A method for predicting bus alighting considering walking

Abstract

The present invention is aimed at the three existing situations of same-station transfer, short-distance walking and BRT transfer. In the existing supervised learning model, which is difficult to train and covers limited scenarios, and does not consider the problem of walking arrival, a bus alighting prediction method considering walking is proposed. The downstream reachable station set is first calculated based on the static line station relationship data of the bus network in the case of boarding at each non-terminal line station. The downstream reachable station set considers the walkable stations within the walking threshold; when there is a BRT transfer station, the virtual and transaction data are added according to the actual route matching to increase the transfer station transfer and walkable stations; based on the downstream station data, the prediction of the alighting station is completed through the construction of the travel chain of the day, the matching of historical travel chains, and the judgment basis of high-frequency riding stations.

Description

Technical Field

[0001] The present invention relates to the field of public transportation data analysis, and in particular to a method for predicting bus getting off taking walking into consideration. Background Art

[0002] In bus network planning, data such as passenger distance and the number of passengers on a vehicle at a certain time are important indicators for network evaluation. However, under the one-ticket bus card swiping rule, it is impossible to directly obtain the passenger's get-off station. Therefore, it is necessary to predict the get-off of bus passengers. The current get-off prediction method mainly determines the get-off by whether the passenger's boarding station for the next transaction is on the ride route of the previous transaction. There are also supervised learning methods for model training. However, the first method covers very limited scenarios, for example, it does not consider the stations that passengers can reach by walking a short distance; the training data in the second method is almost impossible to obtain. At the same time, the characteristics of BRT stations without leaving the station also increase the difficulty of get-off prediction. Summary of the invention

[0003] The present invention aims at the three existing situations of same-station transfer, short-distance walking and BRT transfer. The existing supervised learning model is difficult to train and covers limited scenarios, and does not consider the problem of walking arrival. A bus getting off prediction method taking walking into account is proposed.

[0004] A method for predicting bus alighting taking walking into consideration, comprising:

[0005] Step 1: record the boarding route and boarding station where the current passenger has a ride transaction, and obtain the set of downstream reachable stations;

[0006] Step 2: Record the second ride transaction of the current passenger within a set time. If the second ride stop is in the set of downstream reachable stops of the current ride, a travel chain is formed, and the get-off is predicted and recorded as a travel chain in the travel history of the current passenger.

[0007] Step 3: Search the travel history of the current passenger at the current stop and make a get-off prediction;

[0008] Step 4: Record the boarding stations of the current passenger on the opposite bus within the set number of days, and infer the travel chain based on this and record it in the travel history of the current passenger;

[0009] Step 5: Record the set of downstream reachable stations of the current passenger, set the collection of ride transactions that occurred within a certain number of days, and use this to infer the travel chain and record it in the travel history of the current passenger;

[0010] Step 6: Query all types of travel chains of the current passenger, and determine whether the passenger is a BRT if the travel chain result is that the passenger cannot be directly reached through the boarding line.

[0011] Preferably, the step 1 first obtains the remaining stations of the current boarding route and the stations that can be reached on foot within a set value range of the remaining stations of the boarding route;

[0012] Get the remaining stops of the bus routes that are within walking distance of the station;

[0013] Add up all the remaining sites mentioned above to form the set of downstream reachable sites.

[0014] Preferably, many of the remaining stations are obtained based on the remaining reachable stations of each bus line that can be reached on foot, so there will be overlaps. Therefore, when constructing the remaining stations, the arrival time of each station is calculated, and the plan with the shortest time is recorded in the statistics.

[0015] Preferably, the step 6 first determines whether the vehicle in which the current passenger has a ride transaction is a BRT vehicle, and then determines whether the travel chain uses a BRT vehicle;

[0016] Continue to judge only when BRT vehicles appear;

[0017] When the result of the travel chain is that the BRT vehicle can reach the boarding point directly without boarding, but can reach the boarding point directly through the BRT transfer, a virtual travel chain is constructed.

[0018] Preferably, the basic setting value of the statistical days in step 4 is 1 day, the opposite boarding point of the current passenger is the current travel chain result, and the result of the travel chain of the opposite ride is the current boarding point.

[0019] Preferably, the transaction collection recorded in step five is not included in the set of downstream reachable sites obtained in step one, a scatter diagram is obtained according to the transaction collection recorded in step five, and the center point of the scatter diagram is taken, and the site closest to the center point of the scatter diagram is found in the set of downstream reachable sites as the inferred travel chain result.

[0020] The substantial effect of the present invention is that based on the static line-station relationship data of the bus network, a set of downstream reachable stations is first calculated and generated in the case of boarding at each non-terminal line station, and the downstream reachable station set takes into account the stations reachable on foot within the walking threshold; when a BRT transfer station occurs, virtual and transaction data are added according to the actual route matching to increase the stations that can be transferred and reached on foot at the transfer station; on the basis of the downstream station data, the prediction of the alighting point is completed through the construction of the travel chain of the day, the matching of historical travel chains, high-frequency riding stations and other judgment criteria. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Flow chart of the present invention. DETAILED DESCRIPTION

[0022] The technical solution of the present invention is further described in detail below through specific embodiments in conjunction with the accompanying drawings.

[0023] Example 1

[0024] A method for predicting bus alighting taking walking into consideration, comprising:

[0025] Step 1: record the boarding route and boarding station where the current passenger has a ride transaction, and obtain the set of downstream reachable stations;

[0026] Step 2: Record the second ride transaction of the current passenger within a set time. If the second ride stop is in the set of downstream reachable stops of the current ride, a travel chain is formed, and the get-off is predicted and recorded as a travel chain in the travel history of the current passenger.

[0027] Step 3: Search the travel history of the current passenger at the current stop and make a get-off prediction;

[0028] The current transaction is recorded as , where c represents the card number of the current transaction, r represents the bus route of the current transaction, i represents the boarding station of the current transaction, and t represents the time when the transaction occurs. is the first transaction of card c after time t. , it is called that this transaction can form a travel chain with the next transaction, and the alighting station of this transaction is j.

[0029] Step 4: Record the boarding stations of the current passenger on the opposite bus within the set number of days, and infer the travel chain based on this and record it in the travel history of the current passenger;

[0030] Step 5: Record the set of downstream reachable stations of the current passenger, set the collection of ride transactions that occurred within a certain number of days, and use this to infer the travel chain and record it in the travel history of the current passenger;

[0031] Step 6: Query all types of travel chains of the current passenger, and determine whether the passenger is a BRT if the travel chain result is that the passenger cannot be directly reached through the boarding line.

[0032] The bus route station relationship data table structure is as follows:

[0033] route_name stop_name Site longitude Site Latitude Station No. Up and down 1st Road Canal World 120.XXX 30.XXX 1 Up 1st Road Huayuangang Street, Xiaohe Road intersection 120.XXX 30.XXX 2 Up 1st Road Spinning Bridge 120.XXX 30.XXX 3 Up 1st Road Gao Family Garden 120.XXX 30.XXX 4 Up 1st Road Gongchen Bridge West 120.XXX 30.XXX 5 Up

[0034] The step 1 first obtains the remaining stations of the current boarding route and the stations that can be reached on foot within the set value range of the remaining stations of the boarding route; based on the station location data in the bus route station table, a spatial index is constructed through a GIS algorithm to quickly find stations within a certain distance threshold range (set as D) near all line stations.

[0035] Then, through the pre-calculated station-to-station walking distance table, the stations with walking distance greater than D are filtered out from the obtained station set, and the station set corresponding to each station that can be reached by walking after getting off is obtained, which is recorded as ,in represents line r, represents the i-th site.

[0036] Get the remaining stops of the bus routes that are within walking distance of the station;

[0037] All the remaining stations mentioned above are added together to form the set of downstream reachable stations. According to the bus riding rules, if a passenger gets on the bus at station i of line r, the station number of the possible alighting station must be greater than i. The possible alighting station set is recorded as . Then the set of stations that passengers can reach by bus + walking is .

