roads in the urban road network to obtain the final urban road network.

[0105] In some embodiments of this application, the urban road network can be divided into more uniform and refined road segments by processing the roads between different intersections. These segments serve as basic detection units for assessing the extent of taxi supply-demand imbalance within the road network. Specifically, the final urban road network can be obtained through the following division steps:

[0106] Step 4.1, calculate the length of all roads between different intersections in the urban road network, and calculate the average value of the length of all roads

[0107] Step 4.2, for roads (i.e. road segments) with a length less than or equal to , directly as a road segment; divide the road segments in the urban road network with a length greater than the average value into N equal parts, and each part as a road segment, to obtain the final urban road network; the length of each part is less than or equal to the average value. Wherein, N is the smallest integer value that satisfies the condition that the length of each divided road segment is not greater than .

[0108] It should be noted that step 4 is to subdivide the urban road network, and the corrected points after step 3 are distributed in the final urban road network.

[0109] Step 5, taking the road segments with car rental demand corrected points in the final urban road network as seed road segments, and based on the car rental supply corrected points, each seed road segment is expanded along the final urban road network by iterative optimization, to obtain multiple candidate car rental supply and demand imbalance road segments.

[0110] In some embodiments of the present application, the following steps can be performed respectively for each seed road segment to determine all candidate car rental supply and demand imbalance road segments in the final urban road network:

[0111] Step 5.1, for each car rental demand corrected point on the seed road segment, the number of car rental supply corrected points with a shortest path distance less than a threshold value D from the car rental demand corrected point is taken as the network parity strength of the car rental demand corrected point.

[0112] The above threshold value D is a preset value, which can be set according to actual conditions. Here, it is necessary to determine the network parity strength of each car rental demand corrected point on the seed road segment. Here, the network parity strength is defined as the number of car rental supply corrected points with a shortest path distance less than the threshold value D from the car rental demand corrected point in the urban road network.

[0113] Step 5.2, calculate the log-likelihood ratio of the seed road segment according to the network parity strength, and calculate the abnormality degree of the seed road segment according to the log-likelihood ratio. The abnormality degree is used to describe the degree of car rental supply and demand imbalance of the seed road segment, and the greater the value of the abnormality degree, the higher the degree of car rental supply and demand imbalance.

[0114] Specifically, the log-likelihood ratio log of the seed road segment z can be calculated by the formula *(z); N represents the number of rental car demand calibration points on the final urban road network, μ and σ respectively represent the mean and standard deviation of the network homotopy intensity of all rental car demand calibration points on the final urban road network, and are calculated as μ = X / N, x i represents the network homotopy intensity of the i-th rental car demand calibration point, is the variance of the network homotopy intensity of each rental car supply calibration point within the seed road segment z, x z represents the sum of the network homotopy intensity of the rental car demand calibration points within the seed road segment z, μ z and λ z respectively represent the mean of the network homotopy intensity of the rental car demand calibration points within and outside the seed road segment z, and are calculated as μ z = x z / n z , λ z = (X-x z ) / (N-n z ), n z represents the number of rental car demand calibration points within the seed road segment z, and X represents the sum of the network homotopy intensity of all rental car demand calibration points.

[0115] If the over-supplied road segment needs to be detected, the product of the log-likelihood ratio log * (z) of the seed road segment z and the indicator function I(μ z > λ z ) is taken as the abnormality degree of the seed road segment z, and the indicator function I(μ z > λ z ) takes the value of 1 when μ z > λ z , otherwise it takes the value of 0; if the under-supplied road segment needs to be detected, the product of the log-likelihood ratio log * (z) of the seed road segment z and the indicator function I(μ z < λ z ) is taken as the abnormality degree of the seed road segment z, and the indicator function I(μ z < λ z ) takes the value of 1 when μ z < λ z , otherwise it takes the value of 0. In practical applications, the calculation method of the abnormality degree can be determined according to the actual detection needs.

