A short-term prediction method for inbound passenger flow in a service area based on operator data
Through operator data analysis of individual and group travel characteristics, a prediction model for entering the area passenger flow is constructed, which solves the prediction lag problem caused by relying on historical data in the existing technology, and realizes accurate and timely prediction of the passenger flow in the service area. Improve management efficiency and passenger experience.
Patent Information
- Application Number
- CN202411239852.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-05
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2044-09-05
AI Technical Summary
The existing methods for predicting passenger flows in the service area rely too much on historical data, making it difficult to accurately predict the demand for passenger flows in the short term, especially in case of sharp changes, and it is difficult to respond quickly.
Short-term prediction method for passenger flows entering the zone based on operator data, by analyzing individual historical travel characteristics and group travel data, a prediction model for passenger flows entering the zone is constructed, and the real-time passenger flow prediction is used to use individual real-time travel data, combining the shortest path algorithm and linearity The difference method accurately calculates the longitude, latitude and arrival time of the trajectory point, and constructs a bilayer linear regression model for prediction.
It improves the accuracy of short-term passenger flow forecasts, enhances the responsiveness of service area management, ensures efficient utilization of resource allocation, and improves passenger experience and operational efficiency.
Smart Images

Figure CN118761513B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of expressway operation management, and in particular to a method for short-term prediction of passenger flow entering a service area based on operator data. Background Art
[0002] Short-term passenger flow forecasting in service areas plays a key role in optimizing resource allocation, enhancing customer experience, improving operational efficiency, ensuring safety, and assisting decision-making. With the rapid development of big data and artificial intelligence technologies, the accuracy and real-time performance of short-term passenger flow forecasting will be further improved, providing more convenient and efficient solutions for the operation and management of service areas.
[0003] At present, the research methods for predicting the passenger flow into the service area mainly focus on the video surveillance system, using video recognition technology to estimate the passenger flow in a specific period of time, and further predicting the passenger flow with the help of sequence prediction techniques such as time smoothing. However, these methods rely too much on historical data and ignore the potential impact of road passenger flow characteristics on the passenger flow into the area. In addition, since these methods are mainly based on historical data for prediction, their prediction results often have lags, which is particularly disadvantageous in situations that require rapid response (such as passenger flow management in emergencies). For example, during holidays, when the passenger flow upstream of the service area increases sharply, the service area managers need to quickly formulate a response strategy for the potential increase in passenger flow into the area in the next half hour, one hour, etc. In this case, the prediction method that relies only on historical data is often difficult to accurately predict the passenger flow demand in the short term in the future.
[0004] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention
[0005] In view of the problems in the related art, the present invention proposes a short-term prediction method for the passenger flow entering a service area based on operator data to overcome the above-mentioned technical problems existing in the existing related art.
[0006] To this end, the specific technical solution adopted by the present invention is as follows:
[0007] A short-term prediction method for inbound passenger flow in a service area based on operator data, the method comprising the following steps:
[0008] S1. Analyze individual historical travel characteristics based on the acquired individual historical operator data, where the individual historical travel characteristics include individual actual path, individual time information and individual entry frequency;
[0009] S2. Analyze group travel data using individual historical travel characteristics, and build an inbound passenger flow prediction model based on the group travel data analysis results;
[0010] S3. Use individual real-time travel data to obtain the real-time passenger flow corresponding to each road section, and input the real-time passenger flow into the inbound passenger flow prediction model to obtain the short-term passenger flow in the service area.
[0011] Further, according to the acquired individual historical operator data, analyzing the individual historical travel characteristics, wherein the individual historical travel characteristics include the individual actual path, the individual time information and the individual entry frequency, comprises the following steps:
[0012] S11, converting the acquired individual historical operator data into a trajectory point set, and cleaning the trajectory points in the trajectory point set;
[0013] S12, based on the constructed road network structure, calculate the probability of mapping the cleaned trajectory points to each road section in the road network structure, select the road section with the highest probability, and sort the selected road sections in time according to the time sequence of the trajectory points;
[0014] S13, using the shortest path algorithm to process the time-sorted road sections to obtain individual actual paths;
[0015] S14, using a linear difference method to calculate the individual time information of the trajectory point arriving at each node along the individual actual path;
[0016] S15, matching the position of the cleaned trajectory point with the known service area position, identifying the trajectory point of the individual entering the service area, and obtaining the individual entry frequency.