[0038] Many of the remaining stations are obtained based on the remaining reachable stations of each bus line that can be reached on foot, so there will be overlaps. Therefore, when constructing the remaining stations, the arrival time of each station is calculated, and the plan with the shortest time is recorded in the statistics.

[0039] The set of stations within walking distance of getting off the line There may be duplications in the walkable station sets of different line stations, and it is necessary to select the best alighting station for the repeated walkable stations. The optimal selection scheme is ,in represents the travel time from station i of route r to station j. represents the walking time from station j to station h. The obtained station j is the corresponding alighting station that can be reached by foot from station h. The corresponding station set that can be reached by foot from station i of route r to station j is .

[0040] The step 6 first determines whether the vehicle in which the current passenger has a ride transaction is a BRT vehicle, and then determines whether the travel chain uses a BRT vehicle;

[0041] Continue to judge only when BRT vehicles appear;

[0042] When the result of the travel chain is that the BRT vehicle can reach the boarding point directly without boarding, but can reach the boarding point directly through the BRT transfer, a virtual travel chain is constructed.

[0043] The basic setting value of the statistical days in step 4 is 1 day, the opposite boarding point of the current passenger is the current travel chain result, and the result of the travel chain of the opposite ride is the current boarding point.

[0044] remember is the number of transactions that the passenger has made at station j on route r in history, The passenger's history on the route Site The number of transactions on board, is the opposite line of line r, and In the downstream station set of the current boarding station i. , is the transaction frequency threshold of the determined high-frequency site, then the site is the drop-off point for the current transaction, if and , then the site The drop-off point for the current transaction.

[0045] The transaction collection recorded in step five is not included in the set of downstream reachable sites obtained in step one. A scatter diagram is obtained according to the transaction collection recorded in step five, and the center point of the scatter diagram is taken. The site closest to the center point of the scatter diagram is found in the set of downstream reachable sites as the inferred travel chain result.

[0046] remember is the drop-off point for the current transaction, where It represents the number of transactions that occurred at station j on line r within a certain historical time interval.

[0047] Example 2

[0048] Assume that passenger A gets on bus No. 1 at stop No. 1a in a certain city at 8:00 a.m., and transfers to bus No. 2 at stop No. 1b at 8:40 a.m.

[0049] According to the method of the present invention, in step 1, when A gets on the bus, a set of downstream reachable sites is obtained; wherein 1b belongs to the set of downstream reachable sites.

[0050] Step 2: If passenger A has no travel history, the get-off prediction will not be made immediately.

[0051] Step three, passenger A made the second transaction at 8:40, and the location of the second ride transaction was within the set of downstream reachable sites, thus forming the first travel chain of the day. The prediction result for passenger A is recorded in history and a prediction for getting off is made.

[0052] Example 3

[0053] Assume that passenger A gets on bus No. 1 at stop 1a in a certain city at 8:00 a.m. and takes the oncoming bus of bus No. 1 at stop 1c at 18:00 p.m.

[0054] According to the method of the present invention, in step 1, when A gets on the bus, a set of downstream reachable sites is obtained; among them, 1c belongs to the set of downstream reachable sites.

[0055] Step 2: If passenger A has no travel history, the get-off prediction will not be made immediately.

[0056] No second transaction occurs after the passenger gets on the bus, so the trip chain cannot be obtained from step three.

[0057] Step 4: If the passenger takes the oncoming bus of route 1 at 1c at 18:00 in the afternoon of the same day, the trip chain result in the morning is set to 1c and a prediction is made; at the same time, the trip chain result starting from 1c in the afternoon is set to 1a and a prediction is made.

