CN111339159B - An Analysis and Mining Method for One-ticket Bus Data - Google Patents

Abstract

The application relates to the technical field of traffic, in particular to an analysis and mining method of one-ticket bus data, which comprises the steps of firstly cleaning bus IC card data, bus GPS data, vehicle-mounted machine data and one-way station information; screening bus IC card data including transaction card numbers, transaction time, lines and license plates, firstly extracting information of the transaction card numbers, the transaction time, the lines and the license plates, then deleting useless fields except the bus IC card data, and deleting records of partial field loss; the useful fields in the vehicle-mounted GPS data of the bus are extracted and comprise vehicle-mounted machine numbers, arrival and departure information, positioning time, positioning longitude and latitude, line numbers, sub-line numbers, sequence numbers, bus speeds and bus driving mileage fields, and useless fields are deleted. The application analyzes the site distribution with larger passenger transfer quantity under the existing operation scheme and provides reliable reference data for the adjustment of the subsequent lines.

Description

An Analysis and Mining Method for One-ticket Bus Data

technical field

The invention relates to the technical field of traffic, in particular to a method for analyzing and mining public transport data of a one-ticket system.

Background technique

In recent years, with the rapid development of smart bus systems, mining big data of bus swiping cards has become a new means to guide operation planning, but the one-ticket bus system lacks the information of passengers' alighting stations, which hinders the application of corresponding new data. With the continuous development of technology, traffic big data has become a current research hotspot. Compared with traditional traffic survey methods, traffic big data has lower acquisition costs, but contains more abundant information. The traffic behavior of passengers becomes possible, which provides a new perspective for traditional traffic research; on the other hand, mining traffic big data can also conduct other research such as urban structure detection and urban planning. In many fields, existing studies have shown that traffic big data has broad application prospects.

In recent years, with the rapid development of intelligent public transportation systems across the country, passenger IC card automatic toll collection systems and vehicle-mounted GPS automatic positioning systems have been widely used, and a wealth of traffic big data resources has been accumulated, which provides a basis for obtaining real-time and comprehensive bus passenger flow data. new technical means. However, in most cities in China, only subways and some BRT lines adopt the form of swiping in and out of the station. For conventional buses with wider coverage and greater passenger flow, the one-ticket system is usually adopted. Passengers only need to swipe their cards when boarding and get off. There is no need to swipe the card at the time, resulting in the lack of information such as boarding and exit stations and transfer records in the card swipe record, making these data unable to be directly used.

Contents of the invention

In view of this, the object of the present invention is to provide a method for analyzing and mining bus data of one-ticket system, so as to solve the problems in the background technology.

The invention provides an analysis and mining method for one-ticket bus data, first cleaning bus IC card data, bus GPS data, vehicle-mounted machine data and one-way station information.

Further, the valid fields in the bus passenger IC card data include transaction card number, transaction date, driving line and license plate number. The main fields and their explanations are shown in Table 1.

Table 1 Main fields of bus IC card data and their explanations

field field description CARDNO Trading Card Number TRADEDATE transaction hour ROUTE CODE line VEHICLE CODE number plate

Data cleaning for bus card swiping is mainly to delete logically obviously unreasonable records. The main processing methods are as follows:

S1: First extract the required fields for research, and delete useless fields;

S2: Then delete the records with missing fields.

Furthermore, there are a total of 59 fields in the on-board GPS data of the bus, but some fields are not yet enabled and are empty. Among them, useful fields include on-board machine number, arrival and departure information, positioning time, positioning latitude and longitude, line number, sub-line number, Sequence number, bus speed, bus mileage and other fields, the main fields and their explanations are shown in Table 2.

Table 2 Main fields of GPS data and their explanations

field field description field field description PRODUCTID Vehicle number LATITUDE latitude ISARRLFT to departure information ROUTEID line number ACTDATETIME positioning time SUBROUTEID sub-line number LONGITUDE longitude STATION SEQ NUM Site sequence number

When the bus enters and leaves the station, GPS positioning records will be generated, and arrival and departure data will be generated within 5 meters before and after the station. For GPS data, the main data cleaning process is:

S1: Extract the required fields for research and delete useless fields;

S2: delete the record of latitude and longitude outside the scope of the monitoring area based on ArcGIS;

S3: delete only arrival or only departure data.

Further, vehicle-mounted machine data and vehicle-mounted machine information refer to the license plate number and line name corresponding to the vehicle-mounted machine number, which are used to match the license plate number corresponding to GPS data, as well as the association and fusion of GPS data and IC card data. The data samples are shown in Table 3 shown.

