CN111626479B - Taxi passenger flow distribution point passenger carrying decision method and system based on real-time data - Google Patents

Taxi passenger flow distribution point passenger carrying decision method and system based on real-time data Download PDF

Info

Publication number
CN111626479B
CN111626479B CN202010361903.8A CN202010361903A CN111626479B CN 111626479 B CN111626479 B CN 111626479B CN 202010361903 A CN202010361903 A CN 202010361903A CN 111626479 B CN111626479 B CN 111626479B
Authority
CN
China
Prior art keywords
decision
factor
driver
passenger
income
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
CN202010361903.8A
Other languages
Chinese (zh)
Other versions
CN111626479A (en
Inventor
叶竞成
包兴达
方筠捷
王浩宇
邹辉
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.)
Nanjing University of Posts and Telecommunications
Original Assignee
Nanjing University of Posts and Telecommunications
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 Nanjing University of Posts and Telecommunications filed Critical Nanjing University of Posts and Telecommunications
Priority to CN202010361903.8A priority Critical patent/CN111626479B/en
Publication of CN111626479A publication Critical patent/CN111626479A/en
Application granted granted Critical
Publication of CN111626479B publication Critical patent/CN111626479B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

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
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G06Q10/047—Optimisation of routes or paths, e.g. travelling salesman problem
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • 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
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/067—Enterprise or organisation modelling
    • G06Q50/40—
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/20—Monitoring the location of vehicles belonging to a group, e.g. fleet of vehicles, countable or determined number of vehicles
    • G08G1/202—Dispatching vehicles on the basis of a location, e.g. taxi dispatching
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Abstract

The invention provides a valuation vehicle passenger flow distribution point passenger carrying decision method and a system based on real-time data, wherein the method comprises the following steps: (S1) establishing a multivariate linear regression model in which the decision value varies with the demand factor, the return income factor, the wait income factor and the personal ability factor; (S2) initializing the multiple linear regression model, providing a decision for a driver according to the multiple linear regression model based on the current demand factor, return income factor, waiting income factor and personal ability factor, and training the multiple linear regression model according to a decision satisfaction sample fed back by the driver after the decision; (S3) providing a decision for the driver using the trained multiple linear regression model; the system provides decision advice to the taxi driver according to the method. The invention can realize the real-time prediction of supply and demand relationship with higher accuracy so as to help a driver make a decision and improve the passenger trip efficiency and the driver income efficiency.