[0017] Furthermore, the obtained individual historical operator data is used as a track point set, and the track points in the track point set are cleaned, including the following steps:
[0018] S111, converting the acquired individual historical operator data into a trajectory point set, and arranging the trajectory points in the trajectory point set in ascending order according to arrival time;
[0019] S112, initializing the set, taking the first trajectory point after sorting as the starting point of the set, and searching for trajectory points according to the preset arrival time difference threshold and spatial distance threshold as the search radius;
[0020] S113, based on each trajectory point in the trajectory point set, calculate the time difference and straight-line spatial distance between the next trajectory point and the current trajectory point, and determine whether the time difference and straight-line spatial distance meet the set time difference threshold and spatial distance threshold conditions; if both conditions are met, the next trajectory point is included in the set, and the search is continued with the next trajectory point as the starting point; if any condition is not met, the current trajectory point is used as the starting point of the new set, and step S112 is repeated until both conditions are met;
[0021] S114, repeating steps S112 to S113 until all trajectory points are assigned to corresponding sets;
[0022] S115. Based on the trajectory points in each set, the longitude, latitude, arrival time and departure time of the new trajectory point are calculated using calculation formulas respectively, the new trajectory point is determined, and the new trajectory point replaces the trajectory point set to form a cleaned trajectory point.
[0023] Furthermore, the new trajectory point longitude calculation formula is:
[0024]
[0025] Where, Lng a Indicates the longitude of the ath trajectory point in the set;
[0026] Lng represents the longitude of the trajectory point in the replacement set;
[0027] n represents the number of trajectory points in the replacement set;
[0028] The new trajectory point latitude calculation formula is:
[0029]
[0030] In the formula, Lat a Represents the latitude of the ath trajectory point in the set;
[0031] Lat represents the latitude of the trajectory point in the replacement set;
[0032] n represents the number of trajectory points in the replacement set;
[0033] The calculation formula for the arrival time of the new trajectory point is:
[0034]
[0035] t sa represents the arrival time of the ath trajectory point in the set;
[0036] t s represents the arrival time of the trajectory points in the replacement set;
[0037] a means replacing the ath trajectory point in the set;
[0038] n represents the number of trajectory points in the replacement set;
[0039] The calculation formula for the departure time of the new trajectory point is:
[0040]
[0041] t earepresents the departure time of the ath trajectory point in the set;
[0042] t e represents the departure time of the trajectory point in the replacement set;
[0043] a means replacing the ath trajectory point in the set;
[0044] n represents the number of trajectory points in the replacement set.
[0045] Furthermore, the probability calculation formula for mapping the cleaned trajectory points to each road section in the road network structure is:
[0046]
[0047] In the formula, represents the matching probability between trajectory point p and road segment q;
[0048] q represents the road segment in the road network structure;
[0049] p represents the trajectory point after cleaning;
[0050] Q represents the set of all candidate road segments;
[0051] σ d represents the standard deviation of the distance;
[0052] σ θ Indicates the standard deviation of the direction;
[0053] d(p,q) represents the vertical distance between trajectory point p and road segment q;
[0054] θ(p,q) represents the direction difference between trajectory point p and road segment q;
[0055] q' generally refers to a road section in the road network structure;
[0056] d(p,q') represents the vertical distance between trajectory point p and road segment q';
[0057] θ(p,q') represents the direction difference between trajectory point p and road segment q'.
[0058] Furthermore, the shortest path algorithm is used to process the time-sorted road sections to obtain the individual actual paths, including the following steps:
[0059] S131, based on the time-sorted road sections, using the shortest path algorithm to obtain k driving paths between trajectory points;
[0060] S132, collecting the trajectory points of several users who travel through k driving paths in the same period, and calculating the average speed between the trajectory points of several users;
[0061] S133, obtaining the actual distance of the driving path, combining the average speed, calculating the average time of passing the driving path, and estimating the passing time of each path based on the time difference between the trajectory points;
[0062] S134. Compare the difference between the average duration and the passing time of each path to obtain the individual actual path.