Table 3 Comparison table of on-board machine information

Vehicle frame number number plate line name 20111271 AA1271 42 road 20111601 AA1601 306 Road

Further, the one-way station information table and the one-way station relationship table are the station sequence number, station name and station type corresponding to the line number and sub-line number. Since there are only station sequence numbers in the GPS data, and there is no positioning station name, the one-way station is used The relationship table matches the positioning station name to the GPS data, and the sample data of the table is shown in Table 4. After filtering the line number and sub-line number, the station sequence number and station name have a one-to-one correspondence.

Table 4 One-way station relationship table

line number sub-line number Site Type Number Site sequence number Site name 1 1 3 43 Museum (Single) (East) 1 1 3 44 Archives (East) 1 1 3 45 Democratic Party Building (East)

In the one-way station relationship table, many stations are divided into four directions: east, west, south, north, and there are often many adjacent longitudes and latitudes in the GIS map of the same station. Take the average value of latitude and longitude for fusion to obtain the unique latitude and longitude value of the site, as shown in Figure 1.

The GPS data is fused with the on-board machine information and the fused one-way site relationship table to obtain the GPS data including the license plate number, site name, and longitude and latitude of the site. The data samples are shown in Table 5.

Table 5 One-way station relationship table after matching latitude and longitude

line number sub-line number Site sequence number Site name longitude latitude 2 2001 3 People's Hospital 106.2523192 38.5045019 2 2001 4 People's Hospital 106.2523192 38.5045019 2 2001 5 garden 106.251383 38.4981606

Further, site inference method:

Passenger swiping card data of the "one-ticket system" bus lacks the passenger's boarding and alighting station and transfer station information. In order to complete these information, the following algorithm is proposed:

Passenger boarding site inference:

The bus GPS data is fused with the passenger IC card data, and the boarding site of the passenger is determined by comparing the passenger's card swiping time and the GPS data update time of the station. The inference algorithm is as follows:

Input the original passenger card data UserData; bus GPS data VehiclesGPS; license plate number list Vehicles;

Output: Matched credit card data set Result;

Among them, the Selectdata(data, condition) function means to extract data satisfying the condition condition from data; the ComputeInterval(A, B) function means to calculate the time interval between A and B. Due to the error between GPS positioning time and card swiping time, data with a difference of more than 180 seconds between GPS positioning time and card swiping time is eliminated in the algorithm to ensure the accuracy of the matching results. Specific steps are as follows:

Input: original passenger card data UserData; bus GPS data VehiclesGPS; license plate number list Vehicles;

Output: Matched credit card data set Result;

Define i to represent each license plate number, a is the record of the license plate number i in the passenger swiping card data, b is the record of the license plate number i in the bus GPS data, and j is each card swiping record in a;

First, set the initialization boarding station to be empty; the initialization time interval is infinite, the record station name is k, the boarding station name is added to record j, Selectdata(data, condition) is each record in b, ComputeInterval(A,B) is Record j and record k time intervals, remove the data with a difference between the GPS positioning time and card swiping time greater than 180 seconds, and record the record j after adding the boarding station into Result; then execute the next cycle.

Further, there are two methods for passenger alighting station inference:

Different passengers use the bus to travel different times every day. Some passengers travel multiple times a day, while a large number of passengers only travel once a day. For these two different situations, use the following two methods to complete the passenger drop-off station the guessing process.

Further, the first method of passenger alighting stop inference is based on the alighting stop inference of the passenger trip chain

For passengers who use public transport to travel multiple times in a day, the daily data includes multiple card swiping records, which can form a closed bus travel chain or an unclosed bus travel chain. This paper uses the passenger travel chain to infer the passenger alighting station. The process is as follows:

S1: extract all the card swiping records in the same day of the same day in the passenger card swiping record, and sort by the time of swiping the card;

S2: for a passenger, at first according to the boarding station in the previous record of the passenger, obtain all the stations of the line that the passenger takes this trip;

S3: in the passenger's next ride record, the station with the nearest station space distance of all stations on the last ride line, then this station is the alighting station of the passenger's previous ride;

S4: When the card swiping information in S2 is the last card swiping record of this card number, then utilize the passenger's first card swiping record as the next passenger record, thereby calculating the alighting place when it takes the bus for the last time, the card number The calculation of the alighting station ends;

S5: For all card numbers, continuously run steps S1-S4 until all card numbers complete the inference process.

