A rail transit data processing method, device, equipment and storage medium

Abstract

The embodiments of the present invention provide a rail transit data processing method, device, equipment and storage medium for correcting the passenger flow clearing model and obtaining accurate passenger flow clearing data, thereby reflecting the actual passenger flow situation. The method includes: selecting at least one target site, and determining multiple operating intervals corresponding to the target site, wherein the starting station of the operating interval is the target site, and the terminal station is any station on a single path of the target site without transfer; for any operating interval, traversing the rail transit data of the operating interval in a preset time period, and obtaining a set of non-riding time of passengers in the operating interval; clustering the non-riding time set by unsupervised learning corresponding to the operating interval to obtain the clustering results of the operating interval, and identifying abnormal data in the rail transit data with reverse riding behavior at the target station according to the clustering results; based on the identification results of the abnormal data, clearing the passenger flow.

Description

Technical Field

[0001] The present invention relates to the field of rail transit, and in particular to a rail transit data processing method, device, equipment and storage medium. Background Art

[0002] With the development of society, rail transit has become the main mode of transportation for people in cities, and it has higher travel convenience than ground transportation.

[0003] In order to obtain accurate rail transit passenger flow distribution data, the existing passenger flow classification models, when predicting passenger travel paths, are affected by the full load rate and infer the number of trains that passengers will take based on the proportion of trains they take. During the peak period of rail operation, there is a serious delay in the regular riding direction, that is, passengers cannot take the first train they are waiting for at the platform, so there is a phenomenon of passengers riding in the opposite direction, which leads to inaccurate estimation of passenger paths in passenger flow classification.

[0004] In summary, the passenger flow classification results in the prior art have errors and cannot reflect the actual passenger flow situation. Summary of the invention

[0005] The embodiments of the present invention provide a rail transit data processing method, device, equipment and storage medium for obtaining accurate passenger flow classification results, thereby reflecting the actual passenger flow situation.

[0006] In a first aspect, an embodiment of the present invention provides a rail transit data processing method, comprising:

[0007] Select at least one target station, and determine multiple operating intervals corresponding to the target station, wherein the starting station of the operating interval is the target station, and the terminal station is any station on a single path of the target station without transfer;

[0008] For any operating interval, traverse the rail transit data of the operating interval in the preset time period to obtain the set of non-riding time of passengers in the operating interval;

[0009] The non-riding time set is clustered by unsupervised learning corresponding to the operating interval to obtain the clustering result of the operating interval. The abnormal data with reverse riding behavior at the target station in the rail transit data is identified based on the clustering result.

[0010] Based on the identification results of abnormal data, the passenger flow is sorted.

[0011] The rail transit data processing method provided by the embodiment of the present invention first selects at least one target station and determines multiple operating intervals corresponding to the target station, wherein the starting station of the operating interval is the target station and the terminal station is any station on a single path of the target station without transfer, and then for any operating interval, traverses the rail transit data of the operating interval in a preset time period to obtain a set of non-riding time of passengers in the operating interval, clusters the non-riding time set by an unsupervised learning method corresponding to the operating interval, and obtains a clustering result of the operating interval, identifies abnormal data in the rail transit data with reverse riding behavior at the target station according to the clustering result, and finally sorts the passenger flow based on the identification result of the abnormal data. In an embodiment of the present invention, a set of non-riding time of passengers is obtained based on actual rail transit data, and a mixed Gaussian clustering method is used to cluster the above set of data. According to the clustering results, it is analyzed whether the passengers have reverse riding behavior at the target station, and passenger flow is sorted based on the abnormal data corresponding to the identified reverse riding behavior. Compared with the prior art, the abnormal data with reverse riding behavior and the normal data without reverse riding behavior are respectively sorted by corresponding passenger flow sorting models to obtain accurate passenger flow sorting results, which reflects the actual passenger flow situation.

[0012] In a possible implementation, selecting at least one target site includes:

[0013] For any driving direction of the rail transit vehicle, determine and obtain each target station interval in the driving direction within a preset time period, wherein the target station interval is a station interval with a full load rate greater than a preset full load rate threshold among all station intervals in the driving direction;

[0014] The station that the rail transit vehicle passes first in the target station section is taken as the target station.

[0015] In a possible implementation, traversing rail transit data of an operating section in a preset time period to obtain a set of non-riding time of passengers in the operating section includes:

[0016] Traverse the rail transit data of multiple sub-intervals within the operating interval, the starting station of the sub-interval is the target station, and the terminal station is any station in the operating interval except the target station;

[0017] Each time a piece of rail transit data is traversed, the following process is performed:

[0018] Determine the travel time of passengers in the sub-section according to rail transit data, where the travel time is the time between the time when the passengers enter the station and the time when they leave the station;

[0019] The difference between the passenger's travel time in the sub-interval and the shortest track running time in the sub-interval is taken as the passenger's non-riding time in the running interval.

[0020] In a possible implementation, data clustering is performed on the non-riding time set by unsupervised learning corresponding to the operating interval to obtain a clustering result of the operating interval, including:

[0021] Mixed Gaussian clustering is performed on the non-riding time set to determine the clustering results containing K-type arrays, where the clustering results of the target array represent the proportion of passengers taking the target train in the operating range, the average waiting time of passengers waiting for the target train at the target station, and the standard deviation of the waiting time. The target array is any array in the K-type arrays.

[0022] In a possible implementation, identifying abnormal data in rail transit data with reverse riding behavior at a target station according to the clustering result includes:

[0023] According to the mean difference between the last two types of arrays in the K types of arrays, the abnormal data with reverse riding behavior at the target station in the rail transit data is identified.

[0024] In a possible implementation, mixed Gaussian clustering is performed on the non-travel time set to determine a clustering result containing K types of arrays, including:

[0025] Use K-means algorithm to cluster the non-travel time set and determine the initial clustering result containing K types of initial arrays;

[0026] The initial clustering results are input into the Gaussian mixture model, and the maximum expectation algorithm is used to iterate the initial clustering results with a preset number of times and preset accuracy to converge, and determine the clustering results containing K types of arrays.

[0027] In a possible implementation, according to the mean difference between the last two types of arrays in the K types of arrays, identifying abnormal data with reverse riding behavior at the target station in the rail transit data includes:

[0028] Determine whether the mean difference of the last two types of arrays in the K types of arrays is greater than the standard deviation. If so, it is determined that there is abnormal data in the rail transit data that there is reverse riding behavior at the target station, where the standard deviation is the sum of the departure interval and the stop time of the rail transit vehicle at the target station;

[0029] The last type of data in the K-type array is identified as abnormal data in the rail transit data with reverse riding behavior at the target station.

[0030] In a possible implementation, based on the recognition result of abnormal data, the passenger flow is sorted, including:

[0031] Extracting abnormal data from rail transit data, and inputting the abnormal data into a first passenger flow sorting model for passenger flow sorting;

[0032] The rail transit data after eliminating abnormal data is input into the second passenger flow sorting model for passenger flow sorting.

[0033] In a possible implementation, the method further includes:

[0034] In the rail transit data, the abnormal data identified as reverse riding behavior at the target station is extracted to form an abnormal data set;

[0035] The abnormal data set is clustered by mixed Gaussian clustering, and the turning point for passengers to take the train in the opposite direction from the target station is determined by combining the shortest track running time between the target station and its reverse station.

[0036] In a possible implementation, before performing mixed Gaussian clustering on the non-travel time set and determining a clustering result including K-type arrays, the method further includes:

[0037] Using different preset cluster numbers, mixed Gaussian clustering is performed on the non-travel time set to determine multiple clustering results under different cluster numbers;

[0038] For multiple clustering results under different numbers of clusters, determine the mean difference of adjacent data in each clustering result except the last class, and determine the cluster number corresponding to the data with the mean difference closest to the standard waiting time as the cluster number K corresponding to the operating interval. The standard waiting time is the sum of the departure interval and half of the stop time of rail transit vehicles at the target station.

[0039] In a second aspect, an embodiment of the present invention provides a rail transit data processing device, including:

[0040] A selection unit, configured to select at least one target station and determine a plurality of operating intervals corresponding to the target station, wherein the starting station of the operating interval is the target station and the terminal station is any station on a single path to the target station without transfer;

[0041] An acquisition unit is used to traverse the rail transit data of any operating interval in a preset period of time and acquire a set of non-riding time of passengers in the operating interval;

[0042] A processing unit, used for clustering the non-riding time set by unsupervised learning corresponding to the operating interval, obtaining a clustering result of the operating interval, and identifying abnormal data of reverse riding behavior at a target station in the rail transit data according to the clustering result;

[0043] The execution unit is used to sort the passenger flow based on the recognition result of abnormal data.

[0044] In a possible implementation manner, the selection unit is specifically configured to:

[0045] For any driving direction of the rail transit vehicle, determine and obtain each target station interval in the driving direction within a preset time period, wherein the target station interval is a station interval with a full load rate greater than a preset full load rate threshold among all station intervals in the driving direction;

[0046] The station that the rail transit vehicle passes first in the target station section is taken as the target station.

[0047] In a possible implementation manner, the acquiring unit is specifically configured to:

[0048] Traverse the rail transit data of multiple sub-intervals within the operating interval, the starting station of the sub-interval is the target station, and the terminal station is any station in the operating interval except the target station;

[0049] Each time a piece of rail transit data is traversed, the following process is performed:

[0050] Determine the travel time of passengers in the sub-section according to rail transit data, where the travel time is the time between the time when the passengers enter the station and the time when they leave the station;

[0051] The difference between the passenger's travel time in the sub-interval and the shortest track running time in the sub-interval is taken as the passenger's non-riding time in the running interval.

[0052] In a possible implementation manner, the processing unit is specifically configured to:

[0053] Mixed Gaussian clustering is performed on the non-riding time set to determine the clustering results containing K-type arrays, where the clustering results of the target array represent the proportion of passengers taking the target train in the operating range, the average waiting time of passengers waiting for the target train at the target station, and the standard deviation of the waiting time. The target array is any array in the K-type arrays.

[0054] In a possible implementation manner, the processing unit is specifically configured to:

[0055] According to the mean difference between the last two types of arrays in the K types of arrays, the abnormal data with reverse riding behavior at the target station in the rail transit data is identified.

[0056] In a possible implementation manner, the processing unit is specifically configured to:

[0057] Use K-means algorithm to cluster the non-travel time set and determine the initial clustering result containing K types of initial arrays;

[0058] The initial clustering results are input into the Gaussian mixture model, and the maximum expectation algorithm is used to iterate the initial clustering results with a preset number of times and preset accuracy to converge, and determine the clustering results containing K types of arrays.