[0116] wherein the over-supplied road segment refers to the number of empty rental cars around the rental car demand within the road segment being higher than that in other areas of the urban road network, and the under-supplied road segment refers to the number of empty rental cars around the rental car demand within the road segment being significantly less than that in other areas of the urban road network.

[0117] Step 5.3, combine each road segment connected with the seed road segment respectively with the seed road segment to obtain a plurality of new road segments, and calculate the abnormality degree of each new road segment.

[0118] It should be noted that the calculation method of the abnormality degree of the new road segment is the same as the calculation method of the abnormality degree of the seed road segment in step 5.2, that is, the value of the abnormality degree of the new road segment can be calculated by taking the new road segment as the seed road segment in step 5.2.

[0119] Step 5.4, take the new road segment corresponding to the abnormality degree with the largest value among the abnormality degrees of all new road segments as the seed road segment.

[0120] Step 5.5, judge whether the abnormality degree of the seed road segment in step 5.4 is greater than the abnormality degree of the seed road segment in step 5.2, if greater, take the seed road segment in step 5.4 as the seed road segment in step 5.3, and return to execute step 5.3, if the abnormality degree of the seed road segment in step 5.4 is equal to the abnormality degree of the seed road segment in step 5.2, take the seed road segment in step 5.4 as the candidate taxi supply and demand imbalance road segment.

[0121] That is, in some embodiments of the present application, the candidate taxi supply and demand imbalance road segment can be locked by repeating the iterative expansion process until the abnormality degree of the seed road segment no longer increases.

[0122] It can be understood that if there are overlapping road segments in a plurality of candidate taxi supply and demand imbalance road segments among all candidate taxi supply and demand imbalance road segments, only the candidate taxi supply and demand imbalance road segment with the largest abnormality degree among the plurality of candidate taxi supply and demand imbalance road segments is retained. For example, there are overlapping road segments between the candidate taxi supply and demand imbalance road segment A and the candidate taxi supply and demand imbalance road segment B, then the abnormality degrees of the candidate taxi supply and demand imbalance road segment A and the candidate taxi supply and demand imbalance road segment B are compared, if the abnormality degree of the candidate taxi supply and demand imbalance road segment A is greater than the abnormality degree of the candidate taxi supply and demand imbalance road segment B, the candidate taxi supply and demand imbalance road segment A is retained and the candidate taxi supply and demand imbalance road segment B is deleted.

[0123] Step 6, perform significance test on each candidate taxi supply and demand imbalance road segment by using the method of Monte Carlo simulation, and take the candidate taxi supply and demand imbalance road segment passing the significance test as the taxi supply and demand imbalance road segment.

[0124] The above taxi supply-demand imbalance road sections are divided into two types, one is a taxi supply-demand imbalance road section with excessive supply, and the other is a taxi supply-demand imbalance road section with insufficient supply. It can be understood that, if the road section with excessive supply needs to be detected, the candidate taxi supply-demand imbalance road section finally passing the significance test is the taxi supply-demand imbalance road section with excessive supply; if the road section with insufficient supply needs to be detected, the candidate taxi supply-demand imbalance road section finally passing the significance test is the taxi supply-demand imbalance road section with insufficient supply.

[0125] In some embodiments of the present application, the specific implementation of step 6 comprises the following steps:

[0126] First, steps 6.1 and 6.2 are performed for each seed road section (i.e., the road section with taxi demand correction points in step 5) to obtain an anomaly degree set AL null :

[0127] Step 6.1, randomly reassign a new network parity strength to each taxi demand correction point on the seed road section, and repeat this process K times to obtain K sets of new network parity strengths.

[0128] Step 6.2, for each set of new network parity strengths, execute steps 5.2 to 5.5 with the set of new network parity strengths as the network parity strength in step 5.2, and keep the maximum anomaly degree. That is, for a set of new network parity strengths, if the anomaly degree of the seed road section no longer increases during the iterative expansion process based on the set of new network parity strengths, the value of the anomaly degree is kept (i.e., the value of the anomaly degree of the candidate taxi supply-demand imbalance road section obtained by executing steps 5.2 to 5.5 based on the set of new network parity strengths is kept). It should be noted that the set of all maximum anomaly degrees is the above anomaly degree set AL null .