Description

Valuation vehicle passenger flow distribution point passenger carrying decision method and system based on real-time data
Technical Field
The invention relates to a decision-making method and a decision-making system, in particular to a taxi passenger flow distribution point passenger carrying decision-making method and a system based on real-time data.
Background
After a taxi driver sends a passenger to an airport, the taxi driver often needs to judge whether the passenger stays in the airport to carry a taxi to the airport or returns to the urban area to search for a passenger source. The prior art generally has the passenger to judge whether to receive the order according to the demand that sends through the APP for the taxi driver. The disadvantage of this approach is that it is generally difficult to predict the supply-demand relationship in real time for this type of APP, resulting in slightly lower reaction time and efficiency after the driver takes an order than if the driver were moving forward along with the supply-demand relationship. Only when the supply-demand relation is predicted in real time, the balance between the passenger flow of large-scale passenger distribution places and the supply and demand of taxi drivers can be ensured, and therefore the taxi can meet the requirements of the passengers in real time.
The invention patent application with the application number of 201911036243X discloses a decision-making method for passenger carrying problems of taxis in airports. The method comprises the steps of establishing a profit maximization model, comparing waiting net profits and returning net profits of taxi drivers, and selecting a decision that the net profits are the maximum. However, this application only considers objective factors such as flight and season factors and revenue factors (including waiting for net revenue and returning net revenue), but does not consider at all individual difference factors in driver passenger capacity and decision tendency under the same objective factors. Compared with objective factors, individual difference factors of drivers are difficult to grasp, but the individual difference factors directly cause that different individuals obtain obvious differences under the same decision. For example, the individual passenger capacity of a driver can be visualized by the empty rate. The greater the idle rate, indicating that it is more difficult to pull to the passenger at ordinary times, the less money is earned by the passenger, and the greater the idle rate the driver may be more inclined to remain at the airport for more consistent revenue. In addition, the decision method provided by the application is too rational and lacks elasticity, and the decision proportion of each factor cannot be adjusted according to the experience and feedback of a driver.
Disclosure of Invention
The purpose of the invention is as follows: in order to overcome the defects of the prior art, the invention aims to provide a taxi pricing car route planning suggestion method based on real-time data.
The invention also provides a pricing vehicle route planning and suggesting system based on the real-time data.
The technical scheme is as follows: in one aspect, the invention discloses a valuation vehicle passenger flow distribution point passenger carrying decision method based on real-time data. The method comprises the following steps:
(S1) establishing a multiple linear regression model:
Figure BDA0002475384100000011
wherein p is a decision value, when p is less than 0.5, waiting decision is made for the driver, and when p is more than 0.5, a decision for returning to the urban area is made for the driver; x 1 、X 2 、X 3 、X 4 Respectively corresponding to demand factor, return income factor, waiting income factor and personal ability factor, beta 0 、β 1 、β 2 、β 3 、β 4 First to fifth decision factors, respectively; the demand factor is the passenger flow volume after the weather influence is considered; the return income factor is the expected return net income B of the driver from the passenger flow distribution point to the urban area Return _ net (ii) a The waiting income factor is the expected waiting net income B of drivers returning to the urban area from the passenger flow distribution point Wait _ net (ii) a The personal ability factor is a driver idle rate;
(S2) is beta 0 、β 1 、β 2 、β 3 、β 4 Setting an initial value based on the current demand factor X 1 Returning a profit factor X 2 Waiting for revenue factor X 3 And personal ability factor X 4 The decision value p is calculated by the formula (1) to provide a decision for a driver, and beta is adjusted according to the decision satisfaction fed back by the driver after the decision 0 、β 1 、β 2 、β 3 、β 4 Is repeated until β before and after the correction, by correcting the decision value p 0 、β 1 、β 2 、β 3 、β 4 Is less than a threshold;
(S3) finally correcting the beta 0 、β 1 、β 2 、β 3 、β 4 Is used for the subsequent calculation of the decision value p, instead of the equation (1).
Further, in step (S2), β is adjusted according to the decision satisfaction of the driver feedback after the decision 0 、β 1 、β 2 、β 3 、β 4 To modify the decision value p, in particular by adjusting β according to the following equation 0 、β 1 、β 2 、β 3 、β 4 ：
Figure BDA0002475384100000021
h β ＝β 0 +β 1 X 1 +β 2 X 2 +β 3 X 3 ，
Wherein, beta' j Is beta j The adjustment value of (d); α is the learning rate and is a positive number; m is the total number of samples fed back by the driver; x is the number of i And y i Satisfaction and corresponding decision value in the ith sample, and x, respectively, of driver feedback i Can only take a value of 0 or 1, when x i When the value of (A) is 0, the driver does not recognize the decision, and when x is equal to i When the value of (1) represents a driver acceptance decision;
Figure BDA0002475384100000022
and the demand factor, the return income factor, the waiting income factor and the personal ability factor which correspond to the ith sample are respectively fed back by the driver.
Further, the passenger flow volume after considering the weather influence is equal to the product of the expected passenger flow volume when not considering the weather influence and the weather influence factor; the expected passenger flow volume is a variable which changes along with time; the weather impact factor is a ratio of a passenger volume that takes into account the current weather impact to an expected passenger volume that does not take into account the weather impact.