[0059] In a possible implementation manner, the processing unit is specifically configured to:

[0060] Determine whether the mean difference of the last two types of arrays in the K types of arrays is greater than the standard deviation. If so, it is determined that there is abnormal data in the rail transit data that there is reverse riding behavior at the target station, where the standard deviation is the sum of the departure interval and the stop time of the rail transit vehicle at the target station;

[0061] The last type of data in the K-type array is identified as abnormal data in the rail transit data with reverse riding behavior at the target station.

[0062] In a possible implementation manner, the execution unit is specifically configured to:

[0063] Extracting abnormal data from rail transit data, and inputting the abnormal data into a first passenger flow sorting model for passenger flow sorting;

[0064] The rail transit data after eliminating abnormal data is input into the second passenger flow sorting model for passenger flow sorting.

[0065] In a possible implementation manner, the processing unit is further configured to:

[0066] In the rail transit data, the abnormal data identified as reverse riding behavior at the target station is extracted to form an abnormal data set;

[0067] The abnormal data set is clustered by mixed Gaussian clustering, and the turning point for passengers to take the train in the opposite direction from the target station is determined by combining the shortest track running time between the target station and its reverse station.

[0068] In a possible implementation, before performing mixed Gaussian clustering on the non-travel time set and determining a clustering result including K-type arrays, the processing unit is further configured to:

[0069] Using different preset cluster numbers, mixed Gaussian clustering is performed on the non-travel time set to determine multiple clustering results under different cluster numbers;

[0070] For multiple clustering results under different numbers of clusters, determine the mean difference of adjacent data in each clustering result except the last class, and determine the cluster number corresponding to the data with the mean difference closest to the standard waiting time as the cluster number K corresponding to the operating interval. The standard waiting time is the sum of the departure interval and half of the stop time of rail transit vehicles at the target station.

[0071] In a third aspect, an embodiment of the present invention further provides a rail transit data processing device, including:

[0072] At least one processor, at least one memory, and computer program instructions stored in the memory. When the computer program instructions are executed by the processor, the rail transit data processing method provided in the first aspect of the embodiment of the present invention is implemented.

[0073] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having computer program instructions stored thereon. When the computer program instructions are executed by a processor, a rail transit data processing device is enabled to execute the rail transit data processing method provided in the first aspect of the embodiment of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood by reading the detailed description below with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, in which:

[0075] Figure 1 A schematic diagram of a rail transit route provided by an embodiment of the present invention;

[0076] Figure 2 Another rail transit route schematic diagram provided by an embodiment of the present invention;

[0077] Figure 3 A frequency distribution diagram of passenger travel time corresponding to AFC data during off-peak hours provided by an embodiment of the present invention;

[0078] Figure 4 A frequency distribution diagram of passenger travel time corresponding to AFC data during the morning peak period provided by an embodiment of the present invention;

[0079] Figure 5 A schematic flow chart of a rail transit data processing method provided by an embodiment of the present invention;

[0080] Figure 6 A schematic diagram of the structure of a rail transit data processing device provided by an embodiment of the present invention;

[0081] Figure 7 A schematic diagram of the structure of a rail transit data processing device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0082] The embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0083] In view of the fact that in the prior art, when obtaining rail transit passenger flow distribution data, there is a serious phenomenon of passengers staying in the regular riding direction and passengers riding in the opposite direction, which leads to inaccurate estimation of passenger paths in passenger flow sorting, an embodiment of the present invention provides a rail transit data processing solution for obtaining accurate passenger flow sorting data, thereby reflecting the actual passenger flow situation.

[0084] The solution provided by the embodiment of the present invention is described in detail below.

[0085] S1, during the morning rush hour, for any travel direction of the rail transit vehicle, search for each target station interval in the travel direction, wherein the target station interval is a station interval with a full load rate greater than a preset full load rate threshold in the travel direction, and take each station passed before the target station interval in the travel direction as a target station.

[0086] like Figure 1 FIG. 1 is a schematic diagram of a rail transit route provided by an embodiment of the present invention. Figure 1 The figure shows AG with a total of 7 stations, and its driving direction is A→G. In a possible implementation, the preset full load rate threshold is 120%, the full load rate of the AC station interval is 80%, the full load rate of the CD station interval is 130%, and the full load rate of the DG station interval is 110%. The CD station interval is the target station interval, and the upstream station in the CD station interval in the driving direction of A→G is station C, so station C will be used as the target station; or, the full load rate of the AB station interval is 90%, the full load rate of the BD station interval is 125%, the full load rate of the DE station interval is 100%, and the full load rate of the EG station interval is 130%. Then the BD and EG station intervals are both target station intervals, and the upstream station in the BD station interval in the driving direction of A→G is station B, and the upstream station in the EG station interval is station E, so stations B and E will be used as target stations.

[0087] It should be noted that, among them, full load factor = passenger turnover (person-kilometers) / passenger seat (capacity) kilometers;

[0088] In the formula: passenger seat kilometers is the product of the number of operating vehicle capacity and vehicle kilometers; the number of vehicle capacity is the sum of the number of fixed seats set in the car and the standing capacity quota per square meter of effective area. It reflects the relative value of the passenger load of vehicles running on the line within a certain period of time and is an indicator of vehicle utilization. Once the load rate is too high, the rail transit operator will take dynamic flow control measures to strictly control the density of people in the station and in the car.

[0089] Furthermore, the morning peak period may be set according to an empirical value, such as 7-9 am, or 8-9 am, and the present invention is not limited thereto.