[0063] Furthermore, the difference between the average duration and the path duration is compared to obtain the individual actual path, including:
[0064] If the difference between the average duration and the passing time of each path is different, the path with the smaller difference is selected as the individual actual path;
[0065] If the difference between the average duration and the passing time of each path is the same, the path with the shorter distance is selected as the actual path;
[0066] If no user's trajectory point appears on the path within the same time period, the path with the shorter distance is selected as the actual path.
[0067] Furthermore, using individual historical travel characteristics to analyze group travel data, and building an inbound passenger flow prediction model based on the group travel data analysis results includes the following steps:
[0068] S21, integrating individual actual paths, individual time information and individual entry frequency to obtain group travel data;
[0069] S22, analyzing group travel data to obtain the continuous driving time distribution of the inbound passenger flow and the road section area affected by the inbound passenger flow;
[0070] S23. Based on the continuous driving time distribution of the inbound passenger flow and the road section area affected by the inbound passenger flow, calculate the inbound passenger flow of each road section and the inbound passenger flow of all affected road sections, and build an inbound passenger flow prediction model.
[0071] Further, analyzing the group travel data and obtaining the continuous travel time distribution of the inbound passenger flow includes the following steps:
[0072] Divide the whole day into preset time periods, combine group travel data, filter the inbound passenger flow, and calculate the continuous travel time of the inbound passenger flow;
[0073] The calculated continuous travel time of the in-zone passenger flow is segmented using a preset time unit to obtain a number of tuple structures, and all tuple structures are sorted from high to low according to the passenger flow to obtain a passenger flow list;
[0074] Calculate the minimum and maximum values in the passenger flow list to obtain the continuous driving time distribution of the in-area passenger flow;
[0075] Analyzing group travel data and obtaining the road sections affected by the inbound passenger flow includes the following steps:
[0076] Filter out the passenger flow data entering the zone within the q period from the collected historical data;
[0077] Determine the road section location of the passenger flow data within a preset time before entering the area, and use the road section location as the road section area affected by the passenger flow entering the area.
[0078] Furthermore, the inbound passenger flow prediction model is a two-layer linear regression model, in which the first layer obtains the high-frequency inbound passenger flow and the low-frequency inbound passenger flow for each road section;
[0079] The second layer obtains the coefficient of each road section for all road sections within the affected road section area.
[0080] The beneficial effects of the present invention are:
[0081] 1. By utilizing individual historical operator data, the present invention can deeply analyze the individual's historical travel characteristics, laying a solid foundation for building an inbound passenger flow prediction model, thereby more keenly capturing passenger flow dynamics, enhancing the accuracy of short-term passenger flow prediction, and providing strong support for service area management.
[0082] 2. Through the prediction of real-time passenger flow, the present invention enables service area managers to respond quickly to changes in passenger flow and adjust resource allocation in real time to ensure efficient use of resources in the service area, which not only improves the smoothness and convenience of the service, but also significantly enhances the overall satisfaction and experience value of passengers.
[0083] 3. Through the acquisition and analysis of individual real-time travel data, the present invention enables the service area to grasp the passenger flow of each road section in real time, thereby adjusting the traffic management strategy in a timely manner and improving the overall operational efficiency of the service area. BRIEF DESCRIPTION OF THE DRAWINGS
[0084] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0085] Figure 1 is a flow chart of a method for short-term prediction of passenger flow into a service area based on operator data according to an embodiment of the present invention;
[0086] Figure 2 2 is a schematic diagram of a road network structure in a method for short-term prediction of passenger flow entering a service area based on operator data according to an embodiment of the present invention;
[0087] Figure 3 is a schematic diagram of a path in a method for short-term prediction of passenger flow entering a service area based on operator data according to an embodiment of the present invention;
[0088] Figure 4 It is a schematic diagram of trajectory points in a method for short-term prediction of passenger flow entering a service area based on operator data according to an embodiment of the present invention. DETAILED DESCRIPTION
[0089] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention and are mainly used to illustrate the embodiments. They can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these contents, ordinary technicians in this field should be able to understand other possible implementation methods and advantages of the present invention.
[0090] According to an embodiment of the present invention, a method for short-term prediction of inbound passenger flow in a service area based on operator data is provided.
[0091] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to the method for short-term prediction of passenger flow in a service area based on operator data according to an embodiment of the present invention, the method comprises the following steps:
[0092] S1. Analyze individual historical travel characteristics based on the acquired individual historical operator data, wherein the individual historical travel characteristics include individual actual path, individual time information and individual entry frequency.
[0093] It should be noted that the operator data is the operator's Minimization of Drive-tests (MDT) data, which is a sparse trajectory data with longitude and latitude. Its core fields include ID, longitude, latitude, arrival time, and departure time. This data can be collected by all mobile phone users. It is different from the trajectory in the mobile phone signaling data. The trajectory in the mobile phone signaling data is the location of the base station, while the longitude and latitude of the MDT data are the actual trajectory points of the mobile phone users.
[0094] Specifically, analyzing individual historical travel characteristics based on the acquired individual historical operator data, wherein the individual historical travel characteristics include individual actual path, individual time information and individual entry frequency, includes the following steps:
[0095] S11. Convert the acquired individual historical operator data into a trajectory point set, and clean the trajectory points in the trajectory point set.
[0096] Specifically, the obtained individual historical operator data is used as a track point set, and the track points in the track point set are cleaned, including the following steps:
[0097] S111, converting the acquired individual historical operator data into a trajectory point set, and arranging the trajectory points in the trajectory point set in ascending order according to arrival time;
[0098] S112, initializing the set, taking the first trajectory point after sorting as the starting point of the set, and searching for trajectory points according to the preset arrival time difference threshold and spatial distance threshold as the search radius;
[0099] S113, based on each trajectory point in the trajectory point set, calculate the time difference and straight-line spatial distance between the next trajectory point and the current trajectory point, and determine whether the time difference and straight-line spatial distance meet the set time difference threshold and spatial distance threshold conditions; if both conditions are met, the next trajectory point is included in the set, and the search is continued with the next trajectory point as the starting point; if any condition is not met, the current trajectory point is used as the starting point of the new set, and step S112 is repeated until both conditions are met;
[0100] S114, repeating steps S112 to S113 until all trajectory points are assigned to corresponding sets;
[0101] S115. Based on the trajectory points in each set, the longitude, latitude, arrival time and departure time of the new trajectory point are calculated using calculation formulas respectively, the new trajectory point is determined, and the new trajectory point replaces the trajectory point set to form a cleaned trajectory point.
[0102] Specifically, the new trajectory point longitude calculation formula is:
[0103]
[0104] Where, Lng a Indicates the longitude of the ath trajectory point in the set;
[0105] Lng represents the longitude of the trajectory point in the replacement set;
[0106] n represents the number of trajectory points in the replacement set;
[0107] The new trajectory point latitude calculation formula is:
[0108]
[0109] In the formula, Lat a Represents the latitude of the ath trajectory point in the set;
[0110] Lat represents the latitude of the trajectory point in the replacement set;
[0111] n represents the number of trajectory points in the replacement set;
[0112] The calculation formula for the arrival time of the new trajectory point is:
[0113]
[0114] t sa represents the arrival time of the ath trajectory point in the set;
[0115] t s represents the arrival time of the trajectory points in the replacement set;
[0116] a means replacing the ath trajectory point in the set;
[0117] n represents the number of trajectory points in the replacement set;
[0118] The calculation formula for the departure time of the new trajectory point is:
[0119]
[0120] t ea represents the departure time of the ath trajectory point in the set;
[0121] t e represents the departure time of the trajectory point in the replacement set;
[0122] a means replacing the ath trajectory point in the set;
[0123] n represents the number of trajectory points in the replacement set.
[0124] S12. Based on the constructed road network structure, the probability of mapping the cleaned trajectory points to each road section in the road network structure is calculated, and the road section with the highest probability is selected. Combined with the time sequence of the trajectory points, the selected road sections are time-sorted.
[0125] Specifically, the probability calculation formula for mapping the cleaned trajectory points to each road section in the road network structure is:
[0126]
[0127] In the formula, represents the matching probability between trajectory point p and road segment q;
[0128] q represents the road segment in the road network structure;
[0129] p represents the trajectory point after cleaning;
[0130] Q represents the set of all candidate road segments;
[0131] σ d represents the standard deviation of the distance;
[0132] σ θ Indicates the standard deviation of the direction;
[0133] d(p,q) represents the vertical distance between trajectory point p and road segment q;
[0134] θ(p,q) represents the direction difference between trajectory point p and road segment q;
[0135] q' generally refers to a road section in the road network structure;
[0136] d(p,q') represents the vertical distance between trajectory point p and road segment q';
[0137] θ(p,q') represents the direction difference between trajectory point p and road segment q'.
[0138] It is necessary to add that, if Figure 2 As shown in the figure, the road network structure is constructed. The road network consists of nodes and road sections. In addition, the road network elements also include toll booths, service areas, etc. A, B, etc. are nodes, and the section between AB is a road section. The formula of d(p,q) is:
[0139]
[0140] Where d(p,q) represents the vertical distance between trajectory point p and road segment q;
[0141] p x represents the longitude of trajectory point p;
[0142] p y represents the latitude of trajectory point p;
[0143] Indicates the longitude of the starting point of segment q;
[0144] Indicates the latitude of the starting point of segment q;
[0145] Indicates the longitude of the end point of segment q;
[0146] Indicates the latitude of the end point of segment q;
[0147] Indicates the difference between the latitude of the end point and the latitude of the starting point of the road segment q;
[0148] Indicates the difference between the longitude of the end point and the longitude of the starting point of segment q.
[0149] The formula for θ(p,q) is:
[0150]
[0151] in, and They are:
[0152]
[0153]
[0154] Where θ(p,q) represents the direction difference between trajectory point p and road segment q;
[0155] Represents the direction vector of trajectory point p;
[0156] represents the direction vector of road segment q;
[0157] p x represents the longitude of trajectory point p;
[0158] p y represents the latitude of trajectory point p;
[0159] Indicates the longitude of the next point of point p;
[0160] Indicates the latitude of the next point of point p;
[0161] Indicates the longitude of the starting point of segment q;
[0162] Indicates the latitude of the starting point of segment q;
[0163] Indicates the longitude of the end point of segment q;
[0164] Indicates the latitude of the end point of segment q.
[0165] S13. Use the shortest path algorithm to process the time-sorted road sections to obtain individual actual paths.
[0166] Specifically, using the shortest path algorithm to process the time-sorted road segments to obtain the individual actual paths includes the following steps:
[0167] S131. Based on the time-sorted road sections, use the shortest path algorithm to obtain k driving paths between trajectory points.
[0168] S132: Collect the trajectory points of several users who pass through k driving paths in the same period of time, and calculate the average speed between the trajectory points of the several users.
[0169] S133: Obtain the actual distance of the driving path, calculate the average time of passing the driving path in combination with the average speed, and estimate the passing time of each path in combination with the time difference between the trajectory points.
[0170] S134. Compare the difference between the average duration and the passing time of each path to obtain the individual actual path.
[0171] Specifically, comparing the difference between the average duration and the path duration, obtaining the individual actual path includes:
[0172] If the difference between the average duration and the passing time of each path is different, the path with the smaller difference is selected as the individual actual path;
[0173] If the difference between the average duration and the passing time of each path is the same, the path with the shorter distance is selected as the actual path;
[0174] If no user's trajectory point appears on the path within the same time period, the path with the shorter distance is selected as the actual path.
[0175] It is necessary to add that, if Figure 3 As shown in the figure, there are two possible paths MBCEN and MBDEN between point M and point N. The trajectories of other users passing through BCE and BDE in the same period are extracted. For example, there are N individual users who pass through BCE in the same period. Each of them leaves two or more trajectory points on the BCE path. The path distance and time difference between these trajectory points are calculated to obtain the average speed. The formula for the average speed is:
[0176]
[0177] In the formula, represents the speed of user i on path BCE;
[0178] It represents the average speed of individuals traveling on the BCE path during the same period;
[0179] represents the path distance between the first and last trajectory points observed by user i in path BCE;
[0180] represents the absolute value of the difference between the arrival time of the first trajectory point and the departure time of the last point observed by user i in path BCE;
[0181] i represents user;
[0182] N represents the number of users.
[0183] Based on the average speed of BCE and BDE and , calculate the passing time of each path:
[0184]
[0185]
[0186] Where, t BCE It represents the average time from point B to point E via point C during the same period;
[0187] t BDE It represents the average time from point B to point E via point D during the same period;
[0188] L BCE It represents the length of the path from B to E via C, which is the length of BC plus the length of CE;
[0189] L BDE It represents the length of the path from B to E via D, which is the length of BD plus the length of DE;
[0190] v BCE It represents the average speed of individuals traveling on the BCE path during the same period;
[0191] v BDE It represents the average speed of individuals traveling on the BDE path during the same period.
[0192] The time difference between points M and N is , is the time when the individual reaches point N, is the moment when the individual leaves point M, then if the individual passes through BCE, the estimated time in BCE is: , L MB , L EN are the lengths of MB and EN respectively. If an individual passes through BDE, the time spent in BDE is estimated to be ;calculate and If the former is smaller, it is considered that the individual passes through BCE; if the latter is smaller, it is considered that the individual passes through BDE; if the two are the same, it is considered that the individual passes through the shorter path between BCE and BDE; if no user's trajectory appears on these two paths in the same period of time, the shortest path of the two paths is taken as the actual path.
[0193] S14. Use the linear difference method to calculate the individual time information of the trajectory point arriving at each node along the individual actual path.
[0194] It is necessary to add that, if Figure 4 As shown in the figure, the time when an individual arrives at each node is determined. Since most of the trajectory points are in the middle of the road section, the linear difference method is used to obtain the time of the node. M, N, and Q are three points on the travel path. The three points are on sections AB, BC, and EF respectively. The calculation method for the time when it passes through B, C, D, and E is:
[0195]
[0196] Where, t B Indicates the time when passing through point B;
[0197] △t MN Indicates the time difference between point M and point N;
[0198] L MB represents the path length between point M and point B;
[0199] L BN represents the path length between point B and point N;
[0200] represents the moment when the individual leaves point M;
[0201] t C represents the time when it passes through point C;
[0202] represents the moment when the individual reaches point Q;
[0203] represents the moment when the individual leaves point N;
[0204] L NC represents the path length between point N and point C;
[0205] L CD represents the path length between point C and point D;
[0206] L DE represents the path length between point D and point E;
[0207] L EQ Represents the path length between point E and point Q.
[0208] In addition, moments D and E are similar to moments B and C.
[0209] S15, matching the position of the cleaned trajectory point with the known service area position, identifying the trajectory point of the individual entering the service area, and obtaining the individual entry frequency.
[0210] It should be noted that according to the results of trajectory cleaning, the trajectory point position is matched with the service area position. The service area is a surface area. When the trajectory point position is within the service area and the departure time minus the arrival time of the trajectory point exceeds a threshold (such as 5 minutes), it indicates that the passenger has entered the service area.
[0211] Integrate the data of multiple individual trips. For example, if there are Y trips, and Z of them have entered the service area, then the individual's entry frequency is Z / Y. When this frequency is higher than 0.4, it is considered to be a high-frequency crowd in this service area, otherwise it is a low-frequency crowd.
[0212] S2. Analyze group travel data using individual historical travel characteristics, and build an inbound passenger flow prediction model based on the group travel data analysis results.
[0213] Specifically, analyzing group travel data using individual historical travel characteristics and building an inbound passenger flow prediction model based on the group travel data analysis results includes the following steps:
[0214] S21. Integrate individual actual paths, individual time information, and individual entry frequency to obtain group travel data.
[0215] S22. Analyze group travel data to obtain the continuous driving time distribution of the inbound passenger flow and the road section area affected by the inbound passenger flow.
[0216] It should be noted that the continuous driving duration is as follows: a trip includes a starting point, an end point, and nodes such as service areas that may be entered in the middle, such as "starting point-service area 1-service area 2-end point". At any time, the calculation method of the individual's continuous travel duration is the current time minus the time of the nearest node reached before that time. For example, if the current time is between service area 1 and service area 2, the continuous driving duration is the time difference between the current time and the time when the individual leaves service area 1.
[0217] Specifically, analyzing group travel data and obtaining the continuous travel time distribution of inbound passenger flow includes the following steps:
[0218] Divide the whole day into preset time periods, combine group travel data, filter the inbound passenger flow, and calculate the continuous travel time of the inbound passenger flow;
[0219] The calculated continuous travel time of the in-zone passenger flow is segmented using a preset time unit to obtain a number of tuple structures, and all tuple structures are sorted from high to low according to the passenger flow to obtain a passenger flow list;
[0220] Calculate the minimum and maximum values in the passenger flow list to obtain the continuous driving time distribution of the in-area passenger flow;
[0221] Analyzing group travel data and obtaining the road sections affected by the inbound passenger flow includes the following steps:
[0222] Filter out the passenger flow data entering the zone within the q period from the collected historical data;
[0223] Determine the road section location of the passenger flow data within a preset time before entering the area, and use the road section location as the road section area affected by the passenger flow entering the area.
[0224] It should be noted that the 24 hours of the day are divided into 48 periods of 30 minutes. Starting from the first period, the passenger flow entering the service area is screened in turn, and their continuous driving time is calculated:
[0225]
[0226] In the formula, deltaT represents the continuous driving time;
[0227] Indicates the time when the individual arrives at the service area node (Node);
[0228] Indicates the time when the individual leaves the previous node (starting point or service area where it has stayed before).
[0229] The continuous driving time (deltaT) is calculated and distributed in units of 10 minutes, and several tuple structures are obtained as (i, nums i ), where i is deltaT / / 10 (deltaT is divided by 10 and rounded down), nums is the number of individuals entering the service area, and the number of individuals with continuous driving time segment i is sorted from high to low according to nums, and a matrix consisting of (i, nums i ) passenger flow list composed of; , Nums is the total passenger flow entering the service area during this period, N is the number of continuous driving time segments, and the cumulative sum of nums and nums_cum in the passenger flow list is calculated in sequence. ,when Calculate the smallest value j among the first j tuples in the passenger flow list min and the maximum value j max , the range of the driving time is (j min , j max ), where t min =j min *10,t max =j max *10.
[0230] In addition, as the moment t0, to predict the passenger flow entering the area within the next 30 minutes, it is necessary to obtain the position of the passenger flow that can arrive at the service area within 30 minutes on the road network at the moment t0. In order to obtain historical data, the moment t0 is the starting moment of the q period, and the passenger flow entering the area during the q period in the historical data is filtered out, and the position 30 minutes before entering the area is used as the road section where these positions are located as the influencing section range of the passenger flow entering the service area during the q period.
[0231] S23. Based on the continuous driving time distribution of the inbound passenger flow and the road section area affected by the inbound passenger flow, calculate the inbound passenger flow of each road section and the inbound passenger flow of all the influencing road sections, and build an inbound passenger flow prediction model.
[0232] Specifically, the inbound passenger flow prediction model is a two-layer linear regression model. The first layer obtains the high-frequency inbound passenger flow and the low-frequency inbound passenger flow for each road section;
[0233] The second layer obtains the coefficient of each road section for all road sections within the affected road section area.
[0234] It should be noted that the inbound passenger flow prediction model adopts a double-layer linear regression model, which is not only highly interpretable and computationally efficient, but also adaptable to short-term prediction needs.
[0235] The first layer of the inbound passenger flow prediction model consists of two univariate linear models for each road section:
[0236] Calculate the passenger flow of these sections at time t0, and select the sections with continuous driving time in the range of (t min -30,t max ) passenger flow, and get the total passenger flow of section i all i =high i +low i , among which the high frequency passenger flow is high i , the low frequency passenger flow is low i , the actual passenger flow entering the area among the high-frequency passenger flow is recorded as high i in =f i high (high i ), the actual passenger flow in the low-frequency passenger flow is low i in =f i low (low i ), where f is a linear regression function, , since there is only one independent variable, j=1, f is a univariate linear regression function, using data from the same period of several days; for section i, the passenger flow that is about to enter the area during a certain period is:
[0237]
[0238] In the formula, r i in It indicates the passenger flow that will enter the area on road section i during a certain period of time;
[0239] high iin Indicates the actual passenger flow entering the area among the high-frequency passenger flow entering the area;
[0240] low i in Indicates the actual passenger flow entering the area within the low-frequency passenger flow entering the area;
[0241] f represents a univariate linear regression function;
[0242] high i Indicates high frequency passenger flow into the area;
[0243] low i Indicates low-frequency passenger flow into the area;
[0244] a i high The coefficient representing the actual entry of high-frequency passenger flow into the zone;
[0245] b i high is the intercept term of the high-frequency passenger flow actually entering the area;
[0246] a i low is the coefficient of the actual entry of low-frequency passenger flow into the zone;
[0247] b i low It is the intercept term of the low-frequency passenger flow actually entering the area.
[0248] The second layer of the inbound passenger flow prediction model is composed of a multivariate linear regression model for all road sections within the affected road section area:
[0249] 1. For all the involved sections, the passenger flow K in , use the multiple linear regression method to fit the coefficients of each road section: , where K in is the passenger flow entering the area of all sections, f is the linear regression function, r i in is the passenger flow that is about to enter the area on road section i during a certain period of time.
[0250] 2. Based on the multivariate linear regression method, in order to improve the stability and generalization ability of the model, a mixed penalty term is introduced into the loss function, that is, Lasso (L1) and Ridge (L2) regularization are combined at the same time. The process is as follows:
[0251] For a multiple linear regression model, it is expressed as:
[0252]
[0253] In the formula, K inIndicates the passenger flow into the zone for all sections;
[0254] k represents the number of road sections;
[0255] i represents the road section;
[0256] β i represents the coefficient;
[0257] β0 represents a constant;
[0258] ε represents the error term;
[0259] r i in It indicates the passenger flow that will enter the area on road section i during a certain period of time.
[0260] The loss function is:
[0261]
[0262] In the formula, represents the loss function;
[0263] k represents the number of road sections;
[0264] i represents the road section;
[0265] K i in represents the passenger flow entering the zone at section i;
[0266] β i represents the coefficient;
[0267] r i in It indicates the passenger flow that will enter the area on road section i during a certain period of time;
[0268] β0 represents a constant;
[0269] λ1 represents the weight hyperparameter of the L1 regularization term;
[0270] λ2 represents the weight hyperparameter of the L2 regularization term;
[0271] represents the vector of coefficients;
[0272] represents the L1 regularization term;
[0273] represents the L2 regularization term.
[0274] To optimize the above loss function, the steps are as follows:
[0275] (1) Random initialization and β0;
[0276] (2) Update , where β0 represents a constant, k represents the number of road sections; i represents the road section, K i in represents the passenger flow entering the zone at section i, β i Represents the coefficient, r i in It indicates the passenger flow that will enter the area on road section i during a certain period of time;
[0277] (3) Update
[0278]
[0279] In the formula, β i represents the coefficient, sign represents the function, k represents the number of sections, i represents the section, j represents the section, r j in represents the passenger flow that will enter the area on section j during a certain period of time, K i in represents the passenger flow entering the zone at section i, r i in represents the passenger flow that is about to enter the area on road section i during a certain period of time, β0 represents a constant, λ1 represents the weight hyperparameter of the L1 regularization term, and λ2 represents the weight hyperparameter of the L2 regularization term;
[0280] Among them, the sign function is defined as follows:
[0281]
[0282] In the formula, sign represents a function, and x represents an arbitrary value;
[0283] Repeat (2) and (3) until , from which we can get and β0, where F is the loss function, t is the time, represents the vector of coefficients, β0 represents a constant,
[0284] Based on the above steps, the double-layer linear regression model of the passenger flow entering a service area and the passenger flow of each section in a certain period of time is obtained as follows:
[0285]
[0286] That is, as long as the current passenger flow of each road section is input, the passenger flow entering the corresponding service area in the next half hour can be estimated.
[0287] In the formula, K in represents