Further, the expectation returns a net profit B Return _ Net Calculated by the following way:
B return _ Net ＝B Return to -C Return to ，
B Return to ＝B h ×T w ，
C Return to ＝B h ×T r ，
Wherein, B Return to Return the expected return revenue, i.e. the expected waiting time consumed by leaving the airport for the revenue gained by the passenger after the driver returns to the downtown empty; c Return to No-load cost for drivers to return to the urban area from the passenger flow distribution points in no-load mode; b is h For average hourly income of drivers, B h The calculation method is as follows: calculating the average passenger-carrying mileage of a driver per hour through the historical data of the real-time positioning information of the valuation vehicle and the historical data of the real-time passenger-carrying information of the valuation vehicle, and calculating by combining with the valuation rule to obtain the average passenger-carrying mileage；T w Expected waiting time consumed for staying at an airport, T r The time for returning from the passenger flow distribution point to the downtown; and the urban area positioning is obtained by calculating real-time position data of the taxi and historical data of real-time passenger carrying information.
Further, the expected wait net gain B Wait _ net By the following equations:
B wait _ net ＝B Wait for -C Wait for ，
C Wait for ＝β×B h ×T w ，
Wherein, B Wait for The average waiting income of the urban area is returned from the passenger flow distribution point after the driver stays at the airport and carries passengers for the expected waiting income; c Wait for Wait costs for expected wait durations consumed in the case of staying at an airport; beta is a rest factor and is a constant; the expected waiting revenue B Wait for Determining the average distance from the passenger flow distribution point to the urban area through the pricing rule; and the urban area positioning is obtained by calculating real-time position data of the taxi and historical data of real-time passenger carrying information.
Further, the urban location is calculated by real-time taxi position data and historical real-time passenger information data, and specifically comprises the following steps:
(a) analyzing to obtain all passenger getting-off points according to the real-time position data of the taxi and the historical data of the real-time passenger carrying information;
(b) clustering the getting-off points of the passengers into 4 central points by using a K-means clustering method, which comprises the following steps: selecting 4 alighting points from all passenger alighting points randomly as initial central points by a random point selection mode; forming a cluster by each initial central point and the nearest departure point, so as to form 4 clusters, calculating the central positioning point of each cluster by combining the occurrence times and the positioning of all departure points in each cluster to replace the previous central point, and repeating the calculation until the positioning of the central point is not changed;
(c) and determining the positioning information of the urban area according to the finally obtained positioning of the 4 clustering center points and the corresponding number of the points closest to the getting-off point.
Further, the driver idle rate is calculated by:
(S41) determining the average daily idle distance and the average daily total driving distance of the driver according to the real-time position data of the taxi and the historical data of the real-time passenger carrying information;
(S42) dividing the daily average idling distance by the daily average total driving distance to obtain the idling rate of the driver.
Further, the passenger distribution point includes an airport or a passenger station.
In another aspect, the invention relates to a taxi passenger flow distribution point passenger carrying decision making system based on real-time data. The system comprises:
the demand factor calculation module is used for calculating the passenger flow volume after the weather influence is considered as a demand factor;
the return income factor calculation module is used for calculating expected return net income of the driver from the passenger flow distribution point to the urban area as a return income factor;
the waiting income factor calculating module is used for calculating expected waiting net income of the driver returned to the urban area from the passenger flow distribution point as a waiting income factor;
the personal ability factor calculation module is used for calculating the no-load rate of the driver as a personal ability factor;
a training module, configured to train the established multiple linear regression model, where the multiple linear regression model is expressed as:
Figure BDA0002475384100000041
wherein p is a decision value, when p is less than 0.5, waiting decision is made for the driver, and when p is more than 0.5, a decision for returning to the urban area is made for the driver; x 1 、X 2 、X 3 、X 4 Corresponding to demand factor, return income factor, waiting income factor and personal ability factor, beta 0 、β 1 、β 2 、β 3 、β 4 First to fifth decisions, respectivelyA factor; training the established multiple linear regression model comprises: is beta 0 、β 1 、β 2 、β 3 、β 4 Setting an initial value based on the current demand factor X 1 Return income factor X 2 Waiting for revenue factor X 3 And personal ability factor X 4 A decision value p is calculated according to the multiple linear regression model to provide a decision for a driver, and beta is adjusted according to the decision satisfaction degree fed back by the driver after the decision 0 、β 1 、β 2 、β 3 、β 4 Is repeated until β before and after the correction, by correcting the decision value p 0 、β 1 、β 2 、β 3 、β 4 The difference between the values of (a) is less than a threshold value;
a decision module for adjusting the final beta 0 、β 1 、β 2 、β 3 、β 4 And (3) calculating a decision value p by combining the multiple linear regression model.
Further, the training module is further configured to adjust β according to the following equation 0 、β 1 、β 2 、β 3 、β 4 ：
Figure BDA0002475384100000042
h β ＝β 0 +β 1 X 1 +β 2 X 2 +β 3 X 3 +β 4 X 4 ，
