A method for dynamically predicting passenger waiting time using online car-hailing order data

The Naive Bayes classifier analyzes online car-hailing order data to predict the waiting time for passengers, solves the problem of unpredictable waiting time for passengers during specific periods and improves the travel experience.

CN114548461BActive Publication Date: 2024-12-03BEIJING UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210183024.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-02-27
Publication Date
2024-12-03
Estimated Expiration
2042-02-27

AI Technical Summary

Technical Problem

During transportation hubs, airports, morning rush hour, evening rush hour and other periods, the waiting time of online car-hailing passengers is unpredictable, resulting in immeasurable waiting time for passengers.

Method used

By establishing a naive Bayes classifier, the waiting time of passengers is predicted using online car-hailing order data, and characteristic parameters such as vehicle use period, land type, the number of recent appointments and completed orders, and the accumulated load level are calculated to determine the waiting time status.

Benefits of technology

It achieves rapid and accurate prediction of the waiting time of online car-hailing passengers, helps passengers make decisions, and improves the travel experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114548461B_ABST
    Figure CN114548461B_ABST
Patent Text Reader

Abstract

A method for dynamically predicting the waiting time of passengers using online car-hailing order data belongs to the field of transportation, and can dynamically predict the waiting time of passengers in a specific area based on the order data of online car-hailing users. A waiting time state recognition method based on a naive Bayes classifier is established based on the order data of online car-hailing users, including: establishing a naive Bayes classifier, calculating the probability of a certain type of waiting time state of passengers and the probability of occurrence of characteristic parameter values ​​(car use period, land use type, number of reservation orders, number of completed orders and cumulative load degree in the last t minutes of the grid I where the passenger is located); using a naive Bayes classifier to distinguish the waiting time state of the online car-hailing order data, first pre-processing the data, determining the grid corresponding to each passenger, calculating the value of each characteristic parameter, and using the established naive Bayes classifier to calculate the conditional probability of each type of waiting time, and judging the category of the passenger's waiting time state.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of transportation technology. It can dynamically predict the waiting time for passengers to book a car in a specific area based on the order data of online car-hailing users, and provide information services for urban residents to travel. Background Art

[0002] Online ride-hailing has become a common means of transportation for ordinary people. The order reservation method allows passengers to know in advance how long it will take for the reserved vehicle to arrive at the designated pick-up location, which greatly reduces the waiting time for passengers to arrive at the pick-up location in advance and improves the travel experience of passengers. However, it is often difficult to predict the time when the driver accepts the order. For example, in transportation hubs such as airports and railway stations, during the morning rush hour in residential areas, and during the evening rush hour in employment areas, there are often no drivers to accept orders for a long time, resulting in problems such as the inability to estimate the waiting time for passengers. In fact, there are many factors that affect the time it takes for drivers to accept orders, such as travel time during the morning rush hour, flat peak, evening rush hour, night, etc., land use types such as residential, office, commercial, hub, etc., the number of reservation orders in the recent period, the number of completed orders, historical order load, etc. These factors are intertwined and comprehensively affect the driver's order queue time. Therefore, this patent uses historical statistical data to predict the comprehensive time that affects passenger waiting. Summary of the invention

[0003] The present invention provides a method for dynamically predicting the waiting time of passengers by using online car-hailing order data, which lays a foundation for using online car-hailing order data to infer the waiting time in any situation and help passengers make decisions. The method is divided into two parts, and the process is as follows:

[0004] First, a naive Bayes classifier is established, and a certain amount of training samples are used to calculate the probability of a certain waiting time state for passengers and the probability of the occurrence of characteristic parameter values ​​(time period of use, land type, number of reservation orders in the last t minutes of the grid I where the passenger is located, number of completed orders and cumulative load) when a certain stay state occurs; second, the naive Bayes classifier is used to distinguish the waiting time state of online car-hailing order data. The data is first pre-processed to determine the grid corresponding to each passenger, and the value of each characteristic parameter is calculated. Finally, the established naive Bayes classifier is used to calculate the conditional probability of each type of waiting time and determine the category of the passenger's waiting time state.

[0005] The method comprises the following steps:

[0006] Step 1: Construct a set of passenger waiting time states W, which can be divided into W = {w1,w2,w3,...w k}, where w k Indicates the waiting time. The larger the k value, the longer the waiting time.