Further, the second method of passenger alighting stop inference is based on probabilistic alighting stop inference:

For passengers who do not have consecutive bus trips in one day, this paper applies a passenger alighting station estimation model based on the probability of alighting at the station. Existing studies have shown that the attraction strength and occurrence intensity of bus stations are basically balanced, so the occurrence intensity of the station can be used to replace the station's Strength of attraction. According to the deduction results of passenger boarding stations, the number of people boarding at each station of any line can be counted, and the attraction strength of the stations can be calculated as shown in Equation 2:

In the formula, s _i represents the number of people getting on the train at the i-th station.

The probability p _ij of passengers getting off is related to the average number of bus stops and the attraction strength p _i of the bus stops. The number of bus stops taken by residents is mainly concentrated within a certain range. Statistical experience shows that in a fixed driving direction , the number of bus stops approximately conforms to the Poisson distribution, as shown in Equation 3:

In the formula, Z _ij represents the probability of passengers getting on the bus at station i and getting off at station j; λ represents the average number of bus stops for bus travel. When the number of stations after station i is less than λ, λ=(n-λ), n is a single line The total number of bus stops, from which the probability of passengers getting off at station j from station i can be constructed, as shown in Equation 4:

So far, the total number of passengers boarding at station i and getting off at station j can be obtained as shown in Equation 5:

M _ij ＝s _i ×p _ij (Formula 5)

Further, transfer station inference

The identification of bus interchange can be considered from the perspective of time and space. As shown in Figure 2, bus passengers swipe their cards to get on the bus at the time t1 at the P1 station, and the bus arrives at the P2 station after the time T1 to t2. The walking distance is L, and it takes T2 to arrive at the transfer station P3, and waits for the time T3 to t3 to swipe the card. Car, take the bus station of the transfer route, and arrive at the terminal station P4 at the final running time from T4 to t4 to complete this trip.

Then the T _s consumed during the transfer process is expressed as formula 1:

T _s ＝t ₃ -t ₂ ＝T _walk +T _wait ＝t ₃ -t ₁ -T _v (Formula 1)

Among them, T _walk is the walking time from the getting off site to the transfer site; T _wait is the waiting time at the transfer site; T _v is the last time on the bus.

Further, by analyzing the maximum value of bus ride time T _v , transfer walking time T _walk , and transfer station waiting time T _wait , the maximum transfer time interval can be obtained. This paper combines existing literature and traffic surveys to take the maximum possible The transfer threshold is 60 minutes.

Then the transfer recognition process steps are as follows:

S1: Extract a bus IC card record, record the card swiping time as t1, obtain the adjacent card swiping records of the same card number, and record the card swiping time as t2;

S2: Calculate the card swiping time interval Ti=t2-t1, if Ti≤Tmax, and the distance between transfer stations L<500m, it is considered that the passenger's last trip is a transfer behavior, otherwise it is considered a trip;

S3: judge all data of the same card number, and record the result of identification;

S4: Steps S1-S3 are repeated until the last card to complete passenger transfer behavior identification.

Beneficial effects of a method for analyzing and mining one-ticket bus data of the present invention: a complete one-ticket public bus data mining process is proposed, which includes methods such as original data cleaning, passenger getting on and off station estimation, passenger transfer station estimation, etc. A large amount of passenger flow data has been collected. Based on this data, it is convenient to provide reliable data support for bus operation and route planning. This algorithm identifies the distribution of main passenger flow distribution centers; obtains the daily passenger flow and spatial distribution of the operating lines, and thus macroscopically monitors the residents. It has an intuitive understanding of travel needs; analyzed the distribution of stations with a large number of passenger transfers under the existing operation plan, and provided reliable reference data for subsequent line adjustments.

Description of drawings

Fig. 1 is the schematic diagram before and after the fusion of public transport stations of the present invention;

Fig. 2 is a schematic diagram of different station transfer of the present invention;

Fig. 3 is the figure of passenger flow on the bus stop of the present invention;

Fig. 4 is a figure of passenger flow at a bus stop of the present invention;

Fig. 5 is the top 15 stations of the whole day passenger flow of the present invention;

Fig. 6 is the line of the present invention with a daily passenger flow of over 10,000;

Fig. 7 is each line passenger flow of the present invention;

Fig. 8 is the top 15 stations of the passenger flow of the present invention;

Fig. 9 is the station of the present invention to change passenger flow spatial distribution;

Fig. 10 is a flow chart of the overall algorithm operation of the present invention.