Wherein is beta' j Is beta j The adjustment value of (d); α is a positive scale factor and is a positive number; m is the total number of samples fed back by the driver; x is the number of i And y i Satisfaction and corresponding decision value in the ith sample fed back by the driver, respectively, and x i Can only take a value of 0 or 1, when x is i When the value of (A) is 0, the driver does not recognize the decision, and when x is equal to i When the value of (1) represents a driver acceptance decision;
Figure BDA0002475384100000043
and the demand factor, the return income factor, the waiting income factor and the personal ability factor which correspond to the ith sample are respectively fed back by the driver.
Has the advantages that: compared with the prior art, the invention has the following advantages:
1. the system can predict the supply and demand relationship in real time to help a driver to make a decision, so that taxis near a large-scale passenger collecting and distributing place basically meet the requirement of the passenger, overflow is avoided, and the traveling efficiency of the passenger and the income efficiency of the driver are improved;
2. meanwhile, weather factors, expected return net income, expected waiting net income and individual difference factors of drivers are considered, so that the decision result is closer to the ideal condition;
3. the influence ratio of model benefit factors, personal factors and demand factors on the final decision result can be gradually adjusted according to the decision satisfaction information fed back by the user, and the decision result is corrected, so that the requirements of a driver or a client are better met.
4. The urban area can be divided more effectively by utilizing the K-means clustering, and compared with the rough estimation of the distance returned to the urban area, the more accurate distance returned to the urban area can be obtained more effectively, so that the decision which is as correct as possible and has better user experience can be made.
Drawings
FIG. 1 is a data relationship diagram of one embodiment of the method of the present invention.
Detailed Description
The invention will be explained in more detail below with reference to specific embodiments and the accompanying drawings.
In one embodiment of the invention, an airport is selected as the passenger flow distribution point to illustrate the decision method of the invention. The decision method comprises the following steps:
the method comprises the following steps: and establishing a multivariate linear regression model with the decision value varying with the demand factor, the return income factor, the waiting income factor and the personal ability factor.
Firstly, establishing a multivariate linear regression model of logit (p) and X:
Figure BDA0002475384100000051
where Y ═ 1 denotes the driver decision return, p (Y ═ 1| X) is the probability that Y ═ 1 occurs under the condition that X occurs, X being the set of factors. Meanwhile, according to the regression relationship, there are
logit(p)＝β 0 +β 1 X 1 +β 2 X 2 +…+β k X k
Figure BDA0002475384100000052
Wherein p is a decision value, X 1 、X 2 ...X k Represents k influencing factors, beta 0 、β 1 、β 2 、β 3 、β k Is related to the decision proportion of each factor. Since the factors considered in this embodiment include demand factors, return income factors, waiting income factors, and personal ability factors, the value of k is 4, and the finally obtained multiple linear regression model is:
Figure BDA0002475384100000053
wherein, X 1 、X 2 、X 3 、X 4 Respectively corresponding to demand factor, return income factor, waiting income factor and personal ability factor, beta 0 、β 1 、β 2 、β 3 、β 4 Are the first to fifth decision factors. When p is less than 0.5, a waiting decision is made for the driver, and when p is more than 0.5, a return-to-urban decision is made for the driver.
The calculation of the factors is described below
1. Demand factors
On the one hand, the traffic volume at an airport generally varies with the number of flights, regardless of weather, and the taxi drivers who come to and go to the airport generally have some knowledge of the flights arriving at the airport. When the driver takes the passenger to the airport, if it is found that there are a lot of flights arriving at the airport, or it is found that it is the prime time to arrive at the airport. For example, because the late flight has a lower fare and the nighttime temperature is cooler and more comfortable than the afternoon and midday, people often do not choose the airplane arriving in the midday and the afternoon, but choose the airplane arriving at the evening to achieve higher cost performance. Thus, depending on the driver's experience, the expected traffic at night may be greater, and when the driver takes the passenger to the airport at night, there is a greater likelihood that the passenger will remain at the airport and move to a "pool" ready for pick-up. Likewise, the expected traffic volume at an airport is constantly in dynamic change at different times, i.e. the expected traffic volume is a variable f (t) that changes over time. To simplify the complexity of airport traffic data, it is also believed that the expected traffic volume at the airport will not change significantly every two hours. Thus, t is set to a time period divided every two hours, and f (t) represents the average expected passenger volume at the airport during the corresponding time period.
On the other hand, the weather conditions have various obvious influences on the flight of the airplane, and once the weather conditions are poor, the airplane can be delayed, stopped flying and the like. Thus, the influence of weather factors on passenger traffic does not vary appreciably. In this embodiment, different weather information is classified into 4 categories, and the specific categories are shown in table I:
TABLE I
Weather type Description of the invention
Weather of the first category Sunny and cloudy days
Weather of the second category Rain gust, rain in the small to medium, snow in the small to medium
Weather of type III Rain and snow, fog and haze
Weather of the IV class Heavy rain, heavy snow, thunderstorm, floating dust and fog
The impact of different weather types on passenger flow is indicated by a weather impact factor. For example, a weather effect factor may be defined as: the ratio of the passenger flow volume taking into account the current weather effect to the expected passenger flow volume without taking into account the weather effect. For weather types I, II, III, and IV in Table I, the weather impact factors may be 100%, 77%, 75%, 73%, respectively, based on historical airport statistics.