[0007] Step 2: Urban road grid division: divide the urban road network into M×N grid maps according to longitude and latitude. The set of regions can be expressed as:

[0008] A={I 1×1 ,I 1×2 ,I 1×3 ...I 1×N ,I 2×1 ...I M×N}

[0009] Where I represents a sub-region in the M×N region.

[0010] Step 3: Preprocess the data. The data fields for receiving online car-hailing orders include the order number, the passenger's order time, the longitude and latitude of the passenger's car-hailing location, the longitude and latitude of the destination, and the order type. The single operation data records the complete information of the passenger's online car-hailing trip, including the passenger's order time, boarding time, alighting time, departure location, and arrival location. According to the unique order number of each record, the passenger's order time, boarding time, the longitude and latitude of the passenger's car-hailing location, and the departure location information are retained. If the order is not completed, only the longitude and latitude and the car-hailing time are retained. The boarding time minus the order time is the user's waiting time. According to the longitude and latitude of the passenger's car-hailing location, match it with the grid divided in step 1 to determine the grid I where each passenger is located.

[0011] Step 3: Calculate the characteristic parameters, including the time period T, the type of land L, and the number of reservation orders Q in the last t minutes of the grid I where the passenger is located s 、Complete order quantity Q c And the cumulative load G, the process is as follows:

[0012] (1) The car use period is used to represent the time characteristics of the user's car use. The car use period set T = {t1, t2, t3, t4} is constructed, and the 24 hours of the day are divided into four periods, t1 is the morning peak period, t2 is the off-peak period, t3 is the evening peak period, and t4 is the night period. The user's car use period T is determined according to the user's order time;

[0013] (2) The land use type represents the land use attribute characteristics of the grid to which the user belongs. The land use type set L = {l1, l2, l3, l4, l5} is constructed, where l1 represents residential land, l2 represents office land, l3 represents commercial land, l4 represents transportation hub land, and l5 represents other land use types. Determine the land use type of each grid I;

[0014] (3) The number of reservation orders in the last t minutes of the grid where the passenger is located is calculated by the number of all reservation orders Q within t minutes before the passenger places the order. s This includes both successfully completed orders and cancelled or defaulted orders;

[0015] (4) The number of reservation orders in the last t minutes of the grid where the passenger is located is calculated by the number of all completed orders Q within t minutes before the passenger places the order. c express;

[0016] (5) Cumulative load refers to the ratio of the number of completed orders to the number of scheduled orders in the same area and time period (the last t minutes of the grid I where the passenger is located) for J consecutive days before the passenger places an order. It describes the overall order completion level of the area within t minutes before the passenger places an order. The cumulative load is expressed as Where j represents the j days before the passenger uses the car, Q sj represents the number of reservation orders made by passengers during the same period j days before the date of use of the car. s , Q cj represents the number of orders completed in the same period j days before the passenger's car date Q c ;

[0017] Step 4: Establish a naive Bayes classifier for different categories. The key indicator calculation process of the classifier is as follows:

[0018] (1) Assume that there is a certain amount of sample data R for training the classifier. The order completion status and waiting time are known a priori. The waiting time can be obtained by subtracting the order time from the boarding time of passengers who have successfully completed the order. Define the feature attribute set D, which includes the passenger's car use time T, land type L, and the number of reservation orders Q. s 、Complete order quantity Q c , cumulative load degree G;

[0019] (2) For any passenger data, calculate the probability of being in a certain waiting time Indicates the waiting time is w k The number of records, Q represents the total number of training samples;

[0020] (3) Calculate the discrete probability of different vehicle use periods T under each waiting time condition Indicates that in the training sample, the waiting time is w k The parking period is T i The sample size is calculated; the discrete probability of land use type L occurring under each waiting time condition is calculated. Indicates that in the training sample, the waiting time is w k The land use type in this state is L i Sample size; calculate the number of appointment orders Q under each waiting time state s Discrete probability of occurrence Indicates that in the training sample, the waiting time is w k The number of reservation orders in this state is Q si Sample size; calculate the number of completed orders Q under each waiting time state c Discrete probability of occurrence Indicates that in the training sample, the waiting time is w k The number of completed orders in this state is Q ci The sample size is calculated; the discrete probability of the cumulative load G occurring under each waiting time state is calculated. Indicates that in the training sample, the waiting time is w k The cumulative load in the state is G i The sample size;

[0021] Step 5: Calculate the probability P that R belongs to each type of waiting time state i (w k |T i ,L i ,Q si ,Q ci ,G i ), indicating passenger i, who uses the vehicle for a period of T i , land use type is L i , the number of reservation orders is Q si , the number of completed orders is Q ci , the cumulative load is G i Under the condition of k The probability of is calculated as follows:

