Evaluation method and system for the impact of bus network adjustment on passenger flow

Abstract

The embodiment of the present invention discloses a method and system for evaluating the impact of bus network adjustment on passenger flow, the method comprising: step S1: collecting network static data, line operation information, and passenger flow OD data of a preset historical period; step S2: establishing a machine learning model for the probability of passengers choosing different paths; step S3: obtaining the corresponding affected OD according to the adjustment information of the bus network; step S4: calculating the passenger flow distribution of each affected OD on each path under the adjusted bus network through the machine learning model; step S5: calculating the passenger flow of each line after the bus network adjustment, and then comparing it with the passenger flow before the bus network adjustment, to obtain an evaluation of the impact of the bus network adjustment on passenger flow. The present invention can significantly improve the accuracy of the evaluation of the impact of bus network adjustment on passenger flow, and the present invention is of great significance for guiding the adjustment of bus network.

Description

Technical Field

[0001] The present invention relates to the field of transportation, and in particular to a method and system for evaluating the impact of bus line network adjustment on passenger flow. Background Art

[0002] As urban construction continues, the bus network also needs to be adjusted in a timely manner to adapt to changes in the situation. The ways to adjust the network include adding lines, deleting lines, and adjusting lines. Common reasons for adjusting the bus network include: 1) Changes in crowd gathering places lead to major changes in bus travel demand, such as new residential communities, new commercial buildings, etc.; 2) Competition from other modes of transportation, such as the opening of new rail transit lines, an increase in car ownership, and emerging low-altitude transportation.

[0003] The main problems of bus network adjustment are: 1) how to adjust the lines to adapt to the changes in the situation; 2) how to evaluate the impact of the adjusted line network, especially the impact on passenger flow. The present invention is aimed at evaluating the impact of line network adjustment on passenger flow. The evaluation content includes: which lines will be affected by the line adjustment, whether the passenger flow will increase or decrease, and the specific amount.

[0004] At present, the evaluation of the impact of bus network adjustment on passenger flow mainly uses traffic models to simulate passenger flow distribution, and then analyzes passenger flow changes based on the simulation. The idea of ​​passenger flow distribution simulation can be summarized as follows:

[0005] 1) Pre-set a passenger flow distribution model with parameters such as the Logit model;

[0006] 2) Calibrate the parameters in the model through historical data;

[0007] 3) Carry out passenger flow distribution simulation on the adjusted bus network according to the calibrated parameters.

[0008] The current methods for evaluating the impact of bus network adjustments on passenger flow have the following shortcomings:

[0009] 1) The form of the passenger flow distribution model needs to be pre-set based on experience, such as Logit, and no matter which form the model takes, the degree of approximation to the real world is limited;

[0010] 2) Passenger travel has a certain degree of randomness, and the model calculation results cannot completely match the historical data. This makes it easy to produce evaluation results that are contrary to common sense when evaluating passenger flow impact: for example, after the route adjustment, the travel time becomes shorter, and other conditions remain unchanged, but the passenger flow share is reduced.

[0011] In recent years, the accuracy of bus OD estimation has been increasing, reaching about 90%. It is now possible to evaluate the impact of network adjustments on passenger flow by analyzing all ODs. Summary of the invention

[0012] The technical problem to be solved by the embodiments of the present invention is to provide a method and system for evaluating the impact of bus line network adjustment on passenger flow, so as to evaluate the impact of bus line network adjustment on passenger flow.

[0013] In order to solve the above technical problems, an embodiment of the present invention proposes a method for evaluating the impact of bus line network adjustment on passenger flow, including:

[0014] Step S1: Collecting network static data, line operation information, and passenger flow OD data for a preset historical period;

[0015] Step S2: Building a machine learning model for the probability of passengers choosing different routes;

[0016] Step S3: Obtain the corresponding affected OD according to the adjustment information of the bus network;

[0017] Step S4: Calculate the passenger flow distribution of each affected OD on each path under the adjusted bus network through a machine learning model;

[0018] Step S5: Calculate the passenger flow of each line after the bus network adjustment, and then compare it with the passenger flow before the bus network adjustment to obtain an evaluation of the impact of the bus network adjustment on the passenger flow.