After comprehensively considering the expected passenger flow volume without considering the weather influence and the weather influence factor, determining the demand factor as the passenger flow volume with considering the weather influence, which is equal to the product of the expected passenger flow volume without considering the weather influence and the weather influence factor.
2. Expected return net gain
The driver's decision can be influenced by the magnitude of their expected net gain. If the driver expects that the passenger can get more net income when returning to the urban area, he is likely to be more inclined to return to the urban area to search for the passenger source; while the driver may be more inclined to leave a stable source of passengers at the reception airport if he believes that he may receive a greater net gain in leaving the "storage pool".
The return of net profit B is expected according to the general principles of economics Return _ Net Is equal to the expected return revenue B Return to And no load cost C Return to The difference of (a) is:
B return _ Net ＝B Return to -C No load
For drivers who choose to return to urban passenger carriers, return is expectedYi B Return to I.e. the expected waiting time T consumed in the case of leaving the airport after the driver returns to the urban area without load w The revenue obtained for carrying the passenger, namely:
B return to ＝B h ×T w ，
Wherein, B h Average driver gain per hour, B h The calculation method is as follows: the average passenger-carrying mileage of the driver per hour is calculated through the historical data of the real-time positioning information of the valuation vehicle and the historical data of the real-time passenger-carrying information of the valuation vehicle, and then the average passenger-carrying mileage is calculated by combining with the valuation rule.
And no-load cost C Return to The time cost in returning to the downtown area is mainly the idling cost of the driver returning to the downtown area from the airport where the driver is located without load. C Return to Equal to the product of the time to return to downtown and the average revenue per hour of the driver:
C return to ＝B h ×T r
Wherein, T r The time to return from the airport to the downtown.
Further explanation is needed regarding urban location. Generally, the urban area is a large area, and the departure point of the passenger is uncertain after the driver returns to the urban area. In order to overcome the problem and make the model more accurate, the invention utilizes a K-means clustering method to determine the location of the urban area. The specific method comprises the following steps:
(a) and analyzing to obtain the getting-off points of all passengers according to the real-time position data of the valuation vehicle and the historical data of the real-time passenger carrying information. For example, in the real-time passenger information, no load can be represented by "0", a passenger can be represented by "1", and the position of the pricing car when changing from "1" to "0" is determined by combining the real-time position data, that is, the location of the getting-off point of the passenger can be determined.
(b) Clustering the passenger getting-off points into 4 central points by using a K-means clustering method, comprising the following steps: selecting 4 alighting points from all passenger alighting points randomly as initial central points by a random point selection mode; forming a cluster by each initial central point and the nearest departure point, so as to form 4 clusters, calculating the central positioning point of each cluster by combining the occurrence times and the positioning of all departure points in each cluster to replace the previous central point, and repeating the calculation until the positioning of the central point is not changed;
(c) and determining the positioning information of the urban area according to the finally obtained positioning of the 4 clustering center points and the corresponding number of the points closest to the getting-off point.
After the positioning information of the urban area is obtained through calculation, the average distance of the driver for carrying passengers to the urban area in the airport can be determined, and then the related income is calculated by combining with the pricing rule.
3. Expected wait for net gain
Expected waiting net profit B Wait _ net Is equal to the expected waiting gain B Wait for And waiting cost C Wait for The difference of (a) is:
B wait _ net ＝B Wait for -C Wait for
Waiting revenue B is expected for drivers who choose to stay at the airport for pickup Wait for The average waiting income is the average waiting income returned to the urban area from the airport after the driver stays at the airport to carry the passenger, and the income is determined as the average passenger carrying income through the empirical judgment of the taxi driver and can be determined by the pricing rule of taxi charging and the average distance of returning the passenger from the airport to the urban area. While waiting for a cost C Wait for The latency cost is mainly due to the expected latency consumed in case of staying at the airport. The driver also has a certain rest time of relaxing while waiting in line, and when the driver feels a little tired, the driver may be more inclined to stay at the airport to wait, and take a rest while waiting in line. Therefore, wait for cost C Wait for The calculation method of (A) is as follows:
C wait for ＝β×B h ×T w
Where β is a rest factor, the rest factor β can be statistically analyzed according to questionnaires, and this embodiment is set to 70%.
4. Personal ability factors
The personal ability factor of the driver is mainly determined by the empty load rate. The idling rate of the driver is also an important factor influencing the decision, and is determined by the real-time positioning information and passenger carrying information of the vehicle. The driver's idling rate, and thus the amount of traffic on average, can affect his/her expectation of revenue. For example, a driver with a high idle rate may not receive a large number of orders at ordinary times, may prefer to leave them when transporting passengers to an airport, and earns a certain amount of revenue in the "pool". And if the driver is a driver with extremely strong working capacity, the idling rate is extremely low at ordinary times, and when the driver is confronted with selection, the driver can think that more net benefits can be obtained by returning to the urban area according to the working capacity of the driver, so that the driver is more inclined to return to the urban area. The driver idle rate is set to a variable a. It is defined as the ratio of the average daily idle distance of the driver to the average daily total distance traveled by the driver, i.e.:
Figure BDA0002475384100000081
the daily average idle distance and the daily average total driving distance of the driver can be determined according to the real-time position data of the taxi and the historical data of the real-time passenger carrying information.
Step two: determining each decision proportion beta in the multiple linear regression model established in the step one 0 、β 1 、β 2 、β 3 、β 4 The value of (c). The method mainly comprises the following steps: is beta 0 、β 1 、β 2 、β 3 、β 4 Setting an initial value based on the current demand factor X 1 Return income factor X 2 Waiting for a revenue factor X 3 And personal ability factor X 4 The decision value p is calculated by the formula (1) to provide a decision for a driver, and beta is adjusted according to the decision satisfaction degree fed back by the driver after the decision 0 、β 1 、β 2 、β 3 、β 4 Is repeated until β is before and after correction 0 、β 1 、β 2 、β 3 、β 4 Is less than the threshold value. Wherein, beta 0 、β 1 、β 2 、β 3 、β 4 The initial value setting of (A) can be random, or can be analyzed by the information of the valuation vehicle provided by the company and the likeAn approximate value was obtained.
The correction process mainly comprises adjusting the parameter beta by using a random ascending gradient method j The iterative formula is:
Figure BDA0002475384100000082
where α is a positive scale factor, and is the "learning rate" for setting the step size, J (β) j ) Is a function of the criteria. Alignment rule function J (β) j ) Differentiation was performed to obtain:
Figure BDA0002475384100000083
h β ＝β 0 +β 1 X 1 +β 2 X 2 +β 3 X 3 +β 4 X 4 ， (4)
wherein m is the total number of samples fed back by the driver; x is the number of i And y i Satisfaction and corresponding decision value in the ith sample fed back by the driver, respectively, and x i Can only take a value of 0 or 1, when x is i When the value of (A) is 0, the driver does not recognize the decision, and when x is equal to i When the value of (1) represents that the driver accepts the decision. Therefore, the above iterative equation (1) can be further converted into:
Figure BDA0002475384100000084
the above formula (1/m) is removed without affecting the result, and is equivalent to the following formula:
Figure BDA0002475384100000085
finally, the parameter β is given by the equation (6) j And (4) adjusting.
Step three: will finally correct beta 0 、β 1 、β 2 、β 3 、β 4 Substituting the value of (a) into the multiple linear regression model (i.e. the model in equation (1)) built in the step one for the subsequent calculation of the decision value p, so as to provide the driver with a decision according to the decision value p.
Correspondingly, the system for deciding the passenger carrying of the valuation vehicle passenger flow distribution point based on the real-time data comprises:
the demand factor calculation module is used for calculating the passenger flow volume after the weather influence is considered as a demand factor;
the return income factor calculation module is used for calculating expected return net income of the driver from the passenger flow distribution point to the urban area as a return income factor;
the waiting income factor calculating module is used for calculating expected waiting net income of the driver returned to the urban area from the passenger flow distribution point as a waiting income factor;
the personal ability factor calculation module is used for calculating the no-load rate of the driver as a personal ability factor;
a training module, configured to train the established multiple linear regression model, where the multiple linear regression model is expressed as:
Figure BDA0002475384100000091
wherein, p is a decision value, when p is less than 0.5, a waiting decision is made for the driver, and when p is more than 0.5, a city return decision is made for the driver; x 1 、X 2 、X 3 、X 4 Respectively corresponding to demand factor, return income factor, waiting income factor and personal ability factor, beta 0 、β 1 、β 2 、β 3 、β 4 Are each X 1 、X 2 、X 3 、X 4 The decision ratio is occupied; training the established multiple linear regression model comprises: is beta 0 、β 1 、β 2 、β 3 、β 4 Setting an initial value based on the current demand factor X 1 Return income factor X 2 Waiting for a revenue factor X 3 And personal ability factor X 4 According toThe multiple linear regression model calculates a decision value p to provide a decision for a driver, and beta is adjusted according to the decision satisfaction degree fed back by the driver after the decision 0 、β 1 、β 2 、β 3 、β 4 Is repeated until β before and after the correction, by correcting the decision value p 0 、β 1 、β 2 、β 3 、β 4 The difference between the values of (a) is less than a threshold value;
a decision module for adjusting the final beta 0 、β 1 、β 2 、β 3 、β 4 And (3) calculating a decision value p by combining the multiple linear regression model.
Further, the training module is further configured to adjust β according to the following equation 0 、β 1 、β 2 、β 3 、β 4 ：
Figure BDA0002475384100000092
h β ＝β 0 +β 1 X 1 +β 2 X 2 +β 3 X 3 +β 4 X 4 ，
Wherein is beta' j Is beta j The adjustment value of (d); α is a positive scale factor and is a positive number; m is the total number of samples fed back by the driver; x is a radical of a fluorine atom i And y i Satisfaction and corresponding decision value in the ith sample fed back by the driver, respectively, and x i Can only take a value of 0 or 1, when x is i When the value of (b) is 0, the driver does not recognize the decision, and when x is i When the value of (1) represents a driver acceptance decision;
Figure BDA0002475384100000093
demand factors, return income factors, waiting income factors and personal ability factors corresponding to the ith sample fed back by the driver respectively
The above-described system may be implemented by software and/or programmable hardware.