[0129] After obtaining the anomaly degree set AL null , the final taxi supply-demand imbalance road section can be determined by executing the following steps 6.3 to 6.5:

[0130] Step 6.3, calculate the significance p-value (Z m ) of the mth candidate taxi supply-demand imbalance road section Zm by the following formula:

[0131]

[0132] wherein, represents the maximum anomaly degree corresponding to the kth set of new network parity strengths, AL obs represents the anomaly degree of the mth candidate taxi supply-demand imbalance road section, I represents an indicator function, and I is 1 when If yes, take 1; otherwise, take 0.

[0133] Step 6.4: Sort all candidate taxi supply-demand imbalance road segments in ascending order of significance, obtaining p-value(Z1)≤p-value(Z2)≤……≤p-value(Z2). m )≤……≤p-value(Z M M represents the number of candidate road sections with taxi supply and demand imbalance, Z M Let m represent the Mth candidate road segment with a supply-demand imbalance for taxis; m = 1, 2, ..., M, where M is the number of candidate road segments with a supply-demand imbalance for taxis.

[0134] Step 6.5, if the significance p-value (Zm) of the m-th candidate taxi supply-demand imbalance road segment Zm is... m Less than or equal to the significance level α adj If the m-th candidate taxi supply and demand imbalance segment Zm passes the significance test, then the m-th candidate taxi supply and demand imbalance segment Zm is considered as the taxi supply and demand imbalance segment (i.e., the final taxi supply and demand imbalance segment).

[0135] Where, α adj It is obtained by correcting the significance level α using the false discovery rate control method. α is the preset significance level, which can generally be set to 0.05.

[0136] Specifically, α adj =p-value(Z) q ), q is the condition that is satisfied The maximum subscript number of the significance level at time q (i.e., q is a value from 1 to M), and α is the preset significance level.

[0137] In summary, because this application considers the spatial correlation between taxis and the road network when detecting road sections with taxi supply and demand imbalances, it can accurately locate road sections with such imbalances, greatly improving the accuracy and interpretability of the detection and enhancing the practicality of this application. Furthermore, because this application uses a non-human-determined threshold method to determine the degree of taxi imbalance, it reduces the subjectivity of the evaluation results, thereby enhancing the reliability of this application.

[0138] The following example illustrates the taxi supply and demand imbalance detection method for road sections in this application.

[0139] In this example, the GPS location data of taxis in Xiamen Island area of ​​Xiamen City at 18:00 on May 31, 2019 is used as an example for illustration. In this example, the implementation steps of this application to assist in solving urban traffic management related problems are as follows:

[0140] (1) Identify the taxi supply and demand entries: The original GPS data entries are classified according to the license plate number, and sequentially sorted according to time to form a plurality of GPS data entry sequences. The data entries with the state of carrying passengers and the last time empty are identified as demand point entries, and the data entries with the state of empty and the next time empty are identified as supply point entries. The data entries with time field of 8 and 18 are screened out and projected to the plane space to obtain the spatial distribution diagram of taxi supply and demand points as shown in FIG. 2. Figure 2 Figure 2 FIG. 2 is a spatial distribution diagram of taxi supply and demand points, wherein (a) is a spatial distribution diagram of demand points at 18 o'clock, and (b) is a spatial distribution diagram of supply points at 18 o'clock.

[0141] (2) Match the taxi supply and demand points to the road network to generate the corrected spatial positions of the taxi supply and demand points. The buffer radius r is set to 50 m, and μ and σ are set to 0 and 20, respectively, during the road network matching. c c

[0142] (3) Divide the urban road network into road segments with more uniform lengths as the basic detection unit for detecting the range of the unbalanced road of taxi supply and demand. For a road with a length greater than the average value , the road is broken into N equal parts, and each part is a road segment. N is the minimum integer value that satisfies the condition that the length of each road segment after division is not greater than the average value .