[0090] S2, using the effective path table of transportation travel, find the single path with the target station as the starting station without transfer OD (O stands for Origin, indicating the starting point of the trip, D stands for Destination, indicating the destination of the trip, so OD represents the starting and ending station of transportation), and at the same time find the morning rush hour AFC (Automatic Fare Collection, urban rail transit automatic ticket sales and inspection) data of the above OD within the preset time period.

[0091] It should be noted that the effective transportation path table is established by the rail transit operator, which includes the effective stations in the rail transit path and the travel path planning of each train in the rail operation based on each effective station.

[0092] Furthermore, if Figure 2 Another rail transit route diagram is shown. For example, if Xiyuan Station on Line 4 is the target station, and Anheqiao North → Tiangongyuan is the driving direction, the single-path non-transfer OD of Xiyuan Station includes: Xiyuan-Yuanmingyuan, Xiyuan-East Gate of Peking University, Xiyuan-Zhongguancun, Xiyuan-Haidian Huangzhuang, Xiyuan-Renmin University, etc. Moreover, although Xiyuan-Beigongmen also belongs to the single-path non-transfer OD, its driving direction is Tiangongyuan-Anheqiao North, so it cannot meet the requirements.

[0093] Furthermore, AFC data is obtained through the AFC system, which is an automated system based on computer, communication, network, automatic control and other technologies to realize the entire process of rail transit ticketing, ticket checking, billing, charging, statistics, clearing and management. Therefore, based on the AFC data, the entry and exit stations and entry and exit times of passengers when taking rail transit vehicles can be obtained.

[0094] Furthermore, the AFC data for the above-mentioned OD can be the morning peak period data within one month, or the morning peak period data within two months, and the data generation time is not limited, for example, the morning peak period data from January 1 to January 31, 2020, or the morning peak period data from April 1 to May 30, 2020.

[0095] S3, traverse the above AFC data, determine the entry and exit stations and entry and exit times of passengers when taking rail transit, obtain the passenger's travel time based on the entry and exit times, and subtract the shortest rail running time of the corresponding OD from the passenger's travel time. Since the above ODs are all single-path ODs without transfers, the obtained data set can be considered as the passenger's non-travel time set X = {x1, x2, ..., x n}, and the number of clusters is recorded as K.

[0096] It should be noted that if Figure 3 As shown in the figure, the AFC data of all passengers with OD of Shahe-Xi'erqi Station in the off-peak period between 9:00 and 12:00 in September 2018 are read, the travel time corresponding to the AFC data is calculated and the frequency distribution diagram is drawn. It can be seen that the travel time of passengers conforms to the normal distribution.

[0097] The travel time of passengers from entering the station to exiting the station includes the following parts: entry time, waiting time, riding time, and exit time. The riding time is the shortest track operation time. Therefore, the passenger travel market minus the shortest track operation time corresponding to OD is the passenger entry time, waiting time and exit time.

[0098] Furthermore, the time it takes for passengers to enter and exit the station is related to their walking speed, which changes continuously. Therefore, the time it takes for passengers to enter and exit the station can be regarded as a continuously changing random variable. The waiting time is often related to the number of trains that passengers take. The waiting time of passengers on the same train can be regarded as a continuously changing random variable. Therefore, it can be assumed that the non-ride time of passengers follows a normal distribution to further analyze whether passengers have reverse ride behavior.

[0099] S4, clustering the set X using a K-means clustering method with a preset number of clusters to obtain an initial clustering result.

[0100] It should be noted that the preset numbers of multiple clusters can be set according to the experience of technical personnel in this field. The number of clusters can be a continuous number or a number with intervals, such as 1-6, 2-8, 3 and 5 and 7, 2 and 5 and 6-8, etc. The embodiment of the present invention does not limit the preset value of the number of clusters.

[0101] S5, performing mixed Gaussian clustering on the initial clustering results obtained above to obtain mixed Gaussian clustering results under various cluster numbers, wherein the mixed Gaussian clustering formula can be expressed as:

[0102]

[0103] In the formula, ω i , μ i , σ i are the weight, mean and standard deviation of each Gaussian sub-distribution, respectively, representing the proportion of passengers taking each bus, the average non-riding time and the standard deviation of the non-riding time.

[0104] In specific implementation, the EM (maximum expectation) algorithm is used to obtain the clustering results of mixed Gaussian under each number of clusters, which can have the following steps:

[0105] 1. E-Step: Calculate each data X in the X set iBelongs to the jth distribution Z j The probability calculation formula is:

[0106]

[0107] Among them, ω, μ, σ are the initial input values ​​of the mixed Gaussian model, that is, the initial clustering results obtained by the K-means clustering method in step S4.

[0108] It should be noted that since the EM algorithm is particularly sensitive to the choice of initial values ​​when solving the Gaussian mixture model, before solving it, the original data set is first clustered using the k-means algorithm, and the mean of each category, the standard deviation of each category of samples, and the percentage of each category of samples in the total samples are used as the above initial input values, which can reduce the number of iterative convergence times.

[0109] 2. M-Step: Construct the likelihood function, that is, calculate the logarithmic form of the probability expression of the above sample set under given ω, μ, σ, that is:

[0110]

[0111] 3. Using Jensen's inequality, we can get:

[0112]

[0113] 4. Iterate the ω, μ, σ in the right-hand side until they converge with the preset number of iterations and the preset convergence accuracy, and you can get the ω, μ, σ corresponding to the required likelihood function. Since the right-hand side of the above equation contains ω, μ, σ, the above equation can be quickly iterated by taking partial derivatives of each variable.

[0114] In specific implementation, the number of iterations and the convergence accuracy can be set according to empirical values. For example, the number of iterations can be set to 500 generations and the convergence accuracy can be set to 0.00001. The present invention does not limit this.

[0115] 5. Determine the ω, μ, σ that finally reach the convergence condition as the ω in the mixed Gaussian clustering formula i , μ i , σ i To solve the mixed Gaussian clustering formula and determine the mixed Gaussian clustering results.

[0116] S6. Based on the mixed Gaussian clustering results, calculate the average difference between the means of the first k-1 categories of data under different numbers of clusters, and take the number of clusters that is closest to T (departure interval + 1 / 2 stop time) as the optimal number of clusters.

[0117] It should be noted that Figure 3The figure shows the distribution of passengers' travel time during off-peak hours. Since there is no obvious retention phenomenon during off-peak hours, a normal distribution can be used to represent it. For the morning peak, the reason why there is reverse riding is that there is a serious retention in the regular riding direction, that is, passengers cannot take the first bus they are waiting for at the platform. Therefore, a mixture of multiple normal distributions is needed to represent this distribution, that is, a mixed Gaussian model.

[0118] like Figure 4 As shown in the figure, the AFC data of all OD Shahe-Xi'erqi stations with travel time between 7-9 o'clock (i.e., the morning peak period) in September 2018 are read, the travel time corresponding to the above AFC data is calculated and a frequency distribution diagram is drawn.

[0119] Furthermore, due to the "staying" phenomenon, a more ideal fitting effect can be obtained by adding multiple normal distributions (i.e., Gaussian distributions); at the same time, since different distributions represent different rides, the mean of each distribution is the mean of the travel time under the conditions of the train, and the interval between each mean is roughly T (departure interval + 1 / 2 stop time). Figure 4 The starting station in is Shahe Station. The difference in the means of the last two normal distributions is much greater than T, so the time in the last normal distribution can be identified as the travel time record of passengers riding in the opposite direction.

[0120] Furthermore, during the morning peak hours of rail operation, some lines with relatively crowded passenger flow (i.e., high load rate) have some stations where passengers cannot get on the first train when they arrive at the platform to wait for the train. Passengers need to wait for the next train to arrive before boarding the train. If they cannot board the next train, they need to continue waiting, and so on. According to the above situation, if a passenger cannot board the first train, if he takes the next train, then the longest waiting time is the train departure interval of the route + the train stop time at the station (i.e., boarding the train one second before departure), and the shortest waiting time is the train departure interval (i.e., boarding the train immediately after the vehicle arrives at the platform). Therefore, the T value is set to the average of the above two extreme cases, i.e., the departure interval of the starting station + 1 / 2 of the stop time.

[0121] S7, based on the clustering data results in the optimal number of clusters, determine whether the mean difference between the last category and the second to last category under the optimal number of clusters is greater than the sum of the departure interval and the stop time at the starting station. If so, it is determined that there is a reverse riding phenomenon at the target station.

[0122] It should be noted that when a passenger is waiting for the next train at the target station, the longest waiting time is the sum of the departure interval and the stop time on the route. Therefore, if the difference in the means between the last category and the second-to-last category is greater than the sum of the departure interval and the stop time at the starting station, it is determined that the passenger did not wait for the next train, but took the reverse train.

[0123] S8, the weight of the last category of data under the optimal number of clusters is used as the reverse ride ratio of the target station.

[0124] S9, based on the recognition results of abnormal data, the passenger flow is sorted.

[0125] In specific implementation, passenger flow sorting can be implemented through the following two steps:

[0126] 1. Extract abnormal data from rail transit data and input the abnormal data into the first passenger flow sorting model for passenger flow sorting to identify the turning point and specific travel path of passengers when they take the train in the opposite direction;

[0127] 2. The rail transit data after the abnormal data is removed is input into the second passenger flow sorting model for passenger flow sorting to identify the waiting hours and specific travel routes of passengers at the target station when they take the train normally.

[0128] It should be noted that the first passenger flow sorting model may be two passenger flow sorting models completely different from the second passenger flow sorting model, or may be a branch model of the second passenger flow sorting model, and the present invention does not limit this.

[0129] In a possible implementation, after the abnormal data set is extracted, mixed Gaussian clustering may be performed on it to obtain the selection ratio of each reverse ride turnaround station.

[0130] The rail transit data processing solution provided by the embodiment of the present invention is described in detail below with reference to the accompanying drawings, taking Shahe Station on the Changping Line in Beijing as an example.

[0131] S1, taking Shahe Station as the target station, find all suburban-urban single-path non-transfer ODs with Shahe Station as the starting station, and read the shortest track running time of each OD, as shown in Table 1:

[0132] Table 1

[0133]

[0134] S2, read the morning peak AFC data of the above OD within one month, use the box plot method to filter out abnormal data, and calculate the travel time of the above OD passengers based on the valid data. Then subtract the shortest track operation time of each OD from the travel time, and the resulting data set is the non-travel time collection of passengers at Shahe Station.

[0135] S3, use the k-means algorithm to cluster the above non-riding time set, take the cluster number as 1-6 for calculation, use the obtained results as the initial value for mixed Gaussian clustering, and use the EM (maximum expectation) algorithm to obtain the clustering results under each cluster number, as shown in Table 2:

[0136] Table 2

[0137]

[0138] S4, according to the actual operation diagram data, the train departure interval and stop time are calculated to be 267s and 71s respectively, then T = 267 + 71 / 2 = 302.5s, which is closest to the average difference of the first three categories when clustering into 4 categories. Therefore, the optimal number of clusters is 4.