[0019] Accordingly, an embodiment of the present invention further provides a system for evaluating the impact of bus line network adjustment on passenger flow, comprising:

[0020] Historical data preparation module: collects network static data, line operation information, and passenger flow OD data for a preset historical period;

[0021] Model building module: build a machine learning model of the probability of passengers choosing different routes;

[0022] Affected OD acquisition module: obtains the corresponding affected OD according to the adjustment information of the bus network;

[0023] Passenger flow distribution module: Calculate the passenger flow distribution of each affected OD on each path under the adjusted bus network through machine learning model;

[0024] Impact assessment module: calculates the passenger flow of each line after the bus network adjustment, and then compares it with the passenger flow before the bus network adjustment to obtain an assessment of the impact of the bus network adjustment on passenger flow.

[0025] The beneficial effects of the present invention are as follows: the present invention can significantly improve the accuracy of the evaluation of the impact of bus line network adjustment on passenger flow, and the present invention is of great significance in guiding the adjustment of bus line network. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1It is a flow chart of a method for evaluating the impact of bus line network adjustment on passenger flow according to an embodiment of the present invention.

[0027] Figure 2 It is a structural block diagram of a machine learning model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] It should be noted that, in the absence of conflict, the embodiments in the present application and the features in the embodiments may be combined with each other. The present invention is further described in detail below in conjunction with the drawings and specific embodiments.

[0029] In the embodiments of the present invention, if there are directional indications (such as up, down, left, right, front, back, etc.), they are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0030] In addition, in the present invention, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" or "second" may explicitly or implicitly include at least one of the features.

[0031] The present invention evaluates the impact of the adjustment of the bus network (the network composed of all bus routes in a city is called the bus network) on passenger flow by analyzing the total OD. The process of the present invention is:

[0032] 1) Using historical data, a machine learning model is built to find the probability of a passenger choosing one of two accessible paths under the same OD (i.e., the origin and destination of a trip).

[0033] 2) Find out the travel OD that may be affected based on the network topology changes before and after the line network adjustment;

[0034] 3) For each OD that may be affected, search for accessible paths before and after the network adjustment (paths between ODs: combinations of accessible bus routes between ODs, including transfers. For example, you can take route 1 from O to D, or you can take route 2 from station O, then transfer to route 3 at a certain station and finally arrive at station D. Here, "route 1" and "route 2 transfer to route 3" are both paths between ODs), and then calculate the passenger flow distribution between these accessible paths based on historical passenger path choices and the machine learning model established earlier;

[0035] 4) Based on the passenger flow distribution results of each OD in the previous step, comprehensively calculate the passenger flow changes of each line before and after the line network adjustment.

[0036] Please refer to Figure 1 The method for evaluating the impact of bus line network adjustment on passenger flow in an embodiment of the present invention includes:

[0037] Step S1: Collecting network static data, line operation information, and passenger flow OD data for a preset historical period;

[0038] Step S2: Building a machine learning model for the probability of passengers choosing different routes;

[0039] Step S3: Obtain the corresponding affected OD according to the adjustment information of the bus network;

[0040] Step S4: Calculate the passenger flow distribution of each affected OD on each path under the adjusted bus network through a machine learning model;

[0041] Step S5: Calculate the passenger flow of each line after the bus network adjustment, and then compare it with the passenger flow before the bus network adjustment to obtain an evaluation of the impact of the bus network adjustment on the passenger flow.

[0042] In order to better illustrate the embodiment of the present invention, it is assumed that the adjustment information of the bus line network is:

[0043] 1) Add a route A, given the service time of route A and the departure intervals in different peak periods (peak period: morning peak, off-peak and evening peak. The peak period in each city is different. The common ones are 7:00-9:00 for the morning peak, 17:00-19:00 for the evening peak, and the rest of the time for off-peak);

[0044] 2) A line B was deleted;

[0045] 3) Route C is adjusted to C', given the service time and departure interval of route C' under different peak periods;

[0046] 4) The location of the stations in the network has not changed.

[0047] Step S1, historical data preparation:

[0048] The data required includes the following data for a certain historical period (for example, the past 3 months):

[0049] 1) Static data of the line network: line shape, stations and cross-section distances; station longitude and latitude, and station pairs that can be transferred on foot (for example, if you take a certain line and get off at station X, then walk to station Y to transfer to another line, then X and Y are station pairs that can be transferred);

[0050] 2) Line operation information: the first and last service times of the line, the operating speeds and departure intervals during each peak period;

[0051] 3) Passenger flow OD data: departure station, arrival station, route, travel distance, number of transfers, transfer distance under each OD, as well as travel time and OD volume (passenger flow between ODs, referred to as OD volume) in each time period (such as every half hour).

[0052] At present, the level of informatization of the public transportation industry is constantly increasing. Based on the existing technology and the existing information of the public transportation industry, it is not difficult to obtain the above data.

[0053] Step S2, establishing a machine learning model for the probability of passengers choosing different routes.

[0054] There may be multiple paths from station O to station D in the bus network. Step S1 also gives the actual results of passengers' path selection in history. This step is to establish a mathematical model of the probability of passengers choosing different paths.

[0055] The input of the model is the departure interval, travel time, travel distance, number of transfers and transfer distance of two routes under the same OD in different peak periods. If the route includes multiple routes, the departure interval is the sum of the departure intervals of these routes. The output of the model is the probability of the passenger choosing the first route. p Obviously, the probability that a passenger chooses the second route is 1- p .

[0056] The present invention can use a multi-layer perceptron neural network as a model framework, and train the passenger's choice probability through a big data machine learning method. The neural network model has good flexibility, can simulate any function, avoids the disadvantage of determining the model form in advance, and can better reflect objective reality.

[0057] After the neural network model is trained, a common sense judgment must be made in the specific application, that is, if the parameters of the first route are not worse than those of the second route (the departure interval is not longer, the travel time is not longer, the travel distance is not longer, the number of transfers is not more, and the transfer distance is not longer), then the output of the model needs to be between [0.5, 1], that is, the probability of the passenger choosing the first route is at least 0.5. The purpose of this is to avoid possible common sense errors.

[0058] Let this model be NNM , the structural diagram is as follows Figure 2 The specific training steps are not the focus of the present invention and will not be described in detail here.

[0059] Step S3, finding out the OD that may be affected by the passenger flow.

[0060] For a certain OD, if the line network adjustment causes the original static information of the route (such as the route) to change, or the operational information (such as the departure interval) to change, the passenger flow under the OD may be affected. Use the following method to find the OD that may be affected:

[0061] 1) The departure station is on route A or the arrival station is on route A; or

[0062] 2) The departure station is on route C' or the arrival station is on route C'; or

[0063] 3) The path of OD includes line B; or

[0064] 4) The path of OD contains line C.

[0065] Step S4, calculating the passenger flow distribution of each path of each OD in the adjusted network.

[0066] The calculation in this step is performed by time period (such as half an hour), and the final passenger flow distribution result is the sum of the passenger flow distribution results of each time period:

[0067] Step S41: Find the shortest path between ODs starting from different lines;

[0068] Step S42: Calculate the passenger flow distribution of each path in each time period:

[0069] a. Determine the paths involved in passenger flow distribution;

[0070] b. Calculate the passenger flow distribution weight of each path based on the historical passenger flow distribution results and the NNM model;

[0071] c. Calculate the distribution of OD among each path.

[0072] Step S43: Summarize the passenger flow distribution in each time period.

[0073] Step S41, finding the shortest path between ODs starting from different lines.

[0074] Assume that the line passing through station O has m strip: , the routes passing through station D are n strip: . Calculate the shortest path from different routes from station O to station D, with a maximum of This problem can be regarded as a path search problem of a directed graph, which can be solved by algorithms such as A*, but it is not the focus of the present invention and will not be described in detail here.