[0143] (4) Detect the candidate range of unbalanced road of taxi supply and demand. Each road segment with a demand point is taken as a seed road segment, and each seed road segment is expanded along the road network through iterative optimization to form a candidate range of unbalanced road of taxi supply and demand with locally optimal abnormal degree. The distance threshold D of network parity is 500.

[0144] (5) Perform significance test on each candidate range of unbalanced road of taxi supply and demand by using the Monte Carlo simulation method. Compare the abnormal degree of the candidate range of unbalanced road of taxi supply and demand with the abnormal degree set calculated from the simulation data set, and identify the region that passes the significance test as the range of unbalanced road network of taxi supply and demand. The significance level a is 0.05. The final identification results of the supply and demand imbalance road segments at two time points are shown in FIG. 3. Figure 3

[0145] The above describes the preferred embodiments of the present application. It should be noted that those skilled in the art can make some improvements and refinements without departing from the principles of the present application, and these improvements and refinements should also be considered as the protection scope of the present application.​​​​

Claims

1. A method for detecting a taxi supply and demand imbalance section, characterized in that, The method comprises the following steps: Step 1, extracting all GPS data entries corresponding to each license plate number from a taxi GPS original data set, and sorting all GPS data entries corresponding to each license plate number in chronological order to obtain a GPS data entry sequence of the license plate number for each license plate number; the GPS original data set comprises GPS data entries of all taxis detected in the city; Step 2, determining a taxi supply point and a taxi demand point according to all GPS data entry sequences; Step 3, matching the taxi supply point and the taxi demand point to a city road network of the detected city, correcting the positions of the taxi supply point and the demand point in the city road network to determine a taxi supply correction point of the taxi supply point and a taxi demand correction point of the taxi demand point; Step 4, dividing at least part of the roads in the city road network based on the lengths of all roads between different intersections in the city road network to obtain a final city road network; Step 5, taking a road segment with a taxi demand correction point in the final city road network as a seed road segment, and expanding each seed road segment along the final city road network based on a taxi supply correction point by an iterative optimization method to obtain a plurality of candidate taxi supply-demand imbalance road segments; The method of expanding each seed road segment along the final city road network based on a taxi supply correction point by an iterative optimization method to obtain a plurality of candidate taxi supply-demand imbalance road segments comprises: For each seed road segment, the following steps are performed: Step 5.1, for each taxi demand correction point on the seed road segment, the number of taxi supply correction points with a shortest path distance less than a threshold D from the taxi demand correction point is taken as the network parity strength of the taxi demand correction point; Step 5.2, calculating the log-likelihood ratio of the seed road segment according to the network parity strength, and calculating the abnormality degree of the seed road segment according to the log-likelihood ratio; the abnormality degree is used to describe the taxi supply-demand imbalance degree of the seed road segment; Step 5.3, combining the seed road segment with each road segment connected to the seed road segment to obtain a plurality of new road segments, and calculating the abnormality degree of each new road segment; Step 5.4, taking the new road segment with the maximum abnormality degree among the abnormality degrees of all new road segments as the seed road segment; Step 5.5, judging whether the abnormality degree of the seed road segment in step 5.4 is greater than the abnormality degree of the seed road segment in step 5.2, if greater, taking the seed road segment in step 5.4 as the seed road segment in step 5.3, and returning to execute step 5.3, if the abnormality degree of the seed road segment in step 5.4 is equal to the abnormality degree of the seed road segment in step 5.2, then taking the seed road segment in step 5.4 as a candidate taxi supply-demand imbalance road segment; Step 6, performing significance test on each candidate taxi supply-demand imbalance road segment by a Monte Carlo simulation method, and taking the candidate taxi supply-demand imbalance road segment passing the significance test as a taxi supply-demand imbalance road segment.