Claims (10)

1. A valuation vehicle passenger flow distribution point passenger carrying decision method based on real-time data is characterized by comprising the following steps:
(S1) establishing a multiple linear regression model, said multiple linear regression model being represented by:
Figure FDA0002475384090000011
wherein, p is a decision value, when p is less than 0.5, a waiting decision is made for the driver, and when p is more than 0.5, a city return decision is made for the driver; x 1 、X 2 、X 3 、X 4 Respectively corresponding to demand factor, return income factor, waiting income factor and personal ability factor, beta 0 、β 1 、β 2 、β 3 、β 4 First to fifth decision scale factors, respectively; the demand factor is the passenger flow volume after the weather influence is considered; the return income factor is the expected return net income B of the driver from the passenger flow distribution point to the urban area Return _ Net (ii) a The waiting income factor is the expected waiting net income B of drivers returning to the urban area from the passenger flow distribution point Wait _ net (ii) a The personal ability factor is a driver idle rate;
(S2) is beta 0 、β 1 、β 2 、β 3 、β 4 Setting an initial value based on the current demand factor X 1 Returning a profit factor X 2 Waiting for a revenue factor X 3 And personal ability factor X 4 A decision value p is calculated according to the multiple linear regression model to provide a decision for a driver, and beta is adjusted according to the decision satisfaction degree fed back by the driver after the decision 0 、β 1 、β 2 、β 3 、β 4 Is repeated until β is adjusted before and after 0 、β 1 、β 2 、β 3 、β 4 The difference between the values of (a) is less than a threshold value;
(S3) according to the finally adjusted beta 0 、β 1 、β 2 、β 3 、β 4 In combination with said multiple linear regression modelThe calculation of the subsequent decision value p is performed.
2. The real-time data-based valuation vehicle passenger flow distribution point passenger carrying decision making method according to claim 1, wherein in the step (S2), β is adjusted according to decision satisfaction fed back by a driver after decision making 0 、β 1 、β 2 、β 3 、β 4 To modify the decision value p, in particular by adjusting β according to the following equation 0 、β 1 、β 2 、β 3 、β 4 ：
Figure FDA0002475384090000012
h β ＝β 0 +β 1 X 1 +β 2 X 2 +β 3 X 3 +β 4 X 4 ，
Wherein is beta' j Is beta j The adjustment value of (d); alpha is a positive scale factor and is a positive number; m is the total number of samples fed back by the driver; x is the number of i And y i Satisfaction and corresponding decision value in the ith sample, and x, respectively, of driver feedback i Can only take a value of 0 or 1, when x is i When the value of (b) is 0, the driver does not recognize the decision, and when x is i When the value of (1) represents a driver acceptance decision;
Figure FDA0002475384090000013
and the demand factor, the return income factor, the waiting income factor and the personal ability factor corresponding to the ith sample are respectively fed back by the driver.
3. The real-time data-based pricing vehicle passenger flow distribution point passenger carrying decision method according to claim 1, characterized in that the passenger flow volume after considering weather influence is equal to the product of the expected passenger flow volume without considering weather influence and a weather influence factor; the expected passenger flow volume is a variable which changes along with time; the weather impact factor is a ratio of a passenger volume that takes into account the current weather impact to an expected passenger volume that does not take into account the weather impact.
4. The real-time data based pricing vehicle passenger flow distribution point passenger carrying decision method of claim 1, wherein the expected return net gain B Return _ Net Calculated by the following way:
B return _ Net ＝B Return to -C Return to ，
B Return to ＝B h ×T w ，
C Return to ＝B h ×T r ，
Wherein, B Return to The expected return income, namely the expected waiting time consumed in the case of leaving the airport after the driver returns to the urban area without load is used for the income obtained by carrying passengers; c Return to No-load cost for drivers to return to the urban area from the passenger flow distribution points in no-load mode; b h For average hourly income of drivers, B h The calculation method is as follows: calculating the average passenger-carrying mileage of a driver per hour according to the historical data of the real-time positioning information of the valuation vehicle and the historical data of the real-time passenger-carrying information of the valuation vehicle, and then calculating by combining with the valuation rule; t is w Expected waiting time consumed for staying at an airport, T r The time for returning from the passenger flow distribution point to the downtown; and the urban area positioning is obtained by calculating real-time position data of the taxi and historical data of real-time passenger carrying information.
5. The real-time data based pricing vehicle passenger flow distribution point passenger carrying decision method of claim 1, wherein the expected waiting net gain B Wait _ net By the following equations:
B wait _ net ＝B Wait for -C Wait for ，
C Wait for ＝β×B h ×T w ，
Wherein, B Wait for The average waiting income of the urban area is returned from the passenger flow distribution point after the driver stays at the airport and carries passengers for the expected waiting income; c Wait for Wait costs for expected wait durations consumed in the case of staying at an airport; beta is a rest factor and is a constant; the expected waiting revenue B Wait for Determining the average distance from the passenger flow distribution point to the urban area through the pricing rule; and the urban area positioning is obtained by calculating real-time position data of the taxi and historical data of real-time passenger carrying information.
6. The taxi passenger flow distribution point passenger carrying decision method based on real-time data as claimed in claim 4 or 5, wherein the urban location is calculated from taxi real-time position data and historical data of real-time passenger carrying information, and specifically comprises:
(a) analyzing to obtain all passenger getting-off points according to the real-time position data of the taxi and the historical data of the real-time passenger carrying information;
(b) clustering the passenger getting-off points into 4 central points by using a K-means clustering method, comprising the following steps: selecting 4 alighting points from all passenger alighting points randomly as initial central points by a random point selection mode; forming a cluster by each initial central point and the nearest departure point, so as to form 4 clusters, calculating the central positioning point of each cluster by combining the occurrence times and positioning of all departure points in each cluster to replace the central point of the last time, and repeating the calculation until the positioning of the central point is not changed;
(c) and determining the positioning information of the urban area according to the finally obtained positioning of the 4 clustering center points and the corresponding number of the points closest to the getting-off point.
7. The real-time data-based pricing vehicle passenger flow distribution point passenger carrying decision method according to claim 1, characterized in that the driver idle rate is calculated by:
(S41) determining the average daily idle distance and the average daily total driving distance of the driver according to the real-time position data of the taxi and the historical data of the real-time passenger carrying information;
(S42) dividing the daily average idling distance by the daily average total driving distance to obtain the idling rate of the driver.
8. A real-time data based fare collection and distribution point passenger loading decision method as claimed in claim 1, characterized in that the passenger flow collection and distribution point comprises an airport or a passenger station.
9. A valuation vehicle passenger flow distribution point passenger carrying decision making system based on real-time data is characterized by comprising:
the demand factor calculation module is used for calculating the passenger flow volume after the weather influence is considered as a demand factor;
the return income factor calculation module is used for calculating expected return net income of the driver from the passenger flow distribution point to the urban area as a return income factor;
the waiting income factor calculating module is used for calculating expected waiting net income of the driver returned to the urban area from the passenger flow distribution point as a waiting income factor;
the personal ability factor calculation module is used for calculating the no-load rate of the driver as a personal ability factor;
a training module, configured to train the established multiple linear regression model, where the multiple linear regression model is expressed as:
Figure FDA0002475384090000031
wherein, p is a decision value, when p is less than 0.5, a waiting decision is made for the driver, and when p is more than 0.5, a city return decision is made for the driver; x 1 、X 2 、X 3 、X 4 Respectively corresponding to demand factor, return income factor, waiting income factor and personal ability factor, beta 0 、β 1 、β 2 、β 3 、β 4 First to fifth decision scale factors, respectively; training the established multiple linear regression model comprises: is beta 0 、β 1 、β 2 、β 3 、β 4 Setting an initial value based on the current demand factor X 1 Return income factor X 2 Waiting for a revenue factor X 3 And personal ability factor X 4 A decision value p is calculated according to the multiple linear regression model to provide a decision for a driver, and beta is adjusted according to the decision satisfaction degree fed back by the driver after the decision 0 、β 1 、β 2 、β 3 、β 4 Is repeated until β is before and after correction 0 、β 1 、β 2 、β 3 、β 4 The difference between the values of (a) is less than a threshold value;
a decision module for adjusting the final beta 1 、β 2 、β 3 、β 4 And (3) calculating a decision value p by combining the multiple linear regression model.
10. The real-time data based fare collection and distribution point passenger loading decision system of claim 9, characterized in that the training module is further configured to adjust β according to the following equation 1 、β 2 、β 3 、β 4 ：
Figure FDA0002475384090000032
h β ＝β 0 +β 1 X 1 +β 2 X 2 +β 3 X 3 +β 4 X 4 ，
Wherein, beta' j Is beta j The adjustment value of (a); α is a positive scale factor and is a positive number; m is the total number of samples fed back by the driver; x is a radical of a fluorine atom i And y i Satisfaction and corresponding decision value in the ith sample fed back by the driver, respectively, and x i Can only take a value of 0 or 1, when x is i When the value of (A) is 0, the driver does not recognize the decision, and when x is equal to i When the value of (1) represents a driver acceptance decision;
Figure FDA0002475384090000033
demand factors, return income factors, waiting income factors and personal ability factors corresponding to the ith sample fed back by the driver respectivelyAnd (4) element.
CN202010361903.8A 2020-04-30 2020-04-30 Taxi passenger flow distribution point passenger carrying decision method and system based on real-time data Active CN111626479B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010361903.8A CN111626479B (en) 2020-04-30 2020-04-30 Taxi passenger flow distribution point passenger carrying decision method and system based on real-time data