[0075] Step S42, calculate the passenger flow distribution of each path in each time period:

[0076] a. Determine the paths involved in passenger flow distribution. The paths involved in distribution include:

[0077] 1) For the shortest path found in the previous step, if the service time of the route involved can meet the passengers' travel needs, it will participate in passenger flow allocation;

[0078] 2) If the passenger's travel path before the network adjustment still exists in the adjusted network, this path will also be included in the passenger flow distribution; this path may not be the shortest path between ODs, so it may not be included in the shortest path found in the previous step.

[0079] b. Calculate the passenger flow distribution weight of each path:

[0080] The passenger flow distribution weight of the path is determined as follows:

[0081] 1) The allocation weight of the historical path is the historical passenger flow of the path under this OD; the historical path is the path not involved in the line network adjustment.

[0082] 2) If a route contains adjusted or newly added lines, but the route is "the same" as a route before the network adjustment, the allocation weight of the route is the passenger flow of the route before the network adjustment. "The same" means that the departure interval and route length of the two routes are close enough, and the difference is within the preset threshold range. That is, although the route involves adjusted or newly added lines, under this OD, there is no substantial change between the route and a route before the network adjustment.

[0083] 3) For paths that have undergone substantial changes, the paths in 1) and 2) above are used as reference paths, and the allocation weight of this path is calculated using the following method.

[0084] The calculation method of the distribution weight in step 3) is as follows:

[0085] Assume that the reference paths involved in 1) and 2) are k strip: ,this k The passenger flow distribution weights of the paths are , then the lines involved in 3) L The weight is: .

[0086] in For use NNM Calculated passengers on the route L and reference line Choose between probability.

[0087] It can be seen that the above method utilizes the information of historical passenger route selection, effectively eliminates the impact of the randomness of passenger travel on passenger flow distribution, and avoids counterintuitive passenger flow distribution results.

[0088] c. Calculate the passenger flow assigned to each path.

[0089] Assuming that the demand for public transportation does not change before and after the bus network adjustment, the demand is recorded as od_cnt .

[0090] Based on the result of the previous step, the passenger flow obtained by each path under the current OD is further calculated, that is, the passenger flow is allocated proportionally according to the passenger flow allocation weight. Note that the sum of the weights here is usually not equal to 1.

[0091] Assuming that there is a common r paths participate in passenger flow distribution, and the weights of passenger flow distribution are . Then the passenger flow allocated to each path is: . Where j=1,2,…,r.

[0092] Step S43, summarizing the path allocation results of each time period.

[0093] The above steps are for the path allocation results under different time periods of this OD. After summarizing them, the passenger flow allocation results of different paths under this OD are obtained.

[0094] Step S5, comprehensively calculating the passenger flow impact assessment of each line.

[0095] The previous steps have given the passenger flow distribution of each potentially affected OD under each path. On this basis, the passenger flow of each route after the network adjustment can be calculated. Then, the passenger flow before the network adjustment can be compared to obtain an assessment of the impact of the bus network adjustment on passenger flow:

[0096] 1) If the OD site pair of a line (the OD site pair of a line is a site pair consisting of two sites in the direction of the line. For example, a line has 10 sites in the upstream direction: the 1st site to the 10th site. Then sites 1-3, 2-8, etc. are all OD site pairs, which can be expressed as sites i -Site j , ) are not in the affected OD found in step S3, then the passenger flow of the line remains unchanged before and after the line network adjustment.

[0097] 2) If a line has an OD station pair in the affected OD found in step S3, the final passenger flow distribution of the line can be obtained according to the calculation results of all ODs related to the line in step S4. Then, by comparing with the passenger flow of the line before the line network adjustment, the passenger flow impact assessment of the line can be obtained: whether the passenger flow has increased or decreased, and the specific amount.

[0098] The evaluation system of the impact of bus line network adjustment on passenger flow in the embodiment of the present invention includes:

[0099] Historical data preparation module: collects network static data, line operation information, and passenger flow OD data for a preset historical period;

[0100] Model building module: build a machine learning model of the probability of passengers choosing different routes;

[0101] Affected OD acquisition module: obtains the corresponding affected OD according to the adjustment information of the bus network;

[0102] Passenger flow distribution module: Calculate the passenger flow distribution of each affected OD on each path under the adjusted bus network through machine learning model;

[0103] Impact assessment module: calculates the passenger flow of each line after the bus network adjustment, and then compares it with the passenger flow before the bus network adjustment to obtain an assessment of the impact of the bus network adjustment on passenger flow.