2. The method of claim 1, wherein, The step 2 comprises: For each GPS data entry sequence, the following steps are performed: Step 2.1, for a target GPS data item in the sequence of GPS data items whose state is empty vehicle, if the state of the GPS data item at the next time point of the target GPS data item in the sequence of GPS data items is empty vehicle, the target GPS data item is taken as a taxi supply point GPS data item; Step 2.2, for a target GPS data item in the sequence of GPS data items whose state is passenger, if the state of the GPS data item at the previous time point of the target GPS data item in the sequence of GPS data items is empty vehicle, the target GPS data item is taken as a taxi demand point GPS data item; Step 2.3, projecting the longitude and latitude coordinate field in the taxi supply point GPS data item into two-dimensional space to form a taxi supply point; Step 2.4, projecting the longitude and latitude coordinate field in the taxi demand point GPS data item into two-dimensional space to form a taxi demand point.

3. The method of claim 1, wherein, The step 3 comprises: Taking the determined taxi supply point and taxi demand point as a target point, and respectively for each target point, performing the following steps: Step 3.1, determining the GPS data item corresponding to the target point from the sequence of GPS data items to which the target point belongs, and determining the GPS data item at the previous time point and the GPS data item at the next time point of the GPS data item in the sequence of GPS data items; Step 3.2, projecting the longitude and latitude coordinate field in the determined GPS data item at the previous time point into two-dimensional space to form a previous time point position point; Step 3.3, projecting the longitude and latitude coordinate field in the determined GPS data item at the next time point into two-dimensional space to form a next time point position point; Step 3.4, taking the road segment in the urban road network to the target point whose shortest straight line distance is less than r as a first candidate road segment, and taking the projection point of the target point on each first candidate road segment as a candidate correction point of the target point, and taking the set of all candidate correction points of the target point as a first candidate correction point set; Step 3.5, taking the road segment in the urban road network to the previous time point position point whose shortest straight line distance is less than r as a second candidate road segment, and taking the projection point of the previous time point position point on each second candidate road segment as a candidate correction point of the previous time point position point, and taking the set of all candidate correction points of the previous time point position point as a second candidate correction point set; Step 3.6, taking the road segment in the urban road network to the next time point position point whose shortest straight line distance is less than r as a third candidate road segment, and taking the projection point of the next time point position point on each third candidate road segment as a candidate correction point of the next time point position point, and taking the set of all candidate correction points of the next time point position point as a third candidate correction point set; Step 3.7, combining the candidate correction points in the first candidate correction point set, the second candidate correction point set and the third candidate correction point set to obtain a plurality of correction point combinations; each of the correction point combinations contains one candidate correction point in the first candidate correction point set, one candidate correction point in the second candidate correction point set and one candidate correction point in the third candidate correction point set; Step 3.8, calculating the matching degree of each of the correction point combinations, and taking the candidate correction point belonging to the first candidate correction point set in the correction point combination with the maximum matching degree as the correction point of the target point; wherein the correction point corresponding to the taxi supply point is a taxi supply correction point, and the correction point corresponding to the taxi demand point is a taxi demand correction point.

4. The method of claim 3, wherein, The calculation of the matching degree of each of the correction point combinations comprises: The matching degree F(L j ) of the correction point combination is calculated by the formula F(L ca ) = p(C j )*T(C 1n →C j )+p(C 2m )*T(C j →C 2m ). wherein p(C j ) denotes the probability of a candidate correction point C j being selected as a correction point, σ c denotes the standard deviation of a normal distribution, dist(P,C j ) denotes the straight-line distance between the target point P and a candidate correction point C j , μ c denotes the mean value of a normal distribution, T(C 1n → C j ) denotes the transition probability between a candidate correction point C j belonging to the first candidate correction point set and a candidate correction point C 1n belonging to the second candidate correction point set, dist(P1,P) denotes the straight-line distance between the previous-time position point P1 and the target point P, net_dist(C 1n → C j ) denotes the shortest path distance along the urban road network between a candidate correction point C 1n and a candidate correction point C j , p(C 2m ) denotes the probability of a candidate correction point C 2m being selected as a correction point, dist(P,C 2m ) denotes the straight-line distance between the target point P and a candidate correction point C 2m , T(C j → C 2m ) denotes the transition probability between a candidate correction point C j belonging to the first candidate correction point set and a candidate correction point C 2m belonging to the third candidate correction point set, dist(P,P2) denotes the straight-line distance between the target point P and the next-time position point P2, net_dist(C j → C 2m ) denotes the shortest path distance along the urban road network between a candidate correction point C j and a candidate correction point C 2m .

