CN108230724A - A kind of urban mass-transit system Vehicle station name announcing missing data method for repairing and mending based on maximum probability estimation - Google Patents

Abstract

Aiming at the problem of missing data in the urban bus station announcement system, the present invention provides a method for repairing the missing data of the urban bus station announcement system based on maximum probability estimation, which includes the following steps: by analyzing the characteristics of the missing data and the bus Combining the characteristics of passengers' card swiping behavior with historical data, a maximum probability estimation model with travel time as a set of posterior condition parameters was constructed, and the missing data of automatic station announcement equipment in the urban bus system was accurately inferred with the help of passenger card swiping data. The data repair method provided by the present invention overcomes the influence of the shadow effect of urban buildings in the prior art, and the automatic bus station reporting system is prone to positioning failure or communication packet loss, resulting in incomplete station reporting information and seriously affecting the subsequent data mining effect It improves the success rate of passenger bus travel trajectory restoration and increases the information validity of urban bus passenger flow distribution research.

Description

A missing data repair method for urban public transport system based on maximum probability estimation Complementary method

technical field

The invention relates to the technical field of intelligent public transportation, and relates to a method for repairing missing data of on-vehicle call stations in urban public transportation systems based on maximum probability estimation.

Background technique

The distribution data of bus passenger flow is the basis for optimizing the operation and management of urban public transport, the key to restricting the level of urban public transport management, and the focus of urban public transport researchers. For a long time, managers have generally obtained limited bus travel data through sampling surveys, which is time-consuming, laborious and costly. In recent years, with the widespread application of Automated FareCollection (AFC) and Automated Vehicle Locating (AVL), urban public transport systems have accumulated a large amount of operational management data. How to obtain bus passenger flow information from the existing data of the bus system at low cost through data mining technology has become the focus of urban bus researchers at home and abroad.

In the relevant research on obtaining bus passenger flow, the most critical link is to convert the passenger card swiping data into the passenger's movement trajectory in the bus system. However, the existing bus card swiping system often focuses on the operation and sorting function, and ignores the recording of the card swiping station, especially the most widely used one-ticket bus at present, which can only record the passenger boarding time and bus number, Line number, no information about getting on and off the bus. Therefore, in the process of data processing, it is often necessary to first match the card swiping records with the station reporting records of the GPS driving record system to infer the bus passengers’ swiping card boarding stations, and then infer passengers based on the shortest distance transfer assumption and continuous travel chain assumption. Finally, for the card swiping records where the drop-off site cannot be found, the drop-off site is estimated by the site attraction right method.

The above data processing process relies heavily on the vehicle station reporting data. However, the vehicle automatic station reporting system relies heavily on the GPS module and 2G mobile communication module to obtain and upload location information. This type of equipment is seriously affected by the shadow effect of urban buildings, and is prone to positioning failure or communication failure. Packets are lost, resulting in incomplete station reporting information. By analyzing the AVL station reporting data, shift scheduling data and AFC credit card data of a certain city, it is found that the average missing data in the city accounts for 6.25% during the statistical period, and the areas with high data missing rates are located in densely populated areas of the city. Although the data missing rate is less than 10%, the subsequent matching results of bus passengers' boarding stations show that more than 25% of the card swiping data cannot be matched to the boarding stations due to missing data. It can be seen that the lack of AVL station reporting data that is common in urban public transport systems has significantly affected the process of passenger travel trajectory restoration and seriously restricted the mining effect of bus passenger flow.

Contents of the invention

The purpose of the present invention is to provide a method for repairing the missing data of the urban public transport system based on the estimation of the maximum probability for the ubiquitous data loss situation in the automatic station announcement system of the urban public transport. By analyzing the characteristics of the missing data and the characteristics of bus passengers' credit card swiping behavior, this method constructs a maximum probability estimation model with travel time as the posterior condition parameter set, and accurately infers the missing on-board station announcement data of the urban bus system. Data mining provides the necessary support.