[0022]

[0023] Where: P D =P(T i )·P(L i )·P(Q si )·P(Q ci )·P(G i ), P(T i ) indicates that the vehicle is used for a period of time T i The probability, P(L i ) indicates that the land use type is L i The probability, P(Q si ) indicates that the number of reservation orders is Q si The probability, P(Q ci ) indicates that the number of completed orders is Q ci The probability of P(G i ) indicates the cumulative load is G i The probability of, other symbols have the same meanings as before;

[0024] Step 6: Compare P under different waiting time conditions i (w k |T i ,L i ,Q si ,Q ci ,G i ) value, the largest one corresponds to the waiting time state w k The final determined passenger waiting time status;

[0025] The basic data of this invention comes from the online car-hailing APP. By establishing a naive Bayes classifier, it can quickly and accurately predict the waiting time of online car-hailing passengers and help passengers make decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 .Implementation process of this method DETAILED DESCRIPTION

[0027] Take the mobile phone positioning data generated by a user as an example.

[0028] Step 1: Construct a passenger waiting time state set W, which is divided into W = {w1, w2, w3} according to the passengers' daily waiting experience, where 0≤w1<5 minutes means the waiting time is short; 5≤w2<15 minutes means the waiting time is long; w3≥15 minutes means the waiting time is even longer.

[0029] Step 2: Urban road grid division: divide the urban road network into M×N grid maps according to longitude and latitude. The set of regions can be expressed as:

[0030] L={I 1×1 ,I 1×2 ,I 1×3 ...I 1×N ,I 2×1 ...I M×N}

[0031] Where I represents a sub-region in the M×N region;

[0032] Step 3: Preprocess the data. The data fields for receiving online car-hailing orders include order number, passenger order time, longitude and latitude of the passenger's car-hailing location, longitude and latitude of the destination, order type, and driver ID, as shown in Table 1; single operation data records the complete information of passengers taking online car-hailing trips, including passenger order time, boarding time, disembarkation time, departure location, arrival location, order time, driver ID, and platform abbreviation, as shown in Table 2. According to the unique order number of each record, the passenger's order time, boarding time, longitude and latitude of the passenger's car-hailing location, and departure location information are retained. If the order is not completed, only the longitude and latitude and car-hailing time are retained. The boarding time minus the order time is the user's waiting time; according to the longitude and latitude of the passenger's car-hailing location, match it with the grid divided in step 1 to determine the grid I where each passenger is located, as shown in Table 3;

[0033] Table 1: Order receiving data fields

[0034] Fields meaning Fields meaning ORDER_NO Order Number DESTINATION destination PASSENGER_TEL Passenger phone number DEST_LON Destination longitude USE_LOCALE Passenger car location DEST_LAT Destination Latitude USE_LON Longitude of passenger's car location BESPEAKTYPE Order Type USE_LAT Passenger vehicle location latitude DRIVERIDCODE Driver ID ORDER_TIME Passenger order time DATA_SOURCE Platform Abbreviation

[0035] Table 2: Single operation data fields

[0036] Fields meaning Fields meaning ORDERNO Order Number ONTIME The actual time the passenger boarded the vehicle PASSENGER_TEL Passenger phone number OFFTIME The actual time when the passenger gets off the bus REAL_VEHICLE_LON Longitude of the actual departure point of the vehicle ORDER_TIME Order time REAL_VEHICLE_LAT The latitude of the actual departure point of the vehicle DRIVERIDCODE Driver ID DEST_VEH_LON Longitude of the vehicle's actual arrival location DATA_SOURCE Platform Abbreviation DEST_VEH_LAT The latitude of the vehicle's actual arrival location

[0037] Table 3: Sample data

[0038]

[0039]

[0040] Step 3: Calculate the characteristic parameters, including the time period T, the type of land L, and the number of reservation orders Q in the last t minutes of the grid I where the passenger is located s 、Complete order quantity Q c And the cumulative load G, the process is as follows:

[0041] (1) Calculate the car use time period and construct the car use time period set T = {t1, t2, t3, t4}. The 24 hours of the day are divided into four time periods: t1 is the morning peak period 7:00-9:00, t2 is the off-peak period 9:00-17:00, t3 is the evening peak period 17:00-19:00, and t4 is the night period 19:00-7:00 (the next day). For example, the user with order number 10 placed the order at 9:26 and the car use time period is t2 off-peak.