Detailed ways

The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of the application, rather than all embodiments. Based on the embodiments of the application, those of ordinary skill in the art All other embodiments obtained under the premise of no creative work belong to the scope of protection of this application.

In this embodiment, a method for analyzing and mining bus data based on a one-ticket system of the present invention. Figures 3 and 4 show the passenger flow of bus stops, and the number of stations corresponding to the passenger flow is marked in the legend. From the perspective of spatial distribution, the stations with large passenger flow are concentrated in the eastern part of the city, indicating that the eastern part is the core area of the city; according to the distribution of passenger flow at the stations, there are 282 stations with more than 2000 passengers, and the number of passengers dropped off is 282. There are 174 stations with more than 2,000 passengers, and at the same time, the number of stations with less than 300 passengers and the number of stations with less than 300 passengers are both more than 1,800. There is an imbalance.

Figure 5 shows the top 15 stations with the highest passenger flow throughout the day. It can be seen from the figure that the daily passenger flow of these 15 stations exceeds 3,600, and they are important passenger flow distribution centers. The station with the largest passenger flow throughout the day is Beimen Bus The bus station has a daily passenger flow of 6,511 passengers, and the passenger flow near these stations is relatively large. Relevant optimizations for these stations will help improve the service level of public transport.

In this embodiment, the line operation status is as follows: based on the identification results of passengers getting on and off the bus site, the passenger flow of each line throughout the day can be obtained in the actual operation of the bus. Among them, the average daily passenger flow of Route 81 reached 41,650, which is 1.6 times that of Route 191 and four times that of Route 316. This shows that this line plays an important role in the bus network. On the other hand, By analyzing the passenger flow OD of this line in detail, adjusting the relevant lines and sharing some functions of the line can effectively improve the service level of the bus network.

The spatial distribution of passenger flow through each line is shown in Figure 7. The red lines indicate that the daily passenger flow exceeds 15,000. These lines are the backbone lines of the bus network. As shown in the figure, these lines are mainly used to communicate with the city. The core area and surrounding areas are mainly in the west of the city. On the other hand, the difference in passenger flow between the east and west of the city is relatively large, and the distribution of passenger flow is concentrated on one line. This is in line with the huge difference in passenger flow of different lines reflected in Figure 6. By adjusting the bus lines, the distribution of passenger flow is more uniform. It can effectively improve the service level of public transport under emergencies.

In this embodiment, the transfer station uses the above algorithm to identify the passenger flow of each bus station in Yinchuan City in one day. Figure 8 shows the 15 stations with the largest passenger flow, and the passenger flow at the North Gate Bus Station It has the largest volume, with a daily passenger flow of 1037. Its spatial distribution is shown in Figure 9. Stations with a large number of transfers are concentrated in the eastern core area of the city. Further investigation and research on these stations and adjustments to relevant lines can effectively reduce the number of transfer passengers. Improve passenger satisfaction.

The above embodiments are only used to illustrate the technical solutions of the present invention without limitation. Although the present invention has been described in detail with reference to preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced. Without departing from the spirit and scope of the technical solutions of the present invention, all of them should be included in the scope of the claims of the present invention. The technologies, shapes and construction parts not described in detail in the present invention are all known technologies.

Claims (2)

1. The analysis mining method of the public transport data of a ticket system is specifically implemented according to the following steps:

s1: cleaning public transport IC card data, public transport GPS data, vehicle-mounted machine data and one-way station information;

S _1.1 : the bus IC card data comprises a transaction card number, transaction time, a line and a license plate number, firstly, information of the transaction card number, the transaction time, the line and the license plate number is extracted, then useless fields except the bus IC card data are deleted, and records of partial field loss are deleted;

S _1.2 : a plurality of fields are shared in the vehicle-mounted GPS data of the bus, but part of the fields are not started at present and are null; extracting useful fields including vehicle-mounted machine number, arrival information, positioning time, positioning longitude and latitude, line number, sub-line number, sequence number, bus speed and bus driving mileage field, and deleting useless fields; then deleting records with longitude and latitude outside the range of the information acquisition area based on ArcGIS; deleting only the data which arrives at the station or only leaves the station;

S _1.3 : the vehicle-mounted machine information refers to a license plate number and a line name corresponding to the number of the vehicle-mounted machine and is used for matching the license plate number corresponding to the GPS data and associating the GPS data with the IC card dataFusing, namely extracting data of the number of the vehicle-mounted frame, the license plate number and the line name;