[0139] When clustering into 4 categories, the mean difference between the 4th category and the 3rd category is 546s, which exceeds 267+71=338s. Therefore, it can be identified that there is a reverse riding phenomenon at Shahe Station.

[0140] Among them, the complete data clustering recognition results of Shahe Station are shown in Table 3. The proportion of passengers who take the reverse ride is 12.16%. The recognition results are shown in Table 3:

[0141] Table 3

[0142]

[0143] S5, extracting rail transit data after eliminating abnormal data under the optimal number of clusters, that is, the data of passengers taking the first, second and third trains, inputs it into the passenger flow sorting model to sort the passenger flow.

[0144] In a possible implementation, after determining the proportion of reverse travel, the proportion of passengers choosing different turnaround stations is analyzed. The travel time data set for reverse travel is extracted from the total travel time data set, and mixed Gaussian clustering is performed on it to obtain the selection ratio of each turnaround station. Among all reverse travel behaviors with Shahe Station as the target station, the proportion of choosing Shahe Higher Education Park Station as the turnaround station is 44%, Nanshao Station is 51%, and Beishaowa Station is 5%, as shown in Table 4:

[0145] Table 4

[0146]

[0147] like Figure 5 As shown, an embodiment of the present invention provides a rail transit data processing method, comprising:

[0148] Step 501, select at least one target station, and determine multiple operating intervals corresponding to the target station, wherein the starting station of the operating interval is the target station, and the terminal station is any station on a single path of the target station without transfer.

[0149] Step 502: for any operating section, traverse the rail transit data of the operating section in a preset time period to obtain a set of non-riding time of passengers in the operating section.

[0150] Step 503, clustering the non-riding time set by unsupervised learning corresponding to the operating interval to obtain the clustering result of the operating interval, and identifying abnormal data with reverse riding behavior at the target station in the rail transit data based on the clustering result.

[0151] Step 504: sort the passenger flow based on the recognition result of abnormal data.

[0152] In a possible implementation, selecting at least one target site includes:

[0153] For any driving direction of the rail transit vehicle, determine and obtain each target station interval in the driving direction within a preset time period, wherein the target station interval is a station interval with a full load rate greater than a preset full load rate threshold among all station intervals in the driving direction;

[0154] The station that the rail transit vehicle passes first in the target station section is taken as the target station.

[0155] In a possible implementation, traversing rail transit data of an operating section in a preset time period to obtain a set of non-riding time of passengers in the operating section includes:

[0156] Traverse the rail transit data of multiple sub-intervals within the operating interval, the starting station of the sub-interval is the target station, and the terminal station is any station in the operating interval except the target station;

[0157] Each time a piece of rail transit data is traversed, the following process is performed:

[0158] Determine the travel time of passengers in the sub-section according to rail transit data, where the travel time is the time between the time when the passengers enter the station and the time when they leave the station;

[0159] The difference between the passenger's travel time in the sub-interval and the shortest track running time in the sub-interval is taken as the passenger's non-riding time in the running interval.

[0160] In a possible implementation, data clustering is performed on the non-riding time set by unsupervised learning corresponding to the operating interval to obtain a clustering result of the operating interval, including:

[0161] Mixed Gaussian clustering is performed on the non-riding time set to determine the clustering results containing K-type arrays, where the clustering results of the target array represent the proportion of passengers taking the target train in the operating range, the average waiting time of passengers waiting for the target train at the target station, and the standard deviation of the waiting time. The target array is any array in the K-type arrays.

[0162] In a possible implementation, identifying abnormal data in rail transit data with reverse riding behavior at a target station according to the clustering result includes:

[0163] According to the mean difference between the last two types of arrays in the K types of arrays, the abnormal data with reverse riding behavior at the target station in the rail transit data is identified.

[0164] In a possible implementation, mixed Gaussian clustering is performed on the non-travel time set to determine a clustering result containing K types of arrays, including:

[0165] Use K-means algorithm to cluster the non-travel time set and determine the initial clustering result containing K types of initial arrays;

[0166] The initial clustering results are input into the Gaussian mixture model, and the maximum expectation algorithm is used to iterate the initial clustering results with a preset number of times and preset accuracy to converge, and determine the clustering results containing K types of arrays.

[0167] In a possible implementation, according to the mean difference between the last two types of arrays in the K types of arrays, identifying abnormal data with reverse riding behavior at the target station in the rail transit data includes:

[0168] Determine whether the mean difference of the last two types of arrays in the K types of arrays is greater than the standard deviation. If so, it is determined that there is abnormal data in the rail transit data that there is reverse riding behavior at the target station, where the standard deviation is the sum of the departure interval and the stop time of the rail transit vehicle at the target station;

[0169] The last type of data in the K-type array is identified as abnormal data in the rail transit data with reverse riding behavior at the target station.

[0170] In a possible implementation, based on the recognition result of abnormal data, the passenger flow is sorted, including:

[0171] Extracting abnormal data from rail transit data, and inputting the abnormal data into a first passenger flow sorting model for passenger flow sorting;

[0172] The rail transit data after eliminating abnormal data is input into the second passenger flow sorting model for passenger flow sorting.