In order to achieve the above object, the present invention adopts the following technical solutions:

A method for repairing missing data of vehicle-mounted station announcements in urban public transport systems based on maximum probability estimation of the present invention comprises the following steps:

S1. According to each departure, classify and organize the data source 2 "AVL vehicle station announcement data", and arrange the AVL station announcement data of each departure according to the station sequence of the bus line: {S ₀ , S ₁ , S ₂ ,… ,S _i }, where S _i corresponds to the AVL information of the vehicle arriving at the S _i site;

S2. Comparing the departure data of a certain time in step S1 with the sequence of stops of the bus line corresponding to the data source 1 "bus scheduling data";

S21. If the site information is complete, there is no missing data, and there is no need to repair;

S22. If the site information is incomplete, it means that there is a lack of AVL data, lock the site interval S _l = {S ₀ , S _L0 , S _L1 ,...,S _Lk ,...,S ₁ } with data loss, and calculate The travel time out of the missing data interval, denoted as t _TRIP ;

S3. According to the "passenger AFC card swiping data" provided by data source 3, query the card swiping data of passengers taking the bus on this journey from S ₀ to S _i , and use the threshold segmentation method to extract the first station where passengers swiped their cards to board the bus The card-swiping time stamp of a card-swiping event, where the card-swiping moment of the first boarding passenger at S _Lk station is recorded as

S4, with the help of historical data, calculate the travel time probability distribution function from S ₀ to S _L0 , S _L1 ,...,S _Lk satisfying the control condition Θ: Preferably, Θ=t _TRIP ∈[t _TRIP -0.1×t _TRIP ,t _TRIP +0.1×t _TRIP ];

S5. The card swiping time stamp of the first card swiping event of each station S _Lk in the itinerary of S _l Subtract the outbound time of S ₀ to obtain the travel time of each station. For any travel time, substitute the time distribution function of each station obtained by S4 to obtain the probability value, and the station corresponding to the function that outputs the maximum value is The station where the vehicle arrives at the moment, and the data inference is completed.

As a preferred technical solution, the data source 1 "bus scheduling data" includes: a) shift number; b) license plate number; c) line number; d) departure time; e) station sequence;

The data source 2 "AVL vehicle station reporting data" includes: a) license plate number; b) line number; c) station name; d) entry time; e) exit time;

The data source 3 "passenger AFC swiping card data" includes: a) license plate number; b) line number; c) station name; d) card swiping time; e) passenger identification; f) station matching status.

As a preferred technical solution, step S22 is specifically:

By comparing the AVL station reporting data of the current train number with the complete bus line station sequence, lock the station interval with data loss, and mark the missing station number, assuming S _l ={S ₀ ,S _L0 ,S _L1 ,…,S _Lk ,...,S ₁ }, is a section of station intervals that vehicles pass through continuously during this departure, where S ₀ and S ₁ are stations with complete data, S _L0 , S _L1 ,...,S _Lk are S ₀ and S ₁ Stations with missing station reporting data.

As a preferred technical solution, step S3 is specifically:

Since passenger card swiping is a short-interval continuous event, the time threshold can be set as a reference for segmenting adjacent stations S _i , S _i+1 passengers, that is, if the interval between adjacent card swiping records is less than the set time, the adjacent card swiping The record is regarded as the card swiping record at the same station; otherwise, the last card swiping record is the first card swiping record at the next station.

As a preferred technical solution, in step S4, the historical data is historical AVL data of not less than four weeks.

As a preferred technical solution, in step S4, at an unknown station S _Lk where there are passengers boarding the bus, the time stamp of swiping the card generated by the first passenger who boards the bus is Then for this frequency, the arrival time of the bus at this station can be equivalent to

As a preferred technical solution, in step S4, it is assumed that Follow a Gaussian distribution:

Then, after Θ is determined, all trips from site S ₀ to site S _LK in the historical AVL data that satisfy the control condition Θ can be queried, and the mean value μ and standard deviation σ of the model parameter time can be calculated.

Compared with the prior art, the present invention has the following advantages and beneficial effects:

1. The present invention constructs a maximum probability estimation model with travel time as a set of posterior condition parameters by analyzing the characteristics of missing data and the card swiping behavior of bus passengers, and fully utilizes the advantages of massive data, especially suitable for occasions with huge sample sizes .

2. The method adopted in the present invention is to construct a station set S _t = {S ₀ , S _L0 , S _L1 ,..., S _LK , S ₁ } that a certain operating vehicle passes through continuously in one trip, where S _L0 , S _L1 ,...,S _LK are sites with missing data between the data complete sites S ₀ and S ₁ , which can be applied to the repair of multiple consecutive missing sites with high repair efficiency.

