CN117314504B - Public transportation passenger flow prediction method and system - Google Patents

Public transportation passenger flow prediction method and system Download PDF

Info

Publication number
CN117314504B
CN117314504B CN202311595226.6A CN202311595226A CN117314504B CN 117314504 B CN117314504 B CN 117314504B CN 202311595226 A CN202311595226 A CN 202311595226A CN 117314504 B CN117314504 B CN 117314504B
Authority
CN
China
Prior art keywords
passenger flow
date
passenger
time
period
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202311595226.6A
Other languages
Chinese (zh)
Other versions
CN117314504A (en
Inventor
张世强
邵刚
张光磊
孙宏飞
钱贵涛
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hualu Zhida Technology Co Ltd
Original Assignee
Hualu Zhida Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hualu Zhida Technology Co Ltd filed Critical Hualu Zhida Technology Co Ltd
Priority to CN202311595226.6A priority Critical patent/CN117314504B/en
Publication of CN117314504A publication Critical patent/CN117314504A/en
Application granted granted Critical
Publication of CN117314504B publication Critical patent/CN117314504B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data
    • G06Q30/0202—Market predictions or forecasting for commercial activities
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services

Abstract

The invention relates to the technical field of public transportation passenger flow prediction, and provides a public transportation passenger flow prediction method and a public transportation passenger flow prediction system, wherein the public transportation passenger flow prediction method comprises the following steps: acquiring historical passenger flow data of a certain station of a certain line, wherein the historical passenger flow data comprise the arrival time of vehicles and the number of passengers on the vehicle; date classification is carried out according to the historical passenger flow data change, so that the passenger flows on the same kind of date are distributed similarly throughout the day; assuming that the passenger flows of the sites with the same date are distributed identically throughout the day, dividing the passenger flows of each class of date according to a certain principle, and taking the passenger flows of the same date in the same time period as a certain value; establishing an equality relation between the passenger flow of each period and the real passenger flow according to the relation between the historical passenger flow data and the period division; obtaining estimated values of passenger flows in each period by solving a linear equation set formed by the equation relation, so as to obtain the whole-day distribution estimation of the passenger flows in each type of date; and predicting the passenger flow of the line at the station in any period of the future date according to the category to which the future date belongs.

Description

Public transportation passenger flow prediction method and system
Technical Field
The invention relates to the technical field of public transportation passenger flow prediction, in particular to a public transportation passenger flow prediction method and a public transportation passenger flow prediction system.
Background
Accurate estimation and prediction of passenger flow has a very important role for public transportation services. Along with diversification of the travel modes of passengers and acceleration of the urban process, the conventional public transportation service for protecting the fixed lines, the fixed stations and the fixed departure intervals to be required is more and more difficult to meet the personalized demands of the travel of the passengers. The continuous change of the urban process makes the passenger flow distribution rule referenced in early line planning no longer applicable, often leads to irregular planned line passenger flow distribution, insufficient passenger transportation resource utilization, lower passenger satisfaction and high comprehensive operation cost.
In recent years, the public transportation service industry faces the digital transfer type road from the fixed demand to the fixed demand, dynamically adjusts the input transportation capacity resources according to the individual demands or the space-time regularity demands of the traveling of passengers, provides moderate and sufficient service transportation capacity, improves the utilization efficiency of the transportation capacity resources to the maximum extent on the basis of meeting the overall satisfaction of the passengers, and saves the comprehensive operation cost. In order to provide more accurate personalized public transportation service, the demand information of the passenger travel needs to be acquired, and the passenger flow is the most basic passenger travel demand. According to passenger flow information, corresponding vehicle resources can be scheduled to meet the travel demands of passengers as much as possible at the lowest cost.
The traditional method for counting the passenger flow is to record and sort the number of passengers getting on and off the bus manually, and then roughly estimate and judge the passenger flow according to data summarizing analysis, so that the estimated and predicted accuracy of the obtained passenger flow is low, the required manual recording cost is high, and the line coverage rate and the time coverage rate are low.
Along with the improvement of the informatization intelligent degree of the vehicle-mounted terminal equipment and the ticket terminal equipment with vehicle positioning data, the boarding and alighting time and positioning information of passengers can be accurately acquired, the complexity and uncertainty of manual recording can be greatly reduced, all lines and all vehicles can be covered, and passenger flow data can be accurately estimated and predicted. However, in the prior art, most of black box modeling methods adopting feature extraction such as machine learning, deep learning and the like and model training lack of utilization of physical mechanisms and interpretability of models, the applicability range of the models obtained by learning is narrow, the modeling process is too complex, the complexity of the models is high, and the robustness is poor.
On the other hand, the estimation and prediction methods of the passenger flow in the prior art are mostly based on fixed time interval division modes (divided according to hours, days, weeks and the like), and lack of modeling and prediction methods of finer and more flexible time interval division modes, and passenger flow estimation and prediction results with finer time scales are needed for vehicle scheduling, so that hidden danger is brought to resource matching scheduling of later use of the passenger flow prediction values.
Disclosure of Invention
The invention mainly solves the technical problems that the passenger flow prediction modeling process is too complex, the time scale of passenger flow prediction is single and the like in the prior art, and provides a public transportation passenger flow prediction method and system.
The invention provides a public transportation passenger flow prediction method, which comprises the following steps:
acquiring historical passenger flow data of a certain station of a certain line; the historical passenger flow data at least comprises the arrival time of vehicles and the number of passengers getting on;
date classification is carried out according to the historical passenger flow data change, so that the passenger flows on the same kind of date are distributed similarly throughout the day;
assuming that the passenger flows of the sites with the same date are distributed identically throughout the day, dividing the passenger flows of each class of date according to a certain principle, and taking the passenger flows of the same date in the same time period as a certain value;
establishing an equality relation between the passenger flow of each period and the real passenger flow according to the relation between the historical passenger flow data and the period division;
obtaining estimated values of passenger flows in each period by solving an equation set formed by the equation relation, so as to obtain the whole-day distribution estimation of the passenger flows in each type of date;
predicting the passenger flow of the line at the station in any period of the future date according to the category to which the future date belongs;
the date classification is carried out according to the change of the historical passenger flow data, and the date classification mode comprises the following steps: classifying according to the passenger flow change rule, wherein the passenger flow change rule with the same passenger flow change rule is the same kind of date;
the classification according to the passenger flow change rule comprises the following steps:
acquiring historical passenger flow data of a certain station of a certain line every day;
since the vehicle arrival time per day is uncertain, the whole day passenger flow distribution per day needs to be obtained through interpolation fitting; the daily total-day passenger flow distribution is expressed as a passenger flow time-varying curve between the first and last buses of the station and the station time, and the curve is sampled at fixed time intervals to obtain a daily passenger flow vector;
and defining the distance between the vectors according to the similarity between the daily passenger flow vectors, and clustering the daily passenger flow vectors by using the distance, so that the distance between the passenger flow vectors of the same class is smaller, and the distance between the passenger flow vectors of different classes is larger.
Preferably, the historical passenger flow data is collected by one or more modes of bus-mounted equipment, platform monitoring equipment and manual recording.
Preferably, in the date classification mode, in order to predict a passenger flow category to which a certain date belongs in the future, a relationship between a date attribute and the passenger flow category is established, which comprises the following procedures:
firstly, determining date attribute characteristics;
acquiring attribute characteristics of the date corresponding to the historical passenger flow data to form a characteristic vector of the date;
acquiring a passenger flow category of a date corresponding to the historical passenger flow data, and taking the passenger flow category as a category label of the date;
training a machine learning classification model according to the feature vector and the class label of each date;
and judging the passenger flow category to which the date belongs according to the classification model and the attribute characteristics of the date.
Preferably, the passenger flow of each type of date is divided into time intervals according to a certain principle, including but not limited to one of the following division modes:
one of the time division modes is as follows: equally-spaced division is performed according to fixed intervals;
the second time interval division mode is as follows: the passenger flow is divided at unequal intervals according to the size of the passenger flow, the time period of large passenger flow can be relatively smaller, and the time period of small passenger flow can be properly larger.
Preferably, the equation relationship between the passenger flow of each period and the real passenger flow is established according to the relationship between the historical passenger flow data and the period division, and the equation relationship is as follows:
the sum of the accumulated passenger flows of the relevant time periods in the arrival time of two adjacent public transportation vehicles=the number of boarding persons when the public transportation vehicles arrive at the station one time after the two adjacent public transportation vehicles.
Preferably, the obtaining the estimated value of the passenger flow in each period by solving the linear equation set formed by the equation relation includes: and obtaining an optimal solution with the minimum mean square error by using least square estimation.
Preferably, the predicting the passenger flow of the line at the station in any period of the future day according to the category to which the future date belongs includes:
acquiring date types of future dates, wherein each type of date type has the same all-day passenger flow distribution;
acquiring passenger flow estimated values of all time periods corresponding to the date type;
and calculating the number of passengers in the time period of the future date according to the relation between the time period of the future date and the time period division of the date type.
Correspondingly, the invention also provides a public transport passenger flow prediction system, which comprises:
the data acquisition module is used for acquiring historical passenger flow data of a certain station of a certain line; the historical passenger flow data at least comprises the arrival time of vehicles and the number of passengers getting on;
the date classification module is used for classifying the dates according to the change of the historical passenger flow data so that the passenger flows on the same kind of date are distributed similarly all the day;
the time interval dividing module is used for dividing the time intervals of the passenger flows of each class of date according to a certain principle on the assumption that the passenger flows of the sites with the same class of date are distributed identically throughout the day, and the passenger flows of the same class of date in the same time interval are a certain value;
the passenger flow modeling module is used for establishing an equality relation between the passenger flow of each time period and the real passenger flow according to the relation between the historical passenger flow data and the time period division;
the passenger flow estimation module is used for obtaining the estimated value of the passenger flow in each period by solving an equation set formed by the equation relation, so as to obtain the whole-day distribution estimation of the passenger flow in each type of date;
the passenger flow prediction module is used for predicting the number of passengers of the line at the station in any time period of a future day according to the category to which the future date belongs;
the date classification is carried out according to the change of the historical passenger flow data, and the date classification mode comprises the following steps: classifying according to the passenger flow change rule, wherein the passenger flow change rule with the same passenger flow change rule is the same kind of date;
the classification according to the passenger flow change rule comprises the following steps:
acquiring historical passenger flow data of a certain station of a certain line every day;
since the vehicle arrival time per day is uncertain, the whole day passenger flow distribution per day needs to be obtained through interpolation fitting; the daily total-day passenger flow distribution is expressed as a passenger flow time-varying curve between the first and last buses of the station and the station time, and the curve is sampled at fixed time intervals to obtain a daily passenger flow vector;
and defining the distance between the vectors according to the similarity between the daily passenger flow vectors, and clustering the daily passenger flow vectors by using the distance, so that the distance between the passenger flow vectors of the same class is smaller, and the distance between the passenger flow vectors of different classes is larger.
According to the public transportation passenger flow prediction method and system provided by the invention, for a specific station of a specific public transportation line, the passenger flow in a specific time period is assumed to be kept constant, and the passenger flow estimated value of each time period is solved by establishing an equation relation between the passenger flow in each time period and the passenger flow data in actual historical data and further the passenger flow number of the same station in a certain time period in the future is predicted. The whole process only utilizes the solution of the linear equation set, the calculation complexity is much lower than that of the prior art, and the time interval dividing mode can be flexibly customized according to different scene requirements. The public transportation field of the invention comprises all public transportation modes of people carrying such as highway, railway, civil aviation, water transportation and the like, and the application range is wide.
Drawings
FIG. 1 is a flow chart of an implementation of a public transportation passenger flow prediction method provided by the invention;
FIG. 2 is a flow chart for implementing automatic clustering according to the change rule of passenger flow;
FIG. 3 is a flow chart of an implementation of the relationship between the date attribute and the passenger flow category provided by the present invention;
fig. 4 is a schematic block diagram of a public transportation passenger flow prediction system provided by the invention.
Detailed Description
In order to make the technical problems solved by the invention, the technical scheme adopted and the technical effects achieved clearer, the invention is further described in detail below with reference to the accompanying drawings and the embodiments. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting thereof. It should be further noted that, for convenience of description, only some, but not all of the matters related to the present invention are shown in the accompanying drawings.
Example 1
As shown in fig. 1, the public transportation passenger flow prediction method provided by the embodiment of the invention includes:
step 1, acquiring historical passenger flow data of a certain station of a certain line.
The historical passenger flow data at least comprises the arrival time of the vehicle and the number of passengers on the vehicle. The historical passenger flow data is collected by one or more modes of bus-mounted equipment, platform monitoring equipment and manual recording.
And step 2, classifying dates according to the change of the historical passenger flow data, so that the passenger flows on the same kind of dates are similar in all-day distribution.
The invention is based on the basic assumption that the arrival rate (passenger increment) or number of passengers on a line at a particular station at a particular time period per day is the same, i.e. a fixed constant value, for a particular day of a particular day type. This is generally determined by the nature of the route setup, and passengers (e.g., commuting to work or from work) taking long-term fixed-point rides have a relatively high degree of regularity. If the line is established at the initial stage or the temporary change or other public travel alternative infrastructures are established (such as a new subway is opened), the passenger population and the riding time are not fixed, and the method of the invention is not applicable.
The method comprises the steps of acquiring date attributes and corresponding passenger flow distribution categories, wherein the passenger flow distribution on the same category date is assumed to be the same. The date classification based on historical passenger flow data changes includes, but is not limited to, one of the following classification schemes:
date classification mode one: sorting by time interval between dates, such as by season or month;
date classification mode II: classifying according to the relativity between dates, such as workdays or holidays, further dividing into workdays, etc.;
date classification mode three: according to date related attribute characteristics, such as normal weather and extreme weather, whether major activities or traffic accidents exist or not, and the like;
date classification mode four: classifying according to the passenger flow change rule, wherein the passenger flow change rule is the same kind of date.
Specifically, the invention can divide the date type into typical date types corresponding to working days, legal holidays and four seasons of spring, autumn and winter, further can be further subdivided into common working days, first working day, last working day and the like, and can further be further subdivided by considering the influence of the weather type on traveling. The dates can be automatically clustered according to the change rule of the passenger flow according to the change rule of the historical data, and the dates with similar change rules of the passenger flow are the same type of dates. Date attribute features may be season, month, week, weather, workday or holiday, whether there is a major traffic accident, etc.
As shown in fig. 2, in the fourth date classification mode, the classification according to the passenger flow change rule includes the following steps:
acquiring historical passenger flow data of a certain route and a certain station every day, namely the arrival time of vehicles and the number of passengers getting on the vehicle;
since the vehicle arrival time per day is uncertain, the whole day passenger flow distribution per day needs to be obtained through interpolation fitting; the daily total-day passenger flow distribution is expressed as a passenger flow time-varying curve between the first and last buses of the station and the station time, and the curve is sampled at fixed time intervals to obtain a daily passenger flow vector;
distances between vectors, such as euclidean distance or cosine similarity distance, are defined according to the similarity between the daily passenger vectors, and the distance is used for clustering the daily passenger vectors, so that the distances between the similar passenger vectors are smaller, and the distances between the inter-class passenger vectors are larger.
As shown in fig. 3, in the fourth date classification method, in order to predict a passenger flow category to which a certain date belongs in the future, a relationship between a date attribute and the passenger flow category is established, which includes the following steps:
firstly, determining date attribute characteristics such as the year, month, season, working day or not of the date;
acquiring attribute characteristics of the date corresponding to the historical passenger flow data to form a characteristic vector of the date;
acquiring a passenger flow category of a date corresponding to the historical passenger flow data, and taking the passenger flow category as a category label of the date;
training a machine learning classification model according to the feature vector and the class label of each date;
and judging the passenger flow category to which the date belongs according to the classification model and the attribute characteristics of the date.
And step 3, assuming that the passenger flows of the sites with the same date are distributed identically throughout the day, and dividing the passenger flows of each type of date according to a certain principle, wherein the passenger flows of the same date in the same time are given a certain value.
In the step, time division is carried out on each type of date, and passenger flows are the same in the same time period of the same kind of date. The passenger flow of each type of date is divided into time intervals according to a certain principle, including but not limited to one of the following dividing modes:
one of the time division modes is as follows: equally-spaced division is performed according to fixed intervals;
the second time interval division mode is as follows: the passenger flow is divided at unequal intervals according to the size of the passenger flow, the time period of large passenger flow can be relatively smaller, and the time period of small passenger flow can be properly larger.
In the invention, the specific time period of each day can be simply divided into early peak, late peak and peaked time period, and the division can be further refined according to the change rule of actual historical data. Or an extreme fine division method such as division into one period per minute is employed. The invention adopts the linear least square method to estimate the number of passengers in each time period, and only the increase of parameter variables is brought by fine division, so that the influence on the calculation complexity is not great.
The length of the time period can be divided into a time period as long as the historical passenger flow in the time period is relatively stable, so that the modeling process is flexible and convenient, and the number of unknown variables is reduced to the greatest extent. If the passenger arrival rate for each time period is defined as an unknown quantity, the number of passengers flowing for the time period for the non-uniform division can be obtained by multiplying the passenger arrival rate for the time period by the time period length.
In addition, if the real-time arrival data of the passengers can be obtained according to the passenger flow counter device installed on the platform in addition to the historical data of the passengers, the accurate passenger flow rate can be further obtained, and whether the real-time accurate fine particle data show periodic variation rules in time and space or not can also be considered according to the specific data analysis result.
And 4, establishing an equality relation between the passenger flow of each period and the real passenger flow according to the relation between the historical passenger flow data and the period division.
According to the relation between the historical passenger flow data and the time interval division, establishing an equality relation between the passenger flow of each time interval and the real passenger flow, wherein the equality relation is as follows:
the sum of the accumulated passenger flows of the relevant time periods in the arrival time of two adjacent public transportation vehicles=the number of boarding persons when the public transportation vehicles arrive at the station one time after the two adjacent public transportation vehicles.
And 5, obtaining estimated values of the passenger flows in each period by solving an equation set formed by the equation relation, so as to obtain the whole-day distribution estimation of the passenger flows in each type of date.
The linear equation system formed by solving the equation relation is used for obtaining the estimated value of the passenger flow in each period, and the estimated value comprises one of the following solving methods:
linear equation set solving method one: the Gaussian elimination method is suitable for the case that the unique cube number is equal to the unknown number;
and a second solving method of a linear equation set: and converting into a neural network model, and carrying out iterative solution by using a backward propagation algorithm.
And a linear equation set solving method III: and obtaining an optimal solution with the minimum mean square error by using least square estimation. Specifically, the least square algorithm is utilized to estimate and obtain the optimal solution with the minimum mean square error, which comprises the following steps:
the following general form of a linear equation system of the number of passengers is established:
(1)
let coefficient matrixWherein each coefficient in coefficient matrix ARepresent the firstIn equation (h)The coefficient of the number of unknowns,，is a general representation of the least squares problem. Specific each coefficientSources in practical application: equation (1) coefficient before unknownIn practical application, the relation between the time division and the practical passenger flow data time range is dependent. In general, the time range represented by each actual passenger flow data covers a plurality of time periods, and if the time period covers the whole time period, the coefficient in front of the passenger flow volume of the corresponding time period is 1, and if the time period only covers a part of the time period, the coefficient in front of the passenger flow volume of the corresponding time period is set according to the ratio of the coverage.
Vector of number of accumulated people in arrival at bus in this periodTime-phased arrival people number vectorObtaining an equivalent equation set of the linear equation set (1) of the number of passengers:
(2)
the least squares solution of the set of equivalent equations is in the form of:
(3)
wherein,andrepresenting the transpose and the inverse of the matrix respectively,is thatIs the least mean square error estimate.
And step 6, predicting the passenger flow of the line at the station in any period of the future day according to the category to which the future date belongs.
Predicting the passenger flow of the line at the station in any period of a day according to the category to which the future date belongs, including:
acquiring date types of future dates, wherein each type of date type has the same all-day passenger flow distribution;
acquiring passenger flow estimated values of all time periods corresponding to the date type;
and calculating the number of passengers in the time period of the future date according to the relation between the time period of the future date and the time period division of the date type.
After the invention obtains the whole-day distribution estimation of the passenger flow of each type of date, the passenger flow of the line at the station can be predicted in any time period of a future day. Firstly, selecting a date type to which a future date belongs (if the date type is generated by clustering, judging the date type to which the date belongs according to the date attribute characteristics and the established passenger flow distribution machine learning model), and then carrying out accumulated summation on the passenger flow rate estimated value of the station in the period to obtain the passenger flow quantity of the station in a certain period in the corresponding future.
As a preferred mode of the present invention: obtaining the daily distribution estimation of passenger flows on each type of date, and then further comprising:
calculating fitting errors of historical passenger flow data according to passenger flow estimated values of passenger flows in each period;
obtaining a covariance matrix of passenger flow estimation in each period according to a least square method;
according to the fitting error size and distribution condition of the historical passenger flow data, the date type and time division can be corrected so that the fitting error mean value of the historical passenger flow data is close to zero, and the variance is smaller;
according to the covariance matrix of the passenger flow estimation of each period, the prediction error limit of the passenger flow prediction value of a certain period in the future can be calculated, specifically, according to the linear equation relation between the passenger flow prediction value of the period in the future and the passenger flow estimation of each period and the covariance matrix of the passenger flow estimation of each period, the variance and standard deviation of the passenger flow prediction of the period in the future can be obtained, and then the error limit of the passenger flow prediction value is given according to the confidence coefficient requirement, for example, when the confidence coefficient requirement is not lower than 99.7%, the error limit of the passenger flow prediction is +/-3 times of the standard deviation.
The specific process is as follows:
the fitting error estimated by the least square algorithm is as follows:
(4)
wherein,is thatIs a unit matrix of (a).
The magnitude and distribution (mean and variance) of the fitting error can be used to evaluate the accuracy of the least squares estimate and, in turn, to correct the date type and time period of the division.
If the fitting error is approximately normal, the least squares solution is an unbiased estimate, i.e
(5)
Wherein the method comprises the steps ofRepresenting estimation errorsIs a mathematical expectation of (a).
And can obtain, estimate, the errorThe covariance matrix of (2) is:
(6)
wherein the method comprises the steps ofRepresenting estimation errorsIs a covariance of (c).
The relationship between the time period passenger flow and the above-mentioned time period passenger flow on a future date is assumed as follows:
(7)
wherein the method comprises the steps ofFor future passenger flow at that time period for that date,for the coefficient vector between the time period and the delimited time period, the optimal estimate of the passenger flow of the date in the time period in the future can be obtained from the above equation (7):
(8)
wherein the method comprises the steps ofFor the optimal estimation of passenger flow at that time period for the future date, it is obtained from equation (5) above
(9)
The best estimate is seen to be an unbiased estimate, and the variance of the best estimate obtained from equation (6) above is
(10)
And further obtaining the optimal estimate as the standard deviationI.e. the square root of the variance.
The prediction error limit of the passenger flow in the future period can be calculated according to the standard deviation, if the confidence is required to be not lower than 99.7 percent。
Example two
This example illustrates the implementation and specific steps of the present invention in one example:
acquiring historical passenger flow data of a certain station of a certain line; the historical passenger flow data at least comprises the arrival time of the vehicle and the number of passengers on the vehicle. And classifying the dates according to the historical passenger flow data change, so that the passenger flows on the same kind of date are similar in all-day distribution. Assuming that the passenger flows of the sites with the same date are distributed identically throughout the day, dividing the passenger flows of each class of date according to a certain principle, and setting the passenger flows of the same date in the same time period as a certain value.
And establishing an equation relation between the passenger flow of each time period and the real passenger flow according to the relation between the historical passenger flow data and the time period division. For example, the historical passenger flow data is 8 points in the morning, the number of passengers on a certain bus on a certain station is 8, and the time for the previous bus on the certain bus to reach the station is 7 points 50, so that the accumulated number of passengers under the bus to reach the station is 8 people within 10 minutes from 7 points 50 to 8 points. If the time period is divided into 10 time periods (i.e., one time period per minute) from 7 points 50 to 8 points, the number of passengers in the time period is assumed to be x 1 ,x 2 ,x 3 ,...x 10 The following equation can be obtained:
x 1 +x 2 +x 3 +x 4 +x 5 +x 6 +x 7 +x 8 +x 9 +x 10 =8(11)
similarly, all historical passenger flow data of the same date type can be processed similarly to the above, and the unknown number x can be obtained simultaneously 1 ,x 2 ,x 3 ,...x n As long as enough historical passenger flow data is available to ensure that the number of independent equations is not less than the number n of unknowns, a unique solution of the linear equation set can be obtained.
And (3) adopting the methods from the formula (1) to the formula (3) in the first embodiment to obtain the least square solution of the passenger number linear equation set. And the fitting error of the least square estimation is obtained by using the method of the formula (4) in the first embodiment, the magnitude and the distribution (mean and variance) of the fitting error can be used for evaluating the accuracy of the least square estimation value, and in turn, the fitting error can also be used for correcting the date classification and the time interval division.
If the time interval is divided into a relatively coarse period, such as half an hour, the time interval from 7 points 50 to 8 points is one third, and the passenger flow coefficient of the corresponding time interval isI.e.
(12)
The length of the time period can be divided into a time period as long as the historical passenger flow in the time period is relatively stable, so that the modeling process is flexible and convenient, and the number of unknown variables is reduced to the greatest extent. If the passenger arrival rate for each time period is defined as an unknown quantity, the number of passengers flowing for the time period for the non-uniform division can be obtained by multiplying the passenger arrival rate for the time period by the time period length.
After the invention obtains the whole-day distribution estimation of the passenger flow of each type of date, the passenger flow of the line at the station can be predicted in any time period of a future day.
Assuming that the passenger flow from 8 am to 8 am at a certain station needs to be predicted in the future, knowing the date type corresponding to the date and the passenger flow distribution (least square estimation of the passenger flow in each period) corresponding to the date type, and assuming that the relevant period from 8 am to 8 am is divided into 8:00-8:10, 8:10-8:20, 8:20-8:30, namely every tenOne time period per minute, assuming passenger flow estimation corresponding to the three time periods3,5,4, the predicted 8-point to 8-point half passenger flow value is 3+5+4=12.
Further, if it is assumed that the variance of the predicted passenger flow value is stationary, the variance of the predicted passenger flow value may be estimated using the above-described covariance matrix formula (6) of the least squares estimation, thereby giving a precision range of the predicted value (confidence that the ±3 times standard deviation may reach 99.7% or more under the assumption of normal distribution).
Assuming that the covariance matrix of the passenger flow estimation is:
(13)
the variance between the predicted and actual values of passenger flow from 8 to 8 points half is:
(14)
the accuracy range of the passenger flow predicted value from 8 points to 8 half points。
Example III
As shown in fig. 4, an embodiment of the present invention provides a public transportation passenger flow prediction system, including:
the data acquisition module is used for acquiring historical passenger flow data of a certain station of a certain line; the historical passenger flow data at least comprises the arrival time of vehicles and the number of passengers getting on;
the date classification module is used for classifying the dates according to the change of the historical passenger flow data so that the passenger flows on the same kind of date are distributed similarly all the day;
the time interval dividing module is used for dividing the time intervals of the passenger flows of each class of date according to a certain principle on the assumption that the passenger flows of the sites with the same class of date are distributed identically throughout the day, and the passenger flows of the same class of date in the same time interval are a certain value;
the passenger flow modeling module is used for establishing an equality relation between the passenger flow of each time period and the real passenger flow according to the relation between the historical passenger flow data and the time period division;
the passenger flow estimation module is used for obtaining the estimated value of the passenger flow in each period by solving an equation set formed by the equation relation, so as to obtain the whole-day distribution estimation of the passenger flow in each type of date;
and the passenger flow prediction module is used for predicting the number of passengers of the line at the station in any time period of a future day according to the category to which the future date belongs.
The description and specific examples of a public transportation passenger flow prediction method in the first embodiment are equally applicable to a public transportation passenger flow prediction system in the present embodiment, and from the foregoing detailed description of a public transportation passenger flow prediction method, those skilled in the art will clearly know the implementation method of a public transportation passenger flow prediction system in the present embodiment, so that the description will not be repeated for brevity.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments is modified or some or all of the technical features are replaced equivalently, so that the essence of the corresponding technical scheme does not deviate from the scope of the technical scheme of the embodiments of the present invention.

Claims (6)

1. A method for predicting mass transit passenger flow, comprising:
acquiring historical passenger flow data of a certain station of a certain line; the historical passenger flow data at least comprises the arrival time of vehicles and the number of passengers getting on;
date classification is carried out according to the historical passenger flow data change, so that the passenger flows on the same kind of date are distributed similarly throughout the day;
assuming that the passenger flows of the sites with the same class date are distributed identically throughout the day, dividing the passenger flows of each class of date according to a certain principle in a time interval division mode specifically as follows: the method comprises the steps of carrying out unequal interval division according to the size of the passenger flow, wherein the time interval range of large passenger flow is small, and the time interval range of small passenger flow is large; the passenger flow of the same period of the same kind of date is a certain value;
according to the relation between the historical passenger flow data and the time interval division, establishing an equality relation between the passenger flow of each time interval and the real passenger flow, wherein the equality relation is as follows:
the sum of accumulated passenger flows of relevant time periods in the arrival time of two adjacent public transportation vehicles=the number of boarding persons when the public transportation vehicles arrive at the station after one trip;
obtaining estimated values of passenger flows in each period by solving an equation set formed by the equation relation, so as to obtain the whole-day distribution estimation of the passenger flows in each type of date;
the specific mode is as follows:
the method for obtaining the estimated value of the passenger flow in each period by solving the linear equation set formed by the equation relation comprises the following steps: obtaining an optimal solution with minimum mean square error by using least square estimation; specifically, the least square algorithm is utilized to estimate and obtain the optimal solution with the minimum mean square error, which comprises the following steps:
the following general form of a linear equation system of the number of passengers is established:
；
let coefficient matrixWherein the coefficient matrixAIs>Indicate->In equation>Of unknown numberCoefficient of->Is a general representation of the least squares problem; specific each coefficient->Sources in practical application: the coefficients preceding the unknowns of equation (1)>The relation between the time division and the actual passenger flow data time range depends on the actual application; the time range represented by each actual passenger flow data covers a plurality of time periods, if the time range covers the whole time period, the coefficient in front of the passenger flow volume of the corresponding time period is 1, and if the time range is only partially covered, the coefficient in front of the passenger flow volume of the corresponding time period is set according to the covered proportion;
vector of number of accumulated people in arrival at bus in this periodTime-phased arrival people vector>Then an equivalent equation set of the passenger number linear equation set (1) is obtained:
；
the least squares solution of the set of equivalent equations is in the form of:
；
wherein,and->Representing the transpose and inverse of the matrix, respectively, +.>Is->An estimated value having the smallest mean square error;
predicting the passenger flow of the line at the station in any period of the future date according to the category to which the future date belongs;
the date classification is carried out according to the change of the historical passenger flow data, and the date classification mode comprises the following steps: classifying according to the passenger flow change rule, wherein the passenger flow change rule with the same passenger flow change rule is the same kind of date;
the classification according to the passenger flow change rule comprises the following steps:
acquiring historical passenger flow data of a certain station of a certain line every day;
since the vehicle arrival time per day is uncertain, the whole day passenger flow distribution per day needs to be obtained through interpolation fitting; the daily total-day passenger flow distribution is expressed as a passenger flow time-varying curve between the first and last buses of the station and the station time, and the curve is sampled at fixed time intervals to obtain a daily passenger flow vector;
distances between the vectors are defined according to the similarity between the daily traffic vectors, and the daily traffic vectors are clustered by using the distances.
2. The method of claim 1, wherein the historical passenger flow data is collected by one or more of a bus-mounted device, a platform monitoring device, and a manual recording mode.
3. The public transportation passenger flow prediction method according to claim 1, wherein in the date classification mode, in order to predict the passenger flow category to which a certain date belongs in the future, a relationship between a date attribute and the passenger flow category is established, comprising the following processes:
firstly, determining date attribute characteristics;
acquiring attribute characteristics of the date corresponding to the historical passenger flow data to form a characteristic vector of the date;
acquiring a passenger flow category of a date corresponding to the historical passenger flow data, and taking the passenger flow category as a category label of the date;
training a machine learning classification model according to the feature vector and the class label of each date;
and judging the passenger flow category to which the date belongs according to the classification model and the attribute characteristics of the date.
4. The method for predicting the passenger flow of public transportation according to claim 1, wherein the obtaining the estimated value of the passenger flow of each period by solving the linear equation set formed by the equation relation comprises: and obtaining an optimal solution with the minimum mean square error by using least square estimation.
5. The method for predicting the passenger flow of public transportation according to claim 1, wherein predicting the passenger flow of the line at the station in any period of time from the future date according to the category to which the future date belongs comprises:
acquiring date types of future dates, wherein each type of date type has the same all-day passenger flow distribution;
acquiring passenger flow estimated values of all time periods corresponding to the date type;
and calculating the number of passengers in the time period of the future date according to the relation between the time period of the future date and the time period division of the date type.
6. A mass transit passenger flow prediction system, comprising:
the data acquisition module is used for acquiring historical passenger flow data of a certain station of a certain line; the historical passenger flow data at least comprises the arrival time of vehicles and the number of passengers getting on;
the date classification module is used for classifying the dates according to the change of the historical passenger flow data so that the passenger flows on the same kind of date are distributed similarly all the day;
the time interval dividing module is used for dividing the passenger flow of each type of date according to a certain principle on the assumption that the passenger flow of the same stations with the same date are distributed identically throughout the day, and the time interval dividing mode is specifically as follows: the method comprises the steps of carrying out unequal interval division according to the size of the passenger flow, wherein the time interval range of large passenger flow is small, and the time interval range of small passenger flow is large; the passenger flow of the same period of the same kind of date is a certain value;
the passenger flow modeling module is used for establishing an equality relation between the passenger flow of each time period and the real passenger flow according to the relation between the historical passenger flow data and the time period division, wherein the equality relation is as follows: the sum of accumulated passenger flows of relevant time periods in the arrival time of two adjacent public transportation vehicles=the number of boarding persons when the public transportation vehicles arrive at the station after one trip;
the passenger flow estimation module is used for obtaining the estimated value of the passenger flow in each period by solving an equation set formed by the equation relation, so as to obtain the whole-day distribution estimation of the passenger flow in each type of date; the specific mode is as follows:
the method for obtaining the estimated value of the passenger flow in each period by solving the linear equation set formed by the equation relation comprises the following steps: obtaining an optimal solution with minimum mean square error by using least square estimation; specifically, the least square algorithm is utilized to estimate and obtain the optimal solution with the minimum mean square error, which comprises the following steps:
the following general form of a linear equation system of the number of passengers is established:
；
let coefficient matrixWherein the coefficient matrixAIs>Indicate->In equation>Coefficients of unknown number +.>Is a general representation of the least squares problem; specific each coefficient->Sources in practical application: the coefficients preceding the unknowns of equation (1)>The relation between the time division and the actual passenger flow data time range depends on the actual application; the time range represented by each actual passenger flow data covers a plurality of time periods, if the time range covers the whole time period, the coefficient in front of the passenger flow volume of the corresponding time period is 1, and if the time range is only partially covered, the coefficient in front of the passenger flow volume of the corresponding time period is set according to the covered proportion;
vector of number of accumulated people in arrival at bus in this periodTime-phased arrival people vector>Then an equivalent equation set of the passenger number linear equation set (1) is obtained:
；
the least squares solution of the set of equivalent equations is in the form of:
；
wherein,and->Representing the transpose and inverse of the matrix, respectively, +.>Is->An estimated value having the smallest mean square error;
the passenger flow prediction module is used for predicting the number of passengers of the line at the station in any time period of a future day according to the category to which the future date belongs;
the date classification is carried out according to the change of the historical passenger flow data, and the date classification mode comprises the following steps: classifying according to the passenger flow change rule, wherein the passenger flow change rule with the same passenger flow change rule is the same kind of date;
the classification according to the passenger flow change rule comprises the following steps:
acquiring historical passenger flow data of a certain station of a certain line every day;
since the vehicle arrival time per day is uncertain, the whole day passenger flow distribution per day needs to be obtained through interpolation fitting; the daily total-day passenger flow distribution is expressed as a passenger flow time-varying curve between the first and last buses of the station and the station time, and the curve is sampled at fixed time intervals to obtain a daily passenger flow vector;
distances between the vectors are defined according to the similarity between the daily traffic vectors, and the daily traffic vectors are clustered by using the distances.
CN202311595226.6A 2023-11-28 2023-11-28 Public transportation passenger flow prediction method and system Active CN117314504B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311595226.6A CN117314504B (en) 2023-11-28 2023-11-28 Public transportation passenger flow prediction method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311595226.6A CN117314504B (en) 2023-11-28 2023-11-28 Public transportation passenger flow prediction method and system

Publications (2)

Publication Number Publication Date
CN117314504A CN117314504A (en) 2023-12-29
CN117314504B true CN117314504B (en) 2024-03-29

Family

ID=89255510

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311595226.6A Active CN117314504B (en) 2023-11-28 2023-11-28 Public transportation passenger flow prediction method and system

Country Status (1)

Country Link
CN (1) CN117314504B (en)

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103984993A (en) * 2014-05-13 2014-08-13 东南大学 Rail transit passenger flow OD distribution real-time speculation method
CN105404934A (en) * 2015-11-11 2016-03-16 北京航空航天大学 Urban population mobile data model analysis method based on conditional random field
CN105404850A (en) * 2015-10-21 2016-03-16 中南大学 Station passenger flow monitoring system, station passenger flow monitoring method, station passenger flow control system and station passenger flow control method
CN108694464A (en) * 2018-05-14 2018-10-23 重庆交通开投科技发展有限公司 Passenger flow evacuation method and device
CN110390349A (en) * 2019-06-20 2019-10-29 浙江大学 Bus passenger flow volume based on XGBoost model predicts modeling method
CN111598333A (en) * 2020-05-14 2020-08-28 北京轨道交通路网管理有限公司 Passenger flow data prediction method and device
CN116142264A (en) * 2023-04-23 2023-05-23 北京全路通信信号研究设计院集团有限公司 Urban rail transit operation planning method and system
CN116151538A (en) * 2022-11-18 2023-05-23 东风悦享科技有限公司 Simulation system and method applied to intelligent bus demand balanced scheduling

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112598182B (en) * 2020-12-25 2022-11-29 卡斯柯信号有限公司 Intelligent scheduling method and system for rail transit

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103984993A (en) * 2014-05-13 2014-08-13 东南大学 Rail transit passenger flow OD distribution real-time speculation method
CN105404850A (en) * 2015-10-21 2016-03-16 中南大学 Station passenger flow monitoring system, station passenger flow monitoring method, station passenger flow control system and station passenger flow control method
CN105404934A (en) * 2015-11-11 2016-03-16 北京航空航天大学 Urban population mobile data model analysis method based on conditional random field
CN108694464A (en) * 2018-05-14 2018-10-23 重庆交通开投科技发展有限公司 Passenger flow evacuation method and device
CN110390349A (en) * 2019-06-20 2019-10-29 浙江大学 Bus passenger flow volume based on XGBoost model predicts modeling method
CN111598333A (en) * 2020-05-14 2020-08-28 北京轨道交通路网管理有限公司 Passenger flow data prediction method and device
CN116151538A (en) * 2022-11-18 2023-05-23 东风悦享科技有限公司 Simulation system and method applied to intelligent bus demand balanced scheduling
CN116142264A (en) * 2023-04-23 2023-05-23 北京全路通信信号研究设计院集团有限公司 Urban rail transit operation planning method and system

Also Published As

Publication number Publication date
CN117314504A (en) 2023-12-29

Similar Documents

Publication Publication Date Title
CN108072381B (en) Path planning method and device
CN108417034A (en) A kind of highway festivals or holidays vehicle flowrate prediction technique based on Multi-variable Grey Model
CN103065205A (en) Three-dimensional intelligent transportation junction passenger flow time-space analysis and prediction system
CN110634292B (en) Travel time reliability estimation method based on road resistance performance function
CN109670540B (en) Short-term passenger transport hub region resident number change trend prediction method based on kNN algorithm
CN111583628B (en) Road network heavy truck traffic flow prediction method based on data quality control
CN110853156A (en) Passenger OD identification method integrating bus GPS track and IC card data
Ma et al. Measuring service reliability using automatic vehicle location data
CN108876064B (en) Urban rail transit network key station identification method based on passenger flow
CN112562377A (en) Passenger vehicle real-time scheduling method based on random opportunity constraint
CN111199247B (en) Bus operation simulation method
Xie et al. Multistep prediction of bus arrival time with the recurrent neural network
CN115269758A (en) Passenger-guidance-oriented road network passenger flow state deduction method and system
Zheng et al. Improved iterative prediction for multiple stop arrival time using a support vector machine
Faghih et al. Taxi and mobility: Modeling taxi demand using ARMA and linear regression
CN112258029B (en) Demand prediction method for sharing bicycles around subway station
CN117314504B (en) Public transportation passenger flow prediction method and system
Fontes et al. A deep learning approach for predicting bus passenger demand based on weather conditions
CN110796301B (en) Passenger flow prediction method and device based on IC card data
Navick et al. Estimating Passenger Miles, Origin–Destination Patterns, and Loads with Location-Stamped Farebox Data
Tu et al. Analysis and forecast of passenger flow based on public transportation IC card and GPS data
CN105303246A (en) Multiline arrival time prediction for public transportation
CN115565376A (en) Vehicle travel time prediction method and system fusing graph2vec and double-layer LSTM
Pavlyuk et al. Spatiotemporal dynamics of public transport demand: a case study of Riga
Mattson et al. Marginal cost pricing and subsidy of small urban transit

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant