A dynamic recommendation transfer method for public transportation smart travel services

Abstract

The present invention discloses a method for dynamically recommending transfers for smart public transportation travel services, which relates to the technical field of public transportation services, including: S1, establishing a set of public transportation routes; S2, matching overlapping stations with adjacent stations; S3, obtaining the user's starting point and end point and generating candidate paths; S4, calculating the transfer efficiency density and performing path pruning; S5, calculating the departure time rhythm interleaving coefficient between transfer nodes; S6, calculating the transfer chain risk accumulation function; S7, estimating the total time and waiting time; S8, constructing a comprehensive path score and recommendation ranking. The method for dynamically recommending transfers for smart public transportation travel services provided by the present invention is based on public transportation operation paths and station information, combined with adjacent station distance calculation and departure rhythm feature construction, and can more accurately identify the structural associations between transfer nodes, and realize path screening through the rhythm interleaving factor, effectively reducing the recommendation frequency of high-risk transfer paths, and improving the stability of the transfer process.

Description

Technical Field

[0001] The present invention relates to the technical field of public transportation services, and in particular to a method for dynamically recommending transfers for public transportation smart travel services. Background Art

[0002] With the continuous advancement of smart mobility and urban transportation informatization, public transportation route recommendation systems have become increasingly widespread and possess a high level of intelligence. Existing technologies often use methods such as geographic matching of starting and ending points, mining historical operational data, and analyzing route maps to provide users with multiple feasible public transportation routes. These systems also comprehensively consider key indicators such as the number of transfers, walking distance, and estimated arrival time.

[0003] To improve travel efficiency, some systems have incorporated parameters such as real-time bus arrival information and historical peak-hour operating patterns to optimize route selection or predict potential delays. These technologies have significantly improved recommendation accuracy and user experience, particularly in point-to-point commuting and regular route optimization, achieving promising results.

[0004] After searching, a Chinese patent (publication number: CN103578268B) discloses a location recommendation method based on public transportation routes, which includes the following steps: ① In an offline state, the system searches for all public transportation routes passing through each public transportation stop, then finds all points of interest along each public route in a geographic database and classifies the points of interest; the points of interest are also called associated locations of the public transportation stop; ② Also in an offline state, the system caches various associated locations of each public transportation stop, or each bus route, or each geographic area into the system's server memory; ③ The system filters associated locations based on the user's specified departure location, traverses nearby public transportation stops, and for each transportation stop, the system directly retrieves its associated locations from the memory cache and takes out the categories that meet the user's requirements; finally, given all associated locations that meet the user's requirements, the system sends the associated location that best meets the user's requirements to the user.

[0005] In practical applications, users still desire more personalized, dynamic, and risk-aware recommendations when faced with multiple feasible routes, multiple transfer nodes, or complex transfer environments. Therefore, there is still room for further refinement and technological advancement in areas such as path transfer node feature modeling, delay trend assessment, passenger flow status recognition, and user preference adaptation. Therefore, this paper proposes a method for dynamically recommending transfers for public transportation smart travel services. Summary of the Invention

[0006] The purpose of the present invention is to provide a method for dynamically recommending transfers for public transportation smart travel services to solve the problems mentioned in the above background technology.

[0007] The present invention can be implemented through the following technical solution: a method for dynamically recommending transfers for public transportation smart travel services, the method comprising the following steps:

[0008] S1. Collect the routes and stops of various types of public transportation and construct a bus route set including running time information;

[0009] S2, based on the station number or distance threshold, matching the overlapping stations with the adjacent stations and taking the walking time into account;

[0010] S3. Obtain the user's starting point and destination, and generate a set of candidate routes with a limited number of transfers;

[0011] S4. Calculate transfer efficiency density and eliminate inefficient routes below a preset density threshold;

[0012] S5. Calculate the departure time rhythm stagger coefficient. If it is lower than a preset coefficient threshold, reduce the path priority.

[0013] S6. Calculate the transfer chain risk accumulation function based on the vehicle delay probability, the standard deviation of the waiting time, and the congestion level of the current station to obtain the path confidence;

[0014] S7, estimate the total path time and waiting time;

[0015] S8. Construct a scoring function to rank and recommend paths based on total time consumption, interleaving coefficient, and confidence.

[0016] A further technical improvement of the present invention is that when S8 outputs the recommended routes, each recommended route is accompanied by the waiting time, the estimated total time and the number of transfers.

[0017] A further technical improvement of the present invention is that in step S6, the method for obtaining the standard deviation of the waiting time includes:

[0018] F1. Data collection:

[0019] Obtain the actual arrival time series of public transportation at a certain station (which can be obtained from GPS positioning, arrival records, or subway / bus card swiping records);

[0020] F2. Waiting time calculation:

[0021] For consecutive shifts, calculate the actual waiting time between each shift Where, and are two adjacent time points;

[0022] F3. Statistical processing:

[0023] Calculate the mean and standard deviation of the waiting time series.

[0024] A further technical improvement of the present invention is that in step S6, the vehicle delay probability is determined by extracting the running time of consecutive shifts between adjacent stations on the same route, judging the changing trend of this time, and comparing it with the historical peak period; if it shows an increasing trend and is during the peak period, the vehicle delay probability of the current node is adjusted.

[0025] A further technical improvement of the present invention is that the method for adjusting the vehicle delay probability comprises the following steps:

[0026] A1. Collect the continuous shift operation data of the current node:

[0027] Get the bus route where the current node is located, and get the running time sequence between the adjacent stations of the latest n consecutive trips on the route ;

[0028] A2. Calculate the running time trend:

[0029] Based on the running time sequence , use sliding difference or linear regression fitting to obtain the current time-consuming trend slope ;

[0030] A3. Comparison with historical peak periods:

[0031] Get the current system timestamp , and judge Whether it belongs to the predefined historical peak period set H;

[0032] like >0, and ∈H, then it is judged that there is a delay trend;

[0033] A4. Delay Probability Adjustment:

[0034] By formula: , obtain the adjusted vehicle delay probability .

[0035] The further technical improvement of the present invention is that the delay trend adjustment coefficient The methods of obtaining include:

[0036] Conduct trend analysis on the running time of multiple consecutive shifts between adjacent stations on the same route and calculate the changing trend of running time in the current time period;

[0037] The historical operating trend mean and fluctuation range of the site in the same time period are extracted to determine the degree of deviation between the current trend and the historical mean;

[0038] A normalized index is constructed based on the degree of deviation and mapped to a delay trend adjustment coefficient through a nonlinear response function, so that the more significant the trend and the greater the deviation, the higher the corresponding adjustment coefficient.

[0039] A further technical improvement of the present invention is that: congestion level The acquisition method is based on the total passenger flow data of the station, combined with the historical direction distribution, and the passenger flow is divided according to the bus direction. The different lines in the same direction are then divided again, and the number of passengers in the corresponding direction of each line is estimated. The probability of successfully boarding the bus in this direction is determined in combination with the vehicle capacity.

[0040] A further technical improvement of the present invention is to generate a congestion level The specific steps include:

[0041] Z1. Get the total number of passengers at the current transfer node ;

[0042] Z2, dividing the flow of people based on the direction of bus travel;

[0043] Get all possible departure directions and corresponding line sets L of the corresponding station, and calculate the total number of passengers The distribution according to the departure direction is as follows: Where, Estimated number of passengers for the current time period and direction; is a direction prediction function based on historical proportions, weather, holidays and other factors; H is the set of historical peak hours;

[0044] Z3. Perform secondary line division:

[0045] If there are multiple bus routes in the same direction, the number of passengers is further divided by route using the formula: = , get the estimated number of passengers on the jth bus route in the i-th direction ;

[0046] Z4. Determine the probability of a successful ride:

[0047] For the specific routes and directions involved in the candidate routes, the number of people that can be accommodated by the next bus If the estimated number of passengers is >Capacity , it is judged as "high congestion risk" and the probability of successfully taking the bus is low;

[0048] Z5. Constructing congestion levels ;

[0049] By formula: definition .

[0050] A further technical improvement of the present invention is that the step S8 divides the recommendation order into the following two categories based on the user's travel preference settings:

[0051] o1. Efficiency priority mode: suitable for users who want to shorten their commute time;

[0052] Adopting the original comprehensive scoring function, the routes are sorted from high to low according to the scores, and the top N routes are recommended to the user;

[0053] o2. Comfort Priority Mode: Suitable for users who want to reduce waiting time, avoid crowds, shorten walking distances, or travel in bad weather. Specifically, it includes:

[0054] Y1. Obtain the current environmental status parameters, including weather conditions, temperature level, and severe weather indicator, for dynamic setting of subsequent transfer node distance thresholds and shielding conditions;

[0055] Y2. For each path in the candidate path set, obtain the following technical parameters:

[0056] Estimated transfer preparation time for each transfer node;

[0057] Congestion level indicator for each transfer node;

[0058] The walking distance between stations for each transfer segment;

[0059] Whether there is an indoor transfer channel sign (whether the station has an indoor transfer structure is obtained based on the station structure database and map annotation information. The system stores the structural attributes of each station for path evaluation);

[0060] And screen according to the following technical judgments:

[0061] If the transfer preparation time is less than the preset comfort time threshold, it is excluded;

[0062] If the congestion level exceeds the preset upper threshold, it is excluded;

[0063] If the weather is bad and the transfer distance exceeds the dynamic distance threshold, it will be excluded;

[0064] Except for inclement weather, if the total transfer walking distance exceeds a fixed threshold, it will be excluded;

[0065] Y3. Preserve path set sorting processing:

[0066] Perform a structured comparison on the remaining path sets based on the following data fields, selecting in order:

[0067] Priority will be given to those with fewer transfer nodes;

[0068] Those with longer total transfer preparation time will be given priority;

[0069] The one with the smaller maximum congestion level will be given priority;

[0070] Routes with indoor transfer sections are given priority;

[0071] When multiple conditions coexist, the structured sort instruction is called to perform level-by-level comparison;

[0072] Y4. Generate path recommendation results:

[0073] The remaining route set after processing is used as the recommendation result of the "comfort priority" mode and provided to users for selection. The transfer time, walking distance and transfer environment structure of each transfer segment are synchronously annotated in the route information.

[0074] In addition, the path sorting is based on the structured path data table, and the field priority is sorted in sequence according to the number of transfers, maximum congestion level, and transfer preparation time fields. The system program performs automatic sorting without human intervention.

[0075] Compared with the prior art, the present invention has the following beneficial effects:

[0076] The method for dynamically recommending transfers for smart public transportation travel services provided by this invention is based on public transportation routes and station information, combined with adjacent station distance calculation and departure rhythm feature construction. It can more accurately identify the structural associations between transfer nodes and implement path screening through the rhythm interleaving factor, effectively reducing the frequency of recommending high-risk transfer routes and improving the stability of the transfer process.

[0077] Furthermore, by introducing a transfer risk assessment mechanism and a dynamic delay probability adjustment method, the system can perceive factors such as bus schedule trends, congestion levels, and transfer success probabilities in real time, construct a path-level transfer risk function, and perform comprehensive sorting and pruning of multiple paths, ensuring that recommended paths have higher practical accessibility and timeliness.

[0078] In addition, the present invention further introduces a user preference recognition and environmental adaptation mechanism, which can dynamically adjust the path screening and recommendation logic based on the efficiency priority or comfort priority mode selected by the user, and realize differentiated recommendations of transfer paths under peak congestion, bad weather or complex structure station conditions, meeting diverse actual travel needs and enhancing the flexibility and intelligence level of the path recommendation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] To facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0080] Figure 1 This is a flow chart of the method of Example 1 of the present invention. DETAILED DESCRIPTION

[0081] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0082] Example 1

[0083] See also Figure 1 As shown, the present invention provides a method for dynamically recommending transfers for public transportation smart travel services, comprising the following steps:

[0084] S1. Create a bus route set:

[0085] Collect the formal paths and stops corresponding to different types of public transportation (such as bus, subway, BRT) in the corresponding area, and construct the bus route set R={ }, and each line includes all its stations and actual running time information;

[0086] S2. Match overlapping sites with adjacent sites:

[0087] Compare stops between different bus route sets:

[0088] If two routes include exactly the same station number or the GPS position difference is less than the preset distance threshold, they are marked as "coincident stations";

[0089] If the distance between two sites in map space is within the range threshold, they are marked as "adjacent sites" and their walking distance is included in the calculation of additional transfers;

[0090] Specifically, each bus route establishes a path list through the station number and corresponding geographic coordinates, and uses distance threshold or spatial topology matching to identify overlapping stations and adjacent stations between different routes;

[0091] S3. Obtain the user's starting point and end point and generate candidate paths:

[0092] Get the user's current GPS location and destination station, and build all accessible paths based on the station connection relationship (not exceeding the maximum number of transfers K). Each path includes the transfer point, the route and the estimated running time, and the path set P = { };

[0093] S4. Calculate the transfer efficiency density and perform path pruning:

[0094] Constructing transfer station efficiency density matrix :

[0095] ;

[0096] If there are m nodes in a path < preset density threshold, the path is marked as an “inefficient path” and removed from the path set P to obtain a path subset P;

[0097] m is a positive integer greater than 0;

[0098] Specifically, the rhythm stagger coefficient is linearly normalized and then introduced into the scoring factor to ensure stability under different site densities, and its boundary is defined as [0,1]. The higher the value, the more uniform the rhythm.

[0099] S5. Calculate the staggered coefficient of departure time rhythm between transfer nodes:

[0100] For each path ∈ Get the continuous transfer lines in:

[0101] s51, historical departure time series A={ , ,..., } and B={ , ,..., };

[0102] s52. Calculate the interleaving coefficient α:

[0103] ;

[0104] If α is less than the preset coefficient threshold, it is considered as rhythm disharmony and the path score decreases;

[0105] Where, A reasonable transfer time interval (e.g. 3-10 minutes);

[0106] It is an indicator function: it takes the value 1 when the condition is met, otherwise it takes the value 0;

[0107] For the smaller number of values ​​in the historical departure time series A and B, this ensures a reasonable comparison of transfer rhythms and avoids deviations caused by differences in data volume;

[0108] To count how many pairs ( , ) satisfies that the transfer time is within an acceptable range, i.e. the next train after the user arrives is about to depart;

[0109] The value range of α is [0,1]. The higher the value, the more matched the transfer rhythm is, and the more reasonable the waiting time for users after arriving from the previous train is.

[0110] For example, if A={ , , },B={ , , }, transfer time window for min;

[0111] If satisfied ∈ The combinations are:

[0112] ( , ): Five minutes, ( , ): Five minutes, ( , ): 10 minutes;

[0113] There are three groups of satisfaction. =3, =3, so ;

[0114] S6. Calculate the transfer chain risk accumulation function:

[0115] For each path, calculate the risk value of each transfer station :

[0116] ;

[0117] Where, 、 、 are the weight coefficients of each risk factor, which can be adjusted or obtained through training according to actual conditions;

[0118] is the vehicle delay probability at the current node (obtained based on the on-time rate statistics in each historical period);

[0119] Vehicle delay probability By extracting the running time between adjacent stops of consecutive trains on the same route, we determine the changing trend of this time and compare it with the historical peak hours. If the trend is increasing and it is during the peak hours, we adjust the vehicle delay probability at the current node. Specifically, the following steps are involved:

[0120] A1. Collect the continuous shift operation data of the current node:

[0121] Get the bus route where the current node is located, and get the running time sequence between the adjacent stations (i.e. the current station and the previous or next station) of the latest n consecutive trips on the route ={ , ,..., },in, is the actual time taken by the i-th shift to run at adjacent stations;

[0122] A2. Calculate the running time trend indicator:

[0123] Based on the running time sequence , use sliding difference or linear regression fitting to obtain the current time-consuming trend slope ,Right now , indicating whether the running time is increasing;

[0124] A3. Comparison with historical peak periods:

[0125] Get the current system timestamp , and judge Whether it belongs to the predefined historical peak period set H={[ ],[ ],...[ ]}; where, represents the starting time of the kth peak period; represents the end time of the kth peak period;

[0126] like >0, and ∈H, then it is judged that there is a delay trend and the corresponding path is at the peak period;

[0127] A4. Delay Probability Adjustment:

[0128] By introducing the delay trend adjustment factor ∈ (0,1), and the adjusted vehicle delay probability is obtained by the following formula :

[0129] Where, is the historical mean value of the corresponding operating period;

[0130] Delay Trend Adjustment Factor The methods of obtaining include:

[0131] Conduct trend analysis on the running time of multiple consecutive shifts between adjacent stations on the same route and calculate the changing trend of running time in the current time period;

[0132] The historical operating trend mean and fluctuation range of the site in the same time period are extracted to determine the degree of deviation between the current trend and the historical mean;

[0133] A normalized index is constructed based on the degree of deviation and mapped to a delay trend adjustment coefficient through a nonlinear response function, so that the more significant the trend and the greater the deviation, the higher the corresponding adjustment coefficient;

[0134] In the calculation, the comparison period covers at least 10 consecutive shifts, and a sliding average mechanism is introduced to smooth out abnormal fluctuations and avoid prediction bias caused by too few samples;

[0135] is the standard deviation of waiting time, which is used to measure the uncertainty of waiting time. The acquisition methods include:

[0136] F1. Data collection:

[0137] Obtain the actual arrival time series of public transportation at a certain station (which can be obtained from GPS positioning, arrival records, or subway / bus card swiping records);

[0138] F2. Waiting time calculation:

[0139] For consecutive shifts, calculate the actual waiting time between each shift Where, and are two adjacent time points;

[0140] F3. Statistical processing:

[0141] Calculate the mean and standard deviation of the waiting time series ;

[0142] The congestion level of the current station, derived from historical passenger flow data or real-time sensor data;

[0143] Risk value based on each transfer station Perform path chain risk accumulation to obtain the overall successful transfer confidence of the corresponding path ;

[0144] ;

[0145] Where n is the number of transfer nodes in the path, that is, the number of transfers; is the cumulative failure risk of the entire path;

[0146] For example, there are three transfer points, and their risk values ​​are: =0.1, =0.1, =0.1;

[0147] but = (1-0.1) (1-0.2) (1-0.3)=0.9 0.8 0.7=0.504;

[0148] =1-0.504=0.496;

[0149] =0.504;

[0150] The confidence level for this path is 50.4%, indicating that there is approximately a 50% chance that a transfer node will have a problem.

[0151] S7. Estimated total time and waiting time:

[0152] For each retained path p∈pathsubset , respectively calculate:

[0153] Total driving time ;

[0154] Set different transfer preparation times based on the estimated time to reach the transfer node and whether the stations are overlapping or adjacent. ;

[0155] Combine the latest arrival time of each line to get the waiting time ;

[0156] S8. Comprehensive scoring and recommended ranking of construction paths:

[0157] Construct a comprehensive scoring function for each path and obtain the recommendation score S:

[0158] ;

[0159] Where, 、 、 are the weight parameters of the corresponding factors, which can be adjusted through training or experience;

[0160] Specifically, is the time efficiency item, which means the less the total time consumed, the higher the score;

[0161] For the rhythm matching item, if the transfer rhythm is highly coordinated (for example, the departure interval is reasonable), the score will be improved;

[0162] is the path stability item. The higher the path confidence, the greater the probability of successful transfer and the higher the score.

[0163] Recommend the top N paths from high to low recommendation score S to the user;

[0164] When outputting recommended routes, each recommended route is accompanied by the waiting time, estimated total time and number of transfers.

[0165] Example 2

[0166] A method for dynamically recommending transfers for a public transportation smart travel service, the method comprising the following steps:

[0167] S1. Create a bus route set:

[0168] Collect the formal paths and stops corresponding to different types of public transportation (such as bus, subway, BRT) in the corresponding area, and construct the bus route set R={ }, and each line includes all its stations and actual running time information;

[0169] S2. Match overlapping sites with adjacent sites:

[0170] Compare stops between different bus route sets:

[0171] If two routes include exactly the same station number or the GPS position difference is less than the preset distance threshold, they are marked as "coincident stations";

[0172] If the distance between two sites in map space is within the range threshold, they are marked as "adjacent sites" and their walking distance is included in the calculation of additional transfers;

[0173] Specifically, each bus route establishes a path list through the station number and corresponding geographic coordinates, and uses distance threshold or spatial topology matching to identify overlapping stations and adjacent stations between different routes;

[0174] S3. Obtain the user's starting point and end point and generate candidate paths:

[0175] Get the user's current GPS location and destination station, and build all accessible paths based on the station connection relationship (not exceeding the maximum number of transfers K). Each path includes the transfer point, the route and the estimated running time, and the path set P = { };

[0176] S4. Calculate the transfer efficiency density and perform path pruning:

[0177] Constructing transfer station efficiency density matrix :

[0178] ;

[0179] If there are m nodes in a path < preset density threshold, the path is marked as an “inefficient path” and removed from the path set P to obtain a path subset P;

[0180] m is a positive integer greater than 0;

[0181] Specifically, the rhythm stagger coefficient is linearly normalized and then introduced into the scoring factor to ensure stability under different site densities, and its boundary is defined as [0,1]. The higher the value, the more uniform the rhythm.

[0182] S5. Calculate the staggered coefficient of departure time rhythm between transfer nodes:

[0183] For each path ∈ Get the continuous transfer lines in:

[0184] s51, historical departure time series A={ , ,..., } and B={ , ,..., };

[0185] s52. Calculate the interleaving coefficient α:

[0186] ;

[0187] If α is less than the preset coefficient threshold, it is considered as rhythm disharmony and the path score decreases;

[0188] Where, A reasonable transfer time interval (e.g. 3-10 minutes);

[0189] It is an indicator function: it takes the value 1 when the condition is met, otherwise it takes the value 0;

[0190] For the smaller number of values ​​in the historical departure time series A and B, this ensures a reasonable comparison of transfer rhythms and avoids deviations caused by differences in data volume;

[0191] To count how many pairs ( , ) satisfies that the transfer time is within an acceptable range, i.e. the next train after the user arrives is about to depart;

[0192] The value range of α is [0,1]. The higher the value, the more matched the transfer rhythm is, and the more reasonable the waiting time for users after arriving from the previous train is.

[0193] For example, if A={ , , },B={ , , }, transfer time window for min;

[0194] If satisfied ∈ The combinations are:

[0195] ( , ): Five minutes, ( , ): Five minutes, ( , ): 10 minutes;

[0196] There are three groups of satisfaction. =3, =3, so ;

[0197] S6. Calculate the transfer chain risk accumulation function:

[0198] For each path, calculate the risk value of each transfer station :

[0199] ;

[0200] Where, 、 、 are the weight coefficients of each risk factor, which can be adjusted or obtained through training according to actual conditions;

[0201] is the vehicle delay probability at the current node (which is obtained based on the on-time rate statistics in each historical period). The adjustment method includes the following steps:

[0202] A1. Collect the continuous shift operation data of the current node:

[0203] Get the bus route where the current node is located, and get the running time sequence between the adjacent stations (i.e. the current station and the previous or next station) of the latest n consecutive trips on the route ={ , ,..., },in, is the actual time taken by the i-th shift to run at adjacent stations;

[0204] A2. Calculate the running time trend indicator:

[0205] Based on the running time sequence , use sliding difference or linear regression fitting to obtain the current time-consuming trend slope ,Right now , indicating whether the running time is increasing;

[0206] A3. Comparison with historical peak periods:

[0207] Get the current system timestamp , and judge Whether it belongs to the predefined historical peak period set H={[ ],[ ],...[ ]}; where, represents the starting time of the kth peak period; represents the end time of the kth peak period;

[0208] like >0, and ∈H, then it is judged that there is a delay trend and the corresponding path is at the peak period;

[0209] A4. Delay Probability Adjustment:

[0210] By introducing the delay trend adjustment factor ∈ (0,1), and the adjusted vehicle delay probability is obtained by the following formula :

[0211] Where, is the historical mean value of the corresponding operating period;

[0212] is the standard deviation of waiting time, which is used to measure the uncertainty of waiting time;

[0213] The congestion level of the current station is usually derived from historical passenger flow data or real-time sensor data. The acquisition method is based on the total passenger flow data of the station, combined with the historical direction distribution, and the passenger flow is divided by bus direction. The different routes within the same direction are then divided again. The number of passengers in the corresponding direction of each route is estimated, and the probability of successfully boarding the bus in that direction is determined based on the vehicle capacity. The specific steps include:

[0214] Z1. Get the total number of passengers at the current transfer node ;

[0215] Total number of people The source can be obtained through real-time sensors (such as cameras, people counting radars, thermal infrared), historical ride check-in data, or prediction model data (such as holiday traffic forecasts);

[0216] Z2, dividing the flow of people based on the direction of bus travel;

[0217] Get all possible departure directions and corresponding line sets of the corresponding station L={ };

[0218] And based on historical statistical data or direction recognition algorithm, the total number of people The distribution according to the departure direction is as follows: ;

[0219] Where, Estimated number of passengers for the current time period and direction;

[0220] This is a direction prediction function built based on factors such as historical proportions, weather, and holidays, used to calculate passenger flow in a specific direction at a bus stop. Specifically, the historical proportion factor predicts the flow ratio in each direction by analyzing past passenger data. The weather factor adjusts the flow forecast by obtaining real-time weather conditions (such as precipitation and temperature). The holiday factor adjusts the flow forecast by identifying changes in passenger demand on holidays and special dates.

[0221] H is the historical peak hour set;

[0222] Z3. Perform secondary line division:

[0223] If there are multiple bus routes in the same direction (such as Shared stations), further divided by line, using the formula: = Get the estimated number of passengers on the jth bus route in the i-th direction ;

[0224] Z4. Determine the probability of a successful ride:

[0225] For the specific routes and directions involved in the candidate routes, the number of people that can be accommodated by the next bus If the estimated number of passengers is >Capacity , it is judged as "high congestion risk" and the probability of successfully taking the bus is low;

[0226] Z5. Constructing congestion levels ;

[0227] By formula definition , specifically, The indicator range is [0,1]. The closer to 1, the more congested it is and the more difficult it is to take the bus.

[0228] Risk value based on each transfer station Perform path chain risk accumulation to obtain the overall successful transfer confidence of the corresponding path ;

[0229] ;

[0230] Where n is the number of transfer nodes in the path, that is, the number of transfers; is the cumulative failure risk of the entire path;

[0231] For example, there are three transfer points, and their risk values ​​are: =0.1, =0.1, =0.1;

[0232] but = (1-0.1) (1-0.2) (1-0.3)=0.9 0.8 0.7=0.504;

[0233] =1-0.504=0.496;

[0234] =0.504;

[0235] The confidence level for this path is 50.4%, indicating that there is approximately a 50% chance that a transfer node will have a problem.

[0236] S7. Estimated total time and waiting time:

[0237] For each retained path p∈pathsubset , respectively calculate:

[0238] Total driving time ;

[0239] Set different transfer preparation times based on the estimated time to reach the transfer node and whether the stations are overlapping or adjacent. ;

[0240] Combine the latest arrival time of each line to get the waiting time ;

[0241] S8. Based on the user's travel preference settings, the recommendation order is divided into the following two categories:

[0242] o1. Efficiency priority mode: suitable for users who want to shorten their commute time;

[0243] Using the original comprehensive scoring function: ;

[0244] Sort by scores from high to low and recommend the top N paths to the user;

[0245] o2. Comfort Priority Mode: Suitable for users who want to reduce waiting time, avoid crowds, shorten walking distances, or travel in bad weather. Specifically, it includes:

[0246] Y1. Obtain the current environmental status parameters, including weather conditions, temperature level, and severe weather indicator, for dynamic setting of subsequent transfer node distance thresholds and shielding conditions;

[0247] Y2. For each path in the candidate path set, obtain the following technical parameters:

[0248] Estimated transfer preparation time for each transfer node;

[0249] Congestion level indicator for each transfer node;

[0250] The walking distance between stations for each transfer segment;

[0251] Whether there is an indoor transfer channel sign (whether the station has an indoor transfer structure is obtained based on the station structure database and map annotation information. The system stores the structural attributes of each station for path evaluation);

[0252] And screen according to the following technical judgments:

[0253] If the transfer preparation time is less than the preset comfort time threshold, it is excluded;

[0254] If the congestion level exceeds the preset upper threshold (e.g. 0.6), it is excluded;

[0255] If the weather is bad and the transfer distance exceeds the dynamic distance threshold (e.g. 200 meters), then the transfer is excluded;

[0256] If the total transfer walking distance exceeds a fixed threshold (e.g., 500 meters) except for inclement weather, then it is excluded;

[0257] Y3. Preserve path set sorting processing:

[0258] Perform a structured comparison on the remaining path sets based on the following data fields, selecting in order:

[0259] Priority will be given to those with fewer transfer nodes;

[0260] Those with longer total transfer preparation time will be given priority;

[0261] The one with the smaller maximum congestion level will be given priority;

[0262] Routes with indoor transfer sections are given priority;

[0263] When multiple conditions coexist, the structured sort instruction is called to perform level-by-level comparison;

[0264] Y4. Generate path recommendation results:

[0265] The remaining route set after processing is used as the recommendation result of the "comfort priority" mode and provided to users for selection. The transfer time, walking distance and transfer environment structure of each transfer segment are synchronously annotated in the route information.

[0266] In addition, the path sorting is based on the structured path data table, and the field priority is sorted in sequence according to the number of transfers, maximum congestion level, and transfer preparation time fields. The system program performs automatic sorting without human intervention.

[0267] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any form. Although the present invention has been disclosed as above in terms of preferred embodiments, they are not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A method for dynamically recommending transfers for public transportation smart travel services, characterized in that: The method comprises the following steps: S1. Collect the routes and stops of various types of public transportation and construct a bus route set including running time information; S2, based on the station number or distance threshold, matching the overlapping stations with the adjacent stations and taking the walking time into account; S3. Obtain the user's starting point and destination, and generate a set P of candidate routes with a limited number of transfers; S4. Calculate transfer efficiency density and eliminate inefficient routes below a preset density threshold, including: Constructing transfer station efficiency density matrix : ； If there are m nodes in a path < preset density threshold, the path is marked as an "inefficient path" and removed from the path set P to obtain a path subset P; m is a positive integer greater than 0; S5. Calculate the departure time rhythm stagger coefficient. If it is lower than the preset coefficient threshold, reduce the path priority, including: For each path ∈ Get the continuous transfer lines in: s51, historical departure time series A={ , ,..., } and B={ , ,..., }; s52. Calculate the interleaving coefficient α: ； Where, Reasonable transfer time interval; It is an indicator function: it takes the value 1 when the condition is met, otherwise it takes the value 0; The smaller value in the historical departure time series A and B; To count how many pairs ( , ) The time difference between the two trains satisfies the transfer time to be within an acceptable range; S6. Probability of passing vehicle delay , standard deviation of waiting time and the congestion level of the current station , for each path, calculate the risk value of each transfer station : Where, 、 、 are the weight coefficients of each risk factor; Risk value based on each transfer station Perform path chain risk accumulation to obtain the overall successful transfer confidence of the corresponding path ; ; n is the number of transfer nodes in the path; S7, estimate the total path time and waiting time; S8. Construct a scoring function to rank and recommend paths based on total time consumption, interleaving coefficient, and confidence.

2. A method for dynamically recommending transfers for public transportation smart travel services according to claim 1, characterized in that: When S8 outputs recommended routes, it provides each recommended route with the waiting time, estimated total time, and number of transfers.

3. A method for dynamically recommending transfers for public transportation smart travel services according to claim 1, characterized in that: In step S6, the vehicle delay probability is determined by extracting the running time of consecutive shifts between adjacent stations on the same route, judging the changing trend of this time, and comparing it with the historical peak period; if it shows an increasing trend and is during the peak period, the vehicle delay probability of the current node is adjusted.

4. A method for dynamically recommending transfers for public transportation smart travel services according to claim 3, characterized in that: The method for adjusting the vehicle delay probability includes the following steps: A1. Collect the continuous shift operation data of the current node: Get the bus route where the current node is located, and get the running time sequence between the adjacent stations of the latest n consecutive trips on the route ; A2. Calculate the running time trend: Based on the running time sequence , use sliding difference or linear regression fitting to obtain the current time-consuming trend slope ; A3. Comparison with historical peak periods: Get the current system timestamp , and judge Whether it belongs to the predefined historical peak period set H; like >0, and ∈H, then it is judged that there is a delay trend; A4. Adjust the vehicle delay probability based on the preset delay trend adjustment coefficient.

5. A method for dynamically recommending transfers for public transportation smart travel services according to claim 4, characterized in that: Methods for obtaining the delay trend adjustment coefficient include: Conduct trend analysis on the running time of multiple consecutive shifts between adjacent stations on the same route and calculate the changing trend of running time in the current time period; The historical operating trend mean and fluctuation range of the site in the same time period are extracted to determine the degree of deviation between the current trend and the historical mean; A normalized index is constructed based on the degree of deviation and mapped into a delay trend adjustment coefficient through a nonlinear response function.

6. A method for dynamically recommending transfers for public transportation smart travel services according to claim 4, characterized in that: The congestion level is obtained based on the total passenger flow data at the station. Combined with the historical direction distribution, the passenger flow is divided according to the bus direction. The different lines in the same direction are then divided again. The number of passengers in the corresponding direction of each line is estimated, and the probability of successfully boarding the bus in that direction is determined based on the vehicle capacity.

7. A method for dynamically recommending transfers for public transportation smart travel services according to claim 6, characterized in that: The specific steps to generate the congestion level include: Z1. Get the total number of passengers at the current transfer node ; Z2, dividing the flow of people based on the direction of bus travel; Get all possible departure directions and corresponding line sets L of the corresponding station, and calculate the total number of passengers The distribution according to the departure direction is as follows: Where, Estimated number of passengers for the current time period and direction; is a pre-built direction prediction function; H is the historical peak period set; Z3. Perform secondary line division: If there are multiple bus routes in the same direction, the number of passengers is further divided by route using the formula: = , get the estimated number of passengers on the jth bus route in the i-th direction ; Z4. Determine the probability of a successful ride: For the specific routes and directions involved in the candidate routes, the number of people that can be accommodated by the next bus If the estimated number of passengers is >Capacity , it is judged as "high congestion risk" and the probability of successfully boarding the bus is low; Z5. Constructing congestion levels ; By formula: definition .

8. The method for dynamically recommending transfers for public transportation smart travel services according to claim 1, characterized in that: Based on user travel preference settings, the recommendation order is divided into the following two categories: o1. Efficiency priority mode; o2. Comfort priority mode.

9. A method for dynamically recommending transfers for public transportation smart travel services according to claim 8, characterized in that: The efficiency priority mode adopts the method in step S8.

10. A method for dynamically recommending transfers for public transportation smart travel services according to claim 9, characterized in that: Comfort priority mode includes: Y1. Obtain the current environmental status parameters, including weather conditions, temperature level, and whether there is severe weather; Y2. For each path in the candidate path set, obtain the following technical parameters: Estimated transfer preparation time for each transfer node; Congestion level indicator for each transfer node; The walking distance between stations for each transfer segment; Whether there are signs for indoor transfer passages; And screen according to the following technical judgments: If the transfer preparation time is less than the preset comfort time threshold, it is excluded; If the congestion level exceeds the preset upper threshold, it is excluded; If the weather is bad and the transfer distance exceeds the dynamic distance threshold, it will be excluded; If the total transfer walking distance exceeds a fixed threshold except for inclement weather, it is excluded; Y3. Preserve path set sorting processing: Perform a structured comparison on the remaining path sets based on the following data fields, selecting in order: Priority will be given to those with fewer transfer nodes; Those with longer total transfer preparation time will be given priority; The one with the smaller maximum congestion level will be given priority; Routes with indoor transfer sections are given priority; When multiple conditions coexist, the structured sort instruction is called to perform level-by-level comparison; Y4. Generate path recommendation results: The path set retained after processing is used as the recommendation result of the "comfort priority" mode.

Images