3. Compared with the traditional method, the data restoration accuracy rate of the present invention is high, and the restoration accuracy rate of more than 85% can be maintained in the case of less than or equal to 8 consecutive missing sites.

Description of drawings

Fig. 1 is the overall working flow chart of the method for repairing missing AVL data of the public transportation system in this embodiment.

FIG. 2 shows the success rate of data restoration under different post-verification conditions in this embodiment.

Detailed ways

The present invention will be further described in detail below in conjunction with the embodiments and the accompanying drawings, but the embodiments of the present invention are not limited thereto.

Example

Taking the bus data of a certain city as an example, extract the AVL vehicle station announcement data on March 5, 2017 as the basic data source, scan the station announcement data of each vehicle in the AVL station announcement data, and randomly select the starting station with a probability of 15%. After taking out the AVL station reporting data of the 8 stations and adding them to the candidate test set, a total of 94,750 pieces of test data were obtained. For each ratio of data in the candidate test set, uniformly distributed random interference (range: 0-120s) is added to all station reporting records except the first and last station reporting records (S ₀ , S ₁ ) when entering the station, get the test set. The key information fields contained in each test data in the test set are shown in Table 1:

Table 1 Key information fields of test data

Therefore, the missing data inference task is to scan all records in the test set, and according to the first and last station reporting records (S ₀ , S ₁ ) and timestamps T _L0 ',T _L1 ',…,T _Lk containing disturbances 'Deduce S _L0 , S _L1 ,...,S _Lk .

The present invention provides a method for repairing AVL missing data in public transport systems based on maximum probability estimation, the process of which is as shown in Figure 1, comprising the following steps:

The first step is to extract the AVL data of each departure and sort them in chronological order.

The second step is to compare the station sequence of the line. 1) If the station information is complete, there is no data loss and no need to repair; 2) if the station information is incomplete, it means that there is AVL data missing and needs to be repaired. For the situation that needs to be repaired, construct S _l = {S ₀ , S _L0 , S _L1 ,...,S _Lk ,...,S ₁ } as the station intervals that pass continuously in this trip, where S ₀ and S ₁ are data The complete stations, S _L0 , S _L1 ,..., S _Lk are stations with missing station reporting data between S ₀ and S ₁ .

In the third step, according to the "passenger AFC card swiping data" provided by data source 3, query the card swiping data of passengers taking the bus from S ₀ to S _i on this journey, and use the threshold segmentation method to extract the stations where passengers swiped their cards to board the bus The card swiping time stamp of the first card swiping event in , where the card swiping time of the first boarding passenger at S _Lk station is recorded as

The fourth step is to use historical data to calculate the travel time probability distribution function from S ₀ to S _L0 , S _L1 ,...,S _Lk satisfying the control condition Θ: Θ＝t _TRIP ∈[t _TRIP -0.1×t _TRIP ,t _TRIP +0.1×t _TRIP ] is the optimal scheme after testing, and can be flexibly modified according to the actual situation; therefore Θ can be other control conditions.

The fifth step, the time stamp of the first card swiping event of each station S _Lk in the itinerary of S _l Subtract the outbound time of S ₀ to obtain the travel time of each station. For any travel time, substitute the time distribution function of each station obtained by S4 to obtain the probability value, and the station corresponding to the function that outputs the maximum value is the time The station where the vehicle arrives, and the data inference is completed.

The performance comparison of data patching models constructed with different control conditions is shown in Figure 2. The present invention uses a maximum probability estimation model based on travel time, relative to the prior probability model, and other probability models based on the start time. The optimal scheme provided has the best data repair effect, and maintains a repair accuracy of more than 85% when the number of consecutive missing stations is less than or equal to 8.

The above-mentioned embodiment is a preferred embodiment of the present invention, but the embodiment of the present invention is not limited by the above-mentioned embodiment, and any other changes, modifications, substitutions, combinations, Simplifications should be equivalent replacement methods, and all are included in the protection scope of the present invention.

Claims (7)

1. a method for repairing the missing data of the city public transport system vehicle-mounted reporting station based on maximum probability estimation, is characterized in that, comprises the following steps: S1. According to each departure, classify and organize the data source 2 "AVL vehicle station announcement data", and arrange the AVL station announcement data of each departure according to the station sequence of the bus line: {S ₀ , S ₁ , S ₂ ,… ,S _i }, where S _i corresponds to the AVL information of the vehicle arriving at the S _i site; S2. Comparing the departure data of a certain time in step S1 with the sequence of stops of the bus line corresponding to the data source 1 "bus scheduling data"; S21. If the site information is complete, there is no missing data, and there is no need to repair; S22. If the site information is incomplete, it means that there is a lack of AVL data, lock the site interval S _l = {S ₀ , S _L0 , S _L1 ,...,S _Lk ,...,S ₁ } with data loss, and calculate The travel time out of the missing data interval, denoted as t _TRIP ; S3. According to the "passenger AFC card swiping data" provided by data source 3, query the card swiping data of passengers taking the bus on this journey from S ₀ to S _i , and use the threshold segmentation method to extract the first station where passengers swiped their cards to board the bus The card-swiping time stamp of a card-swiping event, in any station S _Lk with missing data, the card-swiping moment of the first boarding passenger is recorded as S4, with the help of historical data, calculate the travel time probability distribution function from S ₀ to S _L0 , S _L1 ,...,S _Lk satisfying the control condition Θ: Among them, Θ＝t _TRIP ∈[t _TRIP -0.1×t _TRIP ,t _TRIP +0.1×t _TRIP ]; S5. The card swiping time stamp of the first card swiping event of each station S _Lk in the itinerary of S _l Subtract the outbound time of S ₀ to obtain the travel time of each station. For any travel time, substitute the time distribution function of each station obtained by S4 to obtain the probability value, and the station corresponding to the function that outputs the maximum value is The station where the vehicle arrives at the moment, and the data inference is completed. 2. according to the described method of repairing the missing data of the urban public transport system vehicle-mounted station reporting system based on the maximum probability estimation according to claim 1, it is characterized in that, said data source 1 " bus dispatching data " comprises: a) shift number; b) license plate number ; c) line number; d) departure time; e) station sequence; The data source 2 "AVL vehicle station reporting data" includes: a) license plate number; b) line number; c) station name; d) entry time; e) exit time; The data source 3 "passenger AFC swiping card data" includes: a) license plate number; b) line number; c) station name; d) card swiping time; e) passenger identification; f) station matching status. 3. according to claim 1, the method for repairing the missing data of the city public transport system vehicle-mounted station reporting system based on maximum probability estimation, is characterized in that, step S22 is specifically: By comparing the AVL station reporting data of the current train number with the complete bus line station sequence, lock the station interval with data loss, and mark the missing station number, assuming S _l ={S ₀ ,S _L0 ,S _L1 ,…,S _Lk ,…,S ₁ }, is a section of stations that vehicles pass through continuously during this departure, where S ₀ and S ₁ are stations with complete data, S _L0 , S _L1 ,…,S _Lk are stations between S ₀ and S1 Stations with missing station reporting data. 4. according to claim 1 based on the urban public transport system vehicle-mounted report station missing data repairing method estimated based on maximum probability, it is characterized in that, step S3 is specifically: Since passenger card swiping is a short-interval continuous event, the time threshold can be set as a reference for segmenting adjacent stations S _i , S _i+1 passengers, that is, if the interval between adjacent card swiping records is less than the set time, the adjacent card swiping The record is regarded as the card swiping record at the same station; otherwise, the last card swiping record is the first card swiping record at the next station. 5. according to claim 1, the method for repairing the missing data of the vehicle-mounted station announcement of the urban public transport system based on the maximum probability estimation, is characterized in that, in step S4, the historical data is the historical AVL data of no less than four weeks. 6. according to claim 1, the method for repairing the missing data of the vehicle-mounted station announcement of the urban public transport system based on the maximum probability estimation, is characterized in that, in step S4, a certain unknown site S _Lk where there are passengers on the bus, the first swipe card to get on the bus The time stamp of card swiping generated by passengers is Then for this frequency, the arrival time of the bus at this station can be equivalent to 7. according to claim 1, the missing data repairing method based on the urban public transport system vehicle-mounted station announcement based on maximum probability estimation, is characterized in that, in step S4, assume Follow a Gaussian distribution: Then, after Θ is determined, all trips from site S ₀ to site S _LK in the historical AVL data that satisfy the control condition Θ can be queried, and the mean value μ and standard deviation σ of the model parameter time can be calculated.