5. The method of claim 1, wherein, The step 4 comprises: Step 4.1, calculating the lengths of all roads between different intersections in the urban road network, and calculating the average value of the lengths of all roads; Step 4.2, dividing the road segments with lengths greater than the average value in the urban road network into N equal parts, and taking each equal part as a road segment to obtain a final urban road network; the length of each equal part is less than or equal to the average value.

6. The method of claim 1, wherein, The step 5.2 comprises: The log-likelihood ratio log of the seed road segment z is calculated by the formula * (z); N represents the number of taxi demand calibration points on the final urban road network, μ and σ respectively represent the average value and standard deviation of the network intensity of all taxi demand calibration points on the final urban road network, x i represents the network intensity of the i-th taxi demand calibration point, is the variance of the network intensity of each taxi supply calibration point within the seed road segment z, x z represents the sum of the network intensity of the taxi demand calibration points within the seed road segment z, μ z and λ z respectively represent the average value of the network intensity of the taxi demand calibration points within and outside the seed road segment z, n z represents the number of taxi demand calibration points within the seed road segment z, and X represents the sum of the network intensity of all taxi demand calibration points. If the over-supplied road segments are to be detected, the log-likelihood ratio log * (z) of the seed road segment z is taken as the abnormality degree of the seed road segment z; the indicator function I(μ z >λ z ) takes the value 1 when μ z >λ z , and otherwise takes the value 0; and the product of z and the indicator function I(μ z >λ z ) is taken as the abnormality degree of the seed road segment z. If the under-supplied road segments are to be detected, the log-likelihood ratio log * (z) of the seed road segment z is taken as the abnormality degree of the seed road segment z; the indicator function I(μ z <λ z ) takes the value 1 when μ z <λ z <λ z <λ z , otherwise takes the value 0.

7. The method of claim 1, wherein, The detection method further comprises: If there are overlapping road segments in the plurality of candidate taxi supply and demand imbalance road segments, only the candidate taxi supply and demand imbalance road segment with the maximum abnormality degree is retained.

8. The method of claim 1, wherein, The step 6 comprises: Respectively for each seed road segment, steps 6.1 and 6.2 are performed: Step 6.1, randomly reassigning a new network parity strength to each taxi demand correction point on the seed road segment, and repeating the process K times to obtain K groups of new network parity strengths; Step 6.2, respectively for each group of new network parity strengths, taking the group of new network parity strengths as the network parity strength in step 5.2 to perform steps 5.2 to 5.5, and retaining the maximum abnormality degree; Step 6.3, calculate the significance p-value (Z m ) of the mth candidate taxi supply-demand imbalance link Zm by the following formula: wherein, represents the maximum abnormality degree corresponding to the kth group of new network's network strength, AL obs represents the abnormality degree of the mth candidate taxi supply and demand imbalance section, I represents an indicator function, I takes 1 when 0; otherwise, 0. Step 6.4: Sort all candidate taxi supply-demand imbalance road segments in ascending order of significance, obtaining p-value(Z1)≤p-value(Z2)≤……≤p-value(Z2). m )≤……≤p-value(Z M M represents the number of candidate road sections with taxi supply and demand imbalance, Z M This indicates the Mth candidate road segment where taxi supply and demand are imbalanced; Step 6.5, if the significance p-value (Z m ) of the mth candidate taxi supply-demand imbalance road section Zm is less than or equal to the significance level a adj , the mth candidate taxi supply-demand imbalance road section Zm is taken as a taxi supply-demand imbalance road section. wherein a adj = p-value(Z q ), q is the largest subscript value satisfying the condition significance level.

Images