[0173] In a possible implementation, the method further includes:

[0174] In the rail transit data, the abnormal data identified as reverse riding behavior at the target station is extracted to form an abnormal data set;

[0175] The abnormal data set is clustered by mixed Gaussian clustering, and the turning point for passengers to take the train in the opposite direction from the target station is determined by combining the shortest track running time between the target station and its reverse station.

[0176] In a possible implementation, before performing mixed Gaussian clustering on the non-travel time set and determining a clustering result including K-type arrays, the method further includes:

[0177] Using different preset cluster numbers, mixed Gaussian clustering is performed on the non-travel time set to determine multiple clustering results under different cluster numbers;

[0178] For multiple clustering results under different numbers of clusters, determine the mean difference of adjacent data in each clustering result except the last class, and determine the cluster number corresponding to the data with the mean difference closest to the standard waiting time as the cluster number K corresponding to the operating interval. The standard waiting time is the sum of the departure interval and half of the stop time of rail transit vehicles at the target station.

[0179] like Figure 6 As shown, an embodiment of the present invention provides a rail transit data processing device, comprising:

[0180] A selection unit 61 is used to select at least one target station and determine multiple operating intervals corresponding to the target station, wherein the starting station of the operating interval is the target station and the terminal station is any station on a single path of the target station without transfer;

[0181] An acquisition unit 62 is used to traverse the rail transit data of any operating interval in a preset period of time and acquire a set of non-riding time of passengers in the operating interval;

[0182] The processing unit 63 is used to perform data clustering on the non-riding time set by unsupervised learning corresponding to the operating interval, obtain the clustering result of the operating interval, and identify abnormal data with reverse riding behavior at the target station in the rail transit data according to the clustering result;

[0183] The execution unit 64 is used to sort the passenger flow based on the recognition result of the abnormal data.

[0184] In a possible implementation manner, the selection unit 61 is specifically configured to:

[0185] For any driving direction of the rail transit vehicle, determine and obtain each target station interval in the driving direction within a preset time period, wherein the target station interval is a station interval with a full load rate greater than a preset full load rate threshold among all station intervals in the driving direction;

[0186] The station that the rail transit vehicle passes first in the target station section is taken as the target station.

[0187] In a possible implementation manner, the acquisition unit 62 is specifically configured to:

[0188] Traverse the rail transit data of multiple sub-intervals within the operating interval, the starting station of the sub-interval is the target station, and the terminal station is any station in the operating interval except the target station;

[0189] Each time a piece of rail transit data is traversed, the following process is performed:

[0190] Determine the travel time of passengers in the sub-section according to rail transit data, where the travel time is the time between the time when the passengers enter the station and the time when they leave the station;

[0191] The difference between the passenger's travel time in the sub-interval and the shortest track running time in the sub-interval is taken as the passenger's non-riding time in the running interval.

[0192] In a possible implementation manner, the processing unit 63 is specifically configured to:

[0193] Mixed Gaussian clustering is performed on the non-riding time set to determine the clustering results containing K-type arrays, where the clustering results of the target array represent the proportion of passengers taking the target train in the operating range, the average waiting time of passengers waiting for the target train at the target station, and the standard deviation of the waiting time. The target array is any array in the K-type arrays.

[0194] In a possible implementation manner, the processing unit 63 is specifically configured to:

[0195] According to the mean difference between the last two types of arrays in the K types of arrays, the abnormal data with reverse riding behavior at the target station in the rail transit data is identified.

[0196] In a possible implementation manner, the processing unit 63 is specifically configured to:

[0197] Use K-means algorithm to cluster the non-travel time set and determine the initial clustering result containing K types of initial arrays;

[0198] The initial clustering results are input into the Gaussian mixture model, and the maximum expectation algorithm is used to iterate the initial clustering results with a preset number of times and preset accuracy to converge, and determine the clustering results containing K types of arrays.

[0199] In a possible implementation manner, the processing unit 63 is specifically configured to:

[0200] Determine whether the mean difference of the last two types of arrays in the K types of arrays is greater than the standard deviation. If so, it is determined that there is abnormal data in the rail transit data that there is reverse riding behavior at the target station, where the standard deviation is the sum of the departure interval and the stop time of the rail transit vehicle at the target station;

[0201] The last type of data in the K-type array is identified as abnormal data in the rail transit data with reverse riding behavior at the target station.

[0202] In a possible implementation, the execution unit 64 is specifically configured to:

[0203] Extracting abnormal data from rail transit data, and inputting the abnormal data into a first passenger flow sorting model for passenger flow sorting;

[0204] The rail transit data after eliminating abnormal data is input into the second passenger flow sorting model for passenger flow sorting.

[0205] In a possible implementation manner, the processing unit 63 is further configured to:

[0206] In the rail transit data, the abnormal data identified as reverse riding behavior at the target station is extracted to form an abnormal data set;

[0207] The abnormal data set is clustered by mixed Gaussian clustering, and the turning point for passengers to take the train in the opposite direction from the target station is determined by combining the shortest track running time between the target station and its reverse station.

[0208] In a possible implementation, before performing mixed Gaussian clustering on the non-travel duration set and determining a clustering result including K-type arrays, the processing unit 63 is further configured to:

[0209] Using different preset cluster numbers, mixed Gaussian clustering is performed on the non-travel time set to determine multiple clustering results under different cluster numbers;

[0210] For multiple clustering results under different numbers of clusters, determine the mean difference of adjacent data in each clustering result except the last class, and determine the cluster number corresponding to the data with the mean difference closest to the standard waiting time as the cluster number K corresponding to the operating interval. The standard waiting time is the sum of the departure interval and half of the stop time of rail transit vehicles at the target station.