[0104] As an implementation method, the passenger flow distribution module obtains the passenger flow distribution of each path according to the following steps:

[0105] Find the shortest path between ODs starting from different lines;

[0106] Calculate the passenger flow distribution of each path in each time period:

[0107] a. Determine the paths involved in passenger flow distribution;

[0108] b. Calculate the passenger flow distribution weight of each path based on historical passenger flow distribution results and machine learning models;

[0109] c. Calculate the passenger flow allocated to each path;

[0110] Summarize the passenger flow distribution in each time period.

[0111] The present invention comprehensively evaluates the passenger flow changes of the affected routes after the bus network adjustment: a) finds the ODs that may be affected; b) calculates the passenger flow distribution of each path of these ODs under the adjusted bus network; c) comprehensively calculates the passenger flow changes of the affected routes. The present invention significantly improves the accuracy of the evaluation of the impact of bus network adjustment on passenger flow.

[0112] As an implementation method, the passenger flow distribution module determines the passenger flow distribution weight of the path according to the following method:

[0113] 1) The allocation weight of the historical route is the historical passenger flow of the route under this OD; the historical route is the route not involved in the bus network adjustment;

[0114] 2) If a route contains adjusted or newly added lines, but the route is the same as one of the routes before the bus network adjustment, the allocation weight of the route is the passenger flow of the route before the bus network adjustment;

[0115] 3) For the paths that have undergone substantial changes, the paths in 1) and 2) above are used as reference paths, and the allocation weight of this path is calculated using the following method:

[0116] Assume that the reference paths involved in 1) and 2) are k strip: ,this k The passenger flow distribution weights of the paths are , then the lines involved in 3) L The weight is: ;

[0117] in The machine learning model calculates the number of passengers on the route L and reference line Choose between The probability of , i∈k.

[0118] As an implementation method, the passenger flow distribution module distributes the passenger flow of each path in proportion according to the passenger flow distribution weight:

[0119] Assuming that there is a common r paths participate in passenger flow distribution, and the weights of passenger flow distribution are ; The passenger flow allocated to each path is: ; where j=1,2,…,r; The demand for public transportation.

[0120] As an implementation method, the input of the machine learning model is the departure interval, travel time, travel distance, number of transfers and transfer distance of two routes under the same OD in different peak periods. If the route includes multiple routes, the departure interval is the sum of the departure intervals of these routes; the output of the model is the probability of the passenger choosing the first route. p .

[0121] The present invention constructs a neural network model to evaluate the probability of passengers choosing between two routes under the same OD, further improving the accuracy of the evaluation of the impact of bus network adjustment on passenger flow. The input of the model includes the peak section, the departure interval of the two routes, the travel time, the travel distance, the number of transfers and the transfer distance, and the output of the model is the probability of passengers choosing the first route.

[0122] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for evaluating the impact of bus network adjustment on passenger flow, characterized in that: include: Step S1: Collecting network static data, line operation information, and passenger flow OD data for a preset historical period; Step S2: Building a machine learning model for the probability of passengers choosing different routes; Step S3: Obtain the corresponding affected OD according to the adjustment information of the bus network; Step S4: Calculate the passenger flow distribution of each affected OD on each path under the adjusted bus network through a machine learning model; Step S5: Calculate the passenger flow of each route after the bus network adjustment, and then compare it with the passenger flow before the bus network adjustment to obtain an evaluation of the impact of the bus network adjustment on the passenger flow; The step S4 comprises: Step S41: Find the shortest path between ODs starting from different lines; Step S42: Calculate the passenger flow distribution of each path in each time period: a. Determine the paths involved in passenger flow distribution; b. Calculate the passenger flow distribution weight of each path based on historical passenger flow distribution results and machine learning models; c. Calculate the passenger flow allocated to each path; Step S43: Summarize the passenger flow distribution in each time period; In step S42, the passenger flow distribution weight of the path is determined according to the following method: 1) The allocation weight of the historical route is the historical passenger flow of the route under this OD; the historical route is the route not involved in the bus network adjustment; 2) If a route contains adjusted or newly added lines, but the route is the same as one of the routes before the bus network adjustment, the allocation weight of the route is the passenger flow of the route before the bus network adjustment; 3) For the paths that have undergone substantial changes, the paths in 1) and 2) above are used as reference paths, and the allocation weight of this path is calculated using the following method: Assume that the reference paths involved in 1) and 2) are k strip: ,this k The passenger flow distribution weights of the paths are , then the lines involved in 3) L The weight is: ; in The machine learning model calculates the number of passengers on the route L and reference line Choose between The probability of , i∈k.