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010361903.8A CN111626479B (en) 2020-04-30 2020-04-30 Taxi passenger flow distribution point passenger carrying decision method and system based on real-time data

Publications (2)

Publication Number Publication Date
CN111626479A CN111626479A (en) 2020-09-04
CN111626479B true CN111626479B (en) 2022-08-26

Family

ID=72259757

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010361903.8A Active CN111626479B (en) 2020-04-30 2020-04-30 Taxi passenger flow distribution point passenger carrying decision method and system based on real-time data

Country Status (1)

Country Link
CN (1) CN111626479B (en)

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107832882A (en) * 2017-11-03 2018-03-23 上海交通大学 A kind of taxi based on markov decision process seeks objective policy recommendation method
CN110889591A (en) * 2019-10-29 2020-03-17 淮阴工学院 Decision method for passenger carrying problem of airport taxi
CN111047138A (en) * 2019-11-13 2020-04-21 长沙学院 Taxi waiting aid decision-making method and system based on passenger carrying benefit analysis
CN111079008B (en) * 2019-12-11 2023-06-20 山东师范大学 Scheme recommendation method and system for taxi driver to leave in storage pool

Also Published As

Publication number Publication date
CN111626479A (en) 2020-09-04

Similar Documents

Publication Publication Date Title
CN111523722B (en) Intelligent charging station optimization selection system based on deep reinforcement learning
CN110555990A (en) effective parking space-time resource prediction method based on LSTM neural network
Yazici et al. A big data driven model for taxi drivers' airport pick-up decisions in new york city
CN109344991B (en) Method for predicting passenger flow of highest section of bus line
CN110119890B (en) Railway riding scheme sorting method based on AHP-grey correlation analysis
CN110889591A (en) Decision method for passenger carrying problem of airport taxi
Pandit et al. A framework for determining commuter preference along a proposed bus rapid transit corridor
CN115392949A (en) Rail transit early-peak time-sharing pricing method based on passenger departure time selection
CN109409955A (en) Timesharing customizes public transport fares formulating method by stages under changes in demand
CN116070033A (en) Novel shared public transportation transfer demand estimation method based on mobile phone signaling data
Trapani et al. An analysis of the impact of open entry on price and the quality of service in the airline industry
CN111626479B (en) Taxi passenger flow distribution point passenger carrying decision method and system based on real-time data
CN111445091B (en) Intelligent decision-making method for taxi driver in passenger transport center
CN109583634A (en) A kind of take-away Distribution path selection method based on Modern Portfolio Theory
CN111160597B (en) Intelligent scheduling method based on comprehensive satisfaction degree of taxi driver
Bi et al. Real trip costs: Modelling intangible costs of urban online car-hailing in Haikou
CN113537569A (en) Short-term bus passenger flow prediction method and system based on weight stack decision tree
CN106682759B (en) Battery supply system for electric taxi and network optimization method
CN112529288A (en) Multi-train differential pricing strategy optimization method and system
Viton Air deregulation revisited: choice of aircraft, load factors, and marginal-cost fares for domestic air travel
CN112200625A (en) Flight resource recommendation method and device
CN116562574A (en) Inter-city customized passenger transport collaborative optimization method, device, equipment and storage medium
CN116453352A (en) Freight car traffic flow prediction method based on K clustering algorithm and neural network
Wang et al. Aggregate modeling and equilibrium analysis of the crowdsourcing market for autonomous vehicles
CN111079008B (en) Scheme recommendation method and system for taxi driver to leave in storage pool

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