[0211] like Figure 7 As shown, the embodiment of the present invention further provides a rail transit data processing device 70, including:

[0212] At least one processor 71, at least one memory 72, and computer program instructions stored in the memory, when the computer program instructions are executed by the processor 71, the rail transit data processing method provided in the embodiment of the present invention is implemented.

[0213] In an exemplary embodiment, a storage medium including instructions is also provided, such as a memory 72 including instructions. The instructions can be executed by a processor 71 of a rail transit data processing device to complete the above method.

[0214] Alternatively, the storage medium may be a non-transitory computer-readable storage medium, for example, the non-transitory computer-readable storage medium may be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.

[0215] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0216] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A rail transit data processing method, characterized in that: include: Select at least one target station, and determine a plurality of operating intervals corresponding to the target station, wherein the starting station of the operating interval is the target station, and the terminal station is any station on a single path of the target station without transfer; For any operating section, traverse the rail transit data of the operating section in a preset time period to obtain a set of non-riding time of passengers in the operating section; Perform mixed Gaussian clustering on the non-riding time set to determine a clustering result including K-type arrays, and identify abnormal data in the rail transit data that have reverse riding behavior at the target station based on the mean difference between the last two types of arrays in the K-type arrays; wherein the clustering result of the target array represents the proportion of passengers who take the target train in the operating interval, the average waiting time of passengers waiting for the target train at the target station, and the standard deviation of the waiting time, and the target array is any array in the K-type arrays; Based on the identification result of the abnormal data, the passenger flow is sorted.

2. The method according to claim 1, characterized in that The selecting at least one target site comprises: For any driving direction of the rail transit vehicle, determine and obtain each target station interval in the driving direction within the preset time period, wherein the target station interval is a station interval with a full load rate greater than a preset full load rate threshold among all station intervals in the driving direction; The station in the target station section that the rail transit vehicle passes first is taken as the target station.

3. The method according to claim 1, characterized in that The traversing of the rail transit data of the operating section in a preset time period to obtain a set of non-riding time of passengers in the operating section includes: Traversing rail transit data of multiple sub-intervals within the operating interval, the starting station of the sub-interval is the target station, and the terminal station is any station in the operating interval except the target station; Each time a piece of rail transit data is traversed, the following process is performed: Determine the travel time of the passenger in the sub-section according to the rail transit data, wherein the travel time is the time between the passenger's entry time and the passenger's exit time; The difference between the travel time of the passenger in the sub-section and the shortest track running time of the sub-section is taken as the non-riding time of the passenger in the running section.

4. The method according to claim 1, characterized in that The performing mixed Gaussian clustering on the non-riding time set to determine a clustering result including K types of arrays includes: Using a K-means algorithm to perform data clustering on the non-riding time set, and determining an initial clustering result including K types of initial arrays; The initial clustering result is input into a mixed Gaussian model, and the maximum expectation algorithm is used to iterate and converge the initial clustering result with a preset number of times and a preset accuracy to determine a clustering result containing K types of arrays.

5. The method according to claim 1, characterized in that The step of identifying abnormal data in the rail transit data having reverse riding behavior at the target station according to the mean difference between the last two types of arrays in the K types of arrays comprises: Determine whether the mean difference of the last two types of arrays in the K types of arrays is greater than the standard deviation. If so, determine that there is abnormal data in the rail transit data indicating reverse riding behavior at the target station, wherein the standard deviation is the sum of the departure interval and the stop time of the rail transit vehicle at the target station; The last type of data in the K-type array is identified as abnormal data in the rail transit data that has reverse riding behavior at the target station.

6. The method according to any one of claims 1 to 5, characterized in that: The step of classifying the passenger flow based on the recognition result of the abnormal data includes: Extracting the abnormal data from the rail transit data, and inputting the abnormal data into a first passenger flow sorting model for passenger flow sorting; The rail transit data after eliminating the abnormal data is input into the second passenger flow sorting model for passenger flow sorting.

7. The method according to any one of claims 1 to 5, characterized in that: The method further comprises: Extracting abnormal data identified as reverse riding behavior at the target station from the rail transit data to form an abnormal data set; The abnormal data set is subjected to mixed Gaussian clustering, and the turnaround station for passengers to take the train in the opposite direction from the target station is determined in combination with the shortest track running time between the target station and its reverse station.

8. The method according to claim 1, characterized in that Before performing mixed Gaussian clustering on the non-riding time set to determine a clustering result including K-type arrays, the method further includes: Performing mixed Gaussian clustering on the non-travel time set with different preset numbers of clusters to determine multiple clustering results with different numbers of clusters; For multiple clustering results under the different clustering numbers, determine the mean difference of adjacent data in the data of other categories of each clustering result except the last category, and determine the cluster number corresponding to the data whose mean difference is closest to the standard waiting time as the cluster number K corresponding to the operating interval. The standard waiting time is the sum of the departure interval and half of the stop time of the rail transit vehicle at the target station.

9. A rail transit data processing device, characterized in that: include: At least one processor, at least one memory and computer program instructions stored in the memory, when the computer program instructions are executed by the processor, implement the rail transit data processing method as described in any one of claims 1-8.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, a rail transit data processing method as described in any one of claims 1 to 8 is implemented.

Images