[0042] (2) Determine the land use type and construct a land use type set L = {l1,l2,l3,l4,l5}, where l1 represents residential land, l2 represents office land, l3 represents commercial land, l4 represents transportation hub land, and l5 represents other land use types. Determine the land use type of each grid I;

[0043] (3) Calculate the number of reservation orders Q in the last 15 minutes for the grid I where the passenger is located s , which includes both successfully completed orders and cancelled or defaulted orders. For example, the number of scheduled orders for a user with order number 10 is Q s is 7;

[0044] (4) Calculate the number of completed orders Q in the last 15 minutes for the grid I where the passenger is located c For example, the user with order number 10 has the number of scheduled orders Q c is 5.

[0045] (5) Calculate the cumulative load, which refers to the ratio of the number of completed orders to the number of scheduled orders in the same area and time period (the last 15 minutes of the grid I where the passenger is located) for 7 consecutive days before the passenger places the order. The cumulative load is expressed as Where j represents the j days before the passenger uses the car, Q sj represents the number of reservation orders made by passengers during the same period j days before the date of use of the car. s , Q cj represents the number of orders completed in the same period j days before the passenger's car date Q c For example, the user with order number 10 has a load of 5 / 7 = 0.71 on that day. Assuming that the load of the previous 6 same periods is They are 0.83, 1, 0.92, 0.77, 0.87, and 0.73 respectively, so the cumulative load

[0046] Step 4: Establish a naive Bayes classifier for different categories. The key indicator calculation process of the classifier is as follows:

[0047] (1) Assume that there are 10,000 samples for training the classifier. The waiting time is known a priori. This data can be obtained by subtracting the order time from the boarding time of passengers who successfully complete the order. Define the feature attribute set D, which includes the passenger's car use time T, land type L, and the number of reservation orders Q. s 、Complete order quantity Q c , cumulative load degree G;

[0048] (2) For any passenger data, calculate the probability of being in a certain waiting time. There are 6000 samples with a waiting time of w1, 3000 samples with a waiting time of w2, and 1000 samples with a waiting time of w3. Then the probability of waiting for w1 is The probability of waiting for time w2 is The probability of waiting for time w3 is

[0049] (3) Calculate the discrete probability of different vehicle use periods T under each waiting time condition Indicates that in the training sample, the waiting time is w k The parking period is T i The sample size is shown in Table 4; the discrete probability of occurrence of land type Land under each waiting time condition is calculated. Indicates that in the training sample, the waiting time is w k The land use type in this state is L i The sample size is shown in Table 5; the number of appointment orders Q under each waiting time state is calculated s Discrete probability of occurrence Indicates that in the training sample, the waiting time is w k The number of reservation orders in this state is Q si The sample size is shown in Table 6; the number of completed orders Q under each waiting time condition is calculated c Discrete probability of occurrence Indicates that in the training sample, the waiting time is w k The number of completed orders in this state is Q ci The sample size is shown in Table 7; the discrete probability of the cumulative load G occurring under each waiting time state is calculated. Indicates that in the training sample, the waiting time is w k The cumulative load in the state is G i The sample size is shown in Table 8;

[0050] Table 4: Discrete probability of different car use periods T occurring under different waiting time conditions

[0051]

[0052]

[0053] Table 5: Discrete probability of occurrence of different land use types L under different waiting time conditions

[0054] L <![CDATA[P(L|w1)]]> <![CDATA[P(L|w2)]]> <![CDATA[P(L|w3)]]> Residential land 0.13 0.12 0.18 Office space 0.17 0.21 0.14 Commercial land 0.1 0.21 0.33 Transportation hub land 0.53 0.33 0.24 Other land 0.07 0.13 0.11

[0055] Table 6: Different order quantities Q under different waiting time conditions s Discrete probability of occurrence

[0056] <![CDATA[Q s ]]> <![CDATA[P(Q s |w1)]]> <![CDATA[P(Q s |w2)]]> <h2 style=";text-align:left;direction:ltr"><![CDATA[P(Q <h2 style=";text-align:left;direction:ltr"> s <h2 style=";text-align:left;direction:ltr"> |w3)]]><h2 style=";text-align:left;direction:ltr"> 1 0.022 0.016 0.001 2 0.011 0.031 0.003 3 0.042 0.029 0.011 4 0.054 0.041 0.025 5 0.110 0.085 0.044 6 0.157 0.133 0.072 7 0.153 0.171 0.055 8 0.087 0.097 0.075 9 0.065 0.053 0.045 10 0.054 0.069 0.065 11 0.046 0.041 0.071 12 0.035 0.045 0.096 13 0.053 0.035 0.084 14 0.044 0.065 0.12 15 0.026 0.045 0.075 16 0.009 0.015 0.056 17 0.011 0.008 0.014 18 0.014 0.012 0.028 19 0.005 0.003 0.026 20 0.002 0.006 0.034