S _1.4 : the single-pass station relation comprises station sequence numbers, station names and station type data corresponding to line numbers and sub-line numbers, station information in ArcGIS data is combined, the longitude and latitude of different lines in different directions of the same station are fused to obtain unique longitude and latitude values of the station, GPS data is fused with vehicle-mounted machine information and the fused single-pass station relation to obtain GPS data containing license plate numbers, station names and station longitude and latitude to leave the station;

s2, deducing the passenger departure station point based on the probability of the departure station point;

wherein, the passenger card swiping data of the bus lacks information of the boarding and disembarking stations and transfer stations of passengers, and the missing data is calculated by the following method;

S _2.1 the passenger gets on the station to infer;

calling original passenger card swiping data UserData, bus GPS data Vehicles GPS and a license plate number list Vehicles; because of the errors of GPS positioning time and card swiping time, eliminating the data with the difference between the GPS positioning time and the card swiping time being more than 180 seconds, so as to ensure the accuracy of a matching result;

S _2.2 the passenger gets off the station to infer;

S _2.2.1 extracting all card swiping records with the same card number in the passenger card swiping records in one day, and sequencing according to card swiping time;

S _2.2.2 aiming at a passenger, firstly, according to the boarding station in the previous record of the passenger, acquiring all stations of the line taken by the passenger when the passenger travels;

S _2.2.3 calculating a station with the nearest space distance to all stations of the previous riding route in the next riding record of the passenger, wherein the station is a get-off station when the passenger takes the vehicle for the previous time;

S _2.2.4 when S is _2.2.2 When the card swiping information in the bus is the last card swiping record of the card number, the first card swiping record of the passenger is used as the record of the next passenger, so that the last bus taking is estimatedThe calculation of the stop point of the getting-off station of the card number is finished at the getting-off place of the car;

S _2.2.5 step S is continuously operated for all card numbers _2.2.1 —S _2.2.4 Until all the card numbers complete the deduction process;

for passengers who do not have continuous bus trips in one day, a passenger getting-off station point estimation model based on the getting-off probability of stations is applied, and the existing research shows that the attraction strength of bus stations is basically balanced with the occurrence strength, so that the attraction strength of stations is replaced by the equivalent occurrence strength of stations, the number of passengers getting on each station of any line can be counted according to the inferred result of the passenger getting-on station points, and the attraction strength of stations is calculated as shown in formula 2:

wherein s is _i Indicating the number of people getting on the ith station;

probability of getting off p of passenger _ij The number of stops and the attraction strength p of stops when the bus is on average _i In the related art, the number of bus stops for the resident bus travel is concentrated in a fixed traveling direction, and the number of bus stops approximately accords with poisson distribution, as shown in the formula 3:

z in _ij The probability of getting on/off the passenger's ith station and j station is represented; λ represents the average number of stops for bus travel, and when the number of stops after i stops is less than λ, λ= (n- λ), n is the total number of single-line bus stops, so that the probability that a passenger gets off from the stop point j on the stop i can be constructed as shown in formula 4:

so far, the total number of passengers getting on and off any station i and getting off station j is shown as a formula 5:

M _ij ＝s _i ×p _ij formula 5.

2. The method for analyzing and mining one-ticket bus data according to claim 1, wherein the method comprises the following steps: transfer station deducing, wherein the consumption formula 1 is expressed in the transfer process:

T _s ＝t ₃ -t ₂ ＝T _walk +T _wait ＝t ₃ -t ₁ -T _v 1 (1)

Wherein T is _walk The walking time from the station to the transfer station is the walking time from the station to the transfer station; t (T) _wait Waiting time for the transfer station; t (T) _v The last time the car was in;

S _3.1 analyzing bus taking time T _v Transfer walking time T _walk Transfer station waiting time T _wait The maximum time interval of transfer can be obtained, and the maximum possible transfer threshold value is 60min; the transfer identification process is as follows:

S _3.2 extracting a public transport IC card record, wherein the record card swiping time is recorded as t1, and the adjacent card swiping record with the same card number is obtained, and the record card swiping time is recorded as t2;

S _3.3 calculating the card swiping time interval Ti=t2-t 1, if Ti is less than or equal to Tmax, and the distance L between transfer stations<500m, considering the passenger to take the next trip as a transfer behavior, otherwise, considering the passenger to take the next trip;

S _3.4 judging all data of the same card number, and recording the identification result;

S _3.5 repeating step S _3.1 -S _3.3 And (5) completing the passenger transfer behavior identification until the last card.