Claims

1. A method for predicting bus getting off taking into account walking, characterized in that: include: Step 1: record the boarding route and boarding station where the current passenger has a ride transaction, and obtain the set of downstream reachable stations; All remaining stations are added together to form a set of downstream reachable stations. According to the bus riding rules, if a passenger gets on the bus at station i of line r, the station number of the possible alighting station must be greater than i. The possible alighting station set is recorded as Then the set of stations that passengers can reach by bus and walking is Get off the bus on the route and walk to the station set S r,i In the , there may be duplications in the walkable station sets of different line stations, and it is necessary to select the optimal alighting station for the repeated walkable stations; the optimal selection scheme is in represents the travel time from getting on at stop i of route r to getting off at stop j, represents the walking time from station j to station h; The obtained station j is the corresponding alighting station that can be reached by foot at station h; the set of stations that can be reached by foot when a passenger gets on at station i of route r and gets off at station j is S r,i,j ; Step 2: Record the second ride transaction of the current passenger within a set time. If the second ride stop is in the set of downstream reachable stops of the current ride, a travel chain is formed, and the get-off is predicted and recorded as a travel chain in the travel history of the current passenger. Step 3: Search the current passenger's travel history at the current stop and make a get-off prediction; The current transaction is recorded as Where c represents the card number of the current transaction, r represents the bus route of the current transaction, i represents the boarding station of the current transaction, and t represents the time when the transaction occurs. is the first transaction of card c after time t; if i′∈∪ j S r,i,j , it is called that this transaction can form a travel chain with the next transaction, and the alighting station of this transaction is j; Step 4: Record the boarding stations of the current passenger on the opposite bus within the set number of days, and infer the travel chain based on this and record it in the travel history of the current passenger; Step 5: Record the set of downstream reachable stations of the current passenger, set the collection of ride transactions that occurred within a certain number of days, and use this to infer the travel chain and record it in the travel history of the current passenger; remember is the number of transactions that the passenger has made at station j on route r in history, is the number of transactions that the passenger has made in the past at station j’ on route r’, where r’ is the opposite route of route r, and j, j’ are located in the downstream station set of the current boarding station i; if C′ is the transaction frequency threshold of the determined high-frequency site, then the site is the drop-off point for the current transaction, if and Site The drop-off point for the current transaction; Obtain a scatter diagram according to the transaction collection recorded in step 5, take the center point of the scatter diagram, and find the station closest to the center point of the scatter diagram in the set of downstream reachable stations as the inferred travel chain result; Let h = max j CT r,j The drop-off point for the current transaction, where CT r,j It represents the number of transactions that occurred at station j on line r within a certain historical time interval; Step 6: Query all types of travel chains of the current passenger, and determine whether the passenger is a BRT if the travel chain result is that the passenger cannot be directly reached through the boarding line.

2. A method for predicting bus getting off taking into account walking according to claim 1, characterized in that: The step 1 first obtains the remaining stations of the current boarding route and the stations that can be reached on foot within the set value range of the remaining stations of the boarding route; Get the remaining stops of the bus routes that are within walking distance of the station; Add up all the remaining sites mentioned above to form the set of downstream reachable sites.

3. A method for predicting bus getting off taking into account walking according to claim 2, characterized in that: Many of the remaining stations are obtained based on the remaining reachable stations of each bus line that can be reached on foot, so there will be overlaps. Therefore, when constructing the remaining stations, the arrival time of each station is calculated, and the plan with the shortest time is recorded in the statistics.

4. The method for predicting bus getting off taking into account walking according to claim 1, characterized in that: The step 6 first determines whether the vehicle in which the current passenger has a ride transaction is a BRT vehicle, and then determines whether the travel chain uses a BRT vehicle; Continue to judge only when BRT vehicles appear; When the result of the travel chain is that the BRT vehicle can reach the boarding point directly without boarding, but can reach the boarding point directly through the BRT transfer, a virtual travel chain is constructed.

5. The method for predicting bus getting off taking into account walking according to claim 1, characterized in that: The basic setting value of the number of days in step 4 is 1 day, the opposite boarding point of the current passenger is the current travel chain result, and the result of the travel chain of the opposite ride is the current boarding point.

Images