[0057] Table 7: Number of completed orders Q under different waiting time conditions c Discrete probability of occurrence

[0058] <![CDATA[Q c ]]> <![CDATA[P(Q c |w1)]]> <![CDATA[P(Q c |w2)]]> <![CDATA[P(Q c |w2)]]> 1 0.008 0.011 0.005 2 0.011 0.014 0.007 3 0.024 0.027 0.018 4 0.044 0.047 0.034 5 0.092 0.08 0.068 6 0.143 0.131 0.107 7 0.164 0.122 0.109 8 0.116 0.174 0.11 9 0.091 0.084 0.068 10 0.050 0.061 0.058 11 0.076 0.041 0.064 12 0.028 0.046 0.062 13 0.047 0.033 0.065 14 0.022 0.034 0.071 15 0.037 0.048 0.051 16 0.013 0.025 0.035 17 0.014 0.011 0.014 18 0.011 0.008 0.02 19 0.005 0.002 0.015 20 0.004 0.001 0.019

[0059] Table 8: Discrete probability of different cumulative loads G occurring under different waiting time conditions

[0060]

[0061]

[0062] Step 5: Calculate the probability P that R belongs to each type of waiting time state i (w k |T i ,L i ,Q si ,Q ci ,G i ), indicating passenger i, who uses the vehicle for a period of T i , land use type is L i , the number of reservation orders is Q si , the number of completed orders is Q ci , the cumulative load is G i Under the condition of k Assume that there is a passenger order receiving data, and the calculation shows that the car use time period T is t2 off-peak period, the land type is l3 commercial land, and the number of reservation orders in the last 15 minutes of the grid I where the passenger is located is Q s is 7, and the number of completed orders is Q s is 5, and the cumulative load G is 4.2, then the calculation process is as follows:

[0063]

[0064]

[0065]

[0066] Where: P D =P(T i )·P(L i )·P(Q si )·P(Q ci )·P(G i ), P(T i ) indicates that the vehicle is used during the period T i The probability, P(L i ) indicates that the land use type is L i The probability, P(Q si ) indicates that the number of reservation orders is Q si The probability, P(Q ci ) indicates that the number of completed orders is Q ci The probability of P(Gi ) indicates the cumulative load is G i The other symbols have the same meanings as before.

[0067] Step 6: Compare P under different waiting time conditions i (w k |T i ,L i ,Q si ,Q ci ,G i ) value size, P i (w1|T i ,L i ,Q si ,Q ci ,G i ) is the largest, and the passenger’s waiting time status is finally determined to be w1.

[0068] Finally, it should be noted that the above examples are only used to illustrate the present invention and are not intended to limit the technical solutions described in the present invention. Therefore, although the present invention has been described in detail with reference to the above examples, those skilled in the art should understand that the present invention can still be modified or replaced by equivalents. All technical solutions and improvements that do not depart from the spirit and scope of the invention should be included in the scope of the claims of the present invention.