2. The method for evaluating the impact of bus line network adjustment on passenger flow as claimed in claim 1, characterized in that: In step S42, the passenger flow of each path is proportionally distributed according to the passenger flow distribution weight: Assuming that there is a common r paths participate in passenger flow distribution, and the weights of passenger flow distribution are ; The passenger flow allocated to each path is: ; where j=1,2,…,r; The demand for public transportation.

3. The method for evaluating the impact of bus network adjustment on passenger flow as claimed in claim 1, characterized in that: The input of the machine learning model is the departure interval, travel time, travel distance, number of transfers and transfer distance of two routes under the same OD in different peak periods. If the route includes multiple routes, the departure interval is the sum of the departure intervals of these routes; the output of the model is the probability of the passenger choosing the first route. p .

4. A system for evaluating the impact of bus network adjustment on passenger flow, characterized in that: include: Historical data preparation module: collects network static data, line operation information, and passenger flow OD data for a preset historical period; Model building module: build a machine learning model of the probability of passengers choosing different routes; Affected OD acquisition module: obtains the corresponding affected OD according to the adjustment information of the bus network; Passenger flow distribution module: Calculate the passenger flow distribution of each affected OD on each path under the adjusted bus network through machine learning model; Impact assessment module: calculates the passenger flow of each line after the bus network adjustment, and then compares it with the passenger flow before the bus network adjustment to obtain an assessment of the impact of the bus network adjustment on passenger flow; The passenger flow distribution module obtains the passenger flow distribution of each path according to the following steps: Find the shortest path between ODs starting from different lines; Calculate the passenger flow distribution of each path in each time period: a. Determine the paths involved in passenger flow distribution; b. Calculate the passenger flow distribution weight of each path based on historical passenger flow distribution results and machine learning models; c. Calculate the passenger flow allocated to each path; Summarize passenger flow distribution in each time period; The passenger flow distribution module determines the passenger flow distribution weight of the path according to the following method: 1) The allocation weight of the historical route is the historical passenger flow of the route under this OD; the historical route is the route not involved in the bus network adjustment; 2) If a route contains adjusted or newly added lines, but the route is the same as one of the routes before the bus network adjustment, the allocation weight of the route is the passenger flow of the route before the bus network adjustment; 3) For the paths that have undergone substantial changes, the paths in 1) and 2) above are used as reference paths, and the allocation weight of this path is calculated using the following method: Assume that the reference paths involved in 1) and 2) are k strip: ,this k The passenger flow distribution weights of the paths are , then the lines involved in 3) L The weight is: ; in The machine learning model calculates the number of passengers on the route L and reference line Choose between The probability of , i∈k.

5. The evaluation system for the impact of bus network adjustment on passenger flow as claimed in claim 4, characterized in that: The passenger flow distribution module distributes the passenger flow of each path in proportion according to the passenger flow distribution weight: Assuming that there is a common r paths participate in passenger flow distribution, and the weights of passenger flow distribution are ; The passenger flow allocated to each path is: ; where j=1,2,…,r; The demand for public transportation.

6. The evaluation system for the impact of bus network adjustment on passenger flow as claimed in claim 4, characterized in that: The input of the machine learning model is the departure interval, travel time, travel distance, number of transfers and transfer distance of two routes under the same OD in different peak periods. If the route includes multiple routes, the departure interval is the sum of the departure intervals of these routes; the output of the model is the probability of the passenger choosing the first route. p .

Images