Claims

1. A method for dynamically predicting passenger waiting time using online car-hailing order data. It is characterized in that The following steps are involved: Step 1: Construct a set of passenger waiting time states W, and divide W into W = {w1,w2,w3,...w k }, where w k Indicates the waiting time. The larger the k value, the longer the waiting time. Step 2: Urban road grid division: divide the urban road network into M×N grid maps according to longitude and latitude, and the set of regions is expressed as: A={I 1×1 ,I 1×2 ,I 1×3 ...I 1×N ,I 2×1 ...I M×N } Where I represents a sub-region in the M×N region; Step 3: Preprocess the data. The data fields for receiving online car-hailing orders include the order number, the passenger's order time, the longitude and latitude of the passenger's car-hailing location, the longitude and latitude of the destination, and the order type. The single operation data records the complete information of the passenger's online car-hailing trip, including the passenger's order time, boarding time, alighting time, departure location, and arrival location. According to the unique order number of each record, the passenger's order time, boarding time, the longitude and latitude of the passenger's car-hailing location, and the departure location information are retained. If the order is not completed, only the longitude and latitude and the car-hailing time are retained. The boarding time minus the order time is the user's waiting time. According to the longitude and latitude of the passenger's car-hailing location, match it with the grid divided in step 1 to determine the grid I where each passenger is located. Step 3: Calculate the characteristic parameters, including the time period T, the type of land L, and the number of reservation orders Q in the last t minutes of the grid I where the passenger is located s 、Complete order quantity Q c And the cumulative load G, the process is as follows: (1) The car use period is used to represent the time characteristics of the user's car use. The car use period set T = {t1, t2, t3, t4} is constructed, and the 24 hours of the day are divided into four periods, t1 is the morning peak period, t2 is the off-peak period, t3 is the evening peak period, and t4 is the night period. The user's car use period T is determined according to the user's order time; (2) The land use type represents the land use attribute characteristics of the grid to which the user belongs. A land use type set L = {l1, l2, l3, l4, l5} is constructed, where l1 represents residential land, l2 represents office land, l3 represents commercial land, l4 represents transportation hub land, and l5 represents other land use types. The land use type of each grid I is determined; (3) The number of reservation orders in the last t minutes of the grid where the passenger is located is calculated by the number of all reservation orders Q within t minutes before the passenger places the order. s Indicates; including both successfully completed orders and cancelled or defaulted orders; (4) The number of reservation orders in the last t minutes of the grid where the passenger is located is calculated by the number of all completed orders Q within t minutes before the passenger places the order. c express; (5) Cumulative load refers to the ratio of the number of all completed orders to the number of all scheduled orders in the last t minutes of the grid I where the passenger is located in the same area and time period for J consecutive days before the passenger places the order; It describes the overall order completion degree of the area within t minutes before the passenger places an order. The cumulative load is expressed as Where j represents the j days before the passenger uses the car, Q sj represents the number of reservation orders made by passengers during the same period j days before the date of use of the car. s , Q cj represents the number of orders completed in the same period j days before the passenger's car date Q c ; Step 4: Establish a naive Bayes classifier for different categories. The key indicator calculation process of the classifier is as follows: (1) A certain amount of sample data R is used to train the classifier. The order completion status and waiting time are known a priori. The waiting time can be obtained by subtracting the order time from the boarding time of passengers who have successfully completed the order. The feature attribute set D is defined, which includes the passenger's car use time T, land type L, and the number of reservation orders Q. s 、Complete order quantity Q c , cumulative load degree G; (2) For any passenger data, calculate the probability of being in a certain waiting time Indicates the waiting time is w k The number of records, Q represents the total number of training samples; (3) Calculate the discrete probability of different vehicle use periods T under each waiting time condition Indicates that in the training sample, the waiting time is w k The parking period is T i The sample size is calculated; the discrete probability of land use type L occurring under each waiting time condition is calculated. Indicates that in the training sample, the waiting time is w k The land use type in this state is L i Sample size; calculate the number of appointment orders Q under each waiting time state s Discrete probability of occurrence Indicates that in the training sample, the waiting time is w k The number of reservation orders in this state is Q si Sample size; calculate the number of completed orders Q under each waiting time state c Discrete probability of occurrence Indicates that in the training sample, the waiting time is w k The number of completed orders in this state is Q ci The sample size is calculated; the discrete probability of the cumulative load G occurring under each waiting time state is calculated. Indicates that in the training sample, the waiting time is w k The cumulative load in the state is G i The sample size; Step 5: Calculate the probability P that R belongs to each type of waiting time state i (w k |T i ,L i ,Q si ,Q ci ,G i ), indicating passenger i, who uses the vehicle for a period of T i , land use type is L i , the number of reservation orders is Q si , the number of completed orders is Q ci , the cumulative load is G i Under the condition of k The probability of is calculated as follows: Where: P D Indicates that the car is used during the time period T i , land use type is L i , the number of reservation orders is Q si , the number of completed orders is Q ci , the cumulative load is G i The probability of P is constant for all categories. D =P(T i )·P(L i )·P(Q si )·P(Q ci )·P(G i ), P(T i ) indicates that the vehicle is used for a period of time T i The probability, P(L i ) indicates that the land use type is L i The probability, P(Q si ) indicates that the number of reservation orders is Q si The probability, P(Q ci ) indicates that the number of completed orders is Q ci The probability of P(G i ) indicates the cumulative load is G i probability; Step 6: Compare P under different waiting time conditions i (w k |T i ,L i ,Q si ,Q ci ,G i ) value, the largest one corresponds to the waiting time state w k The final determined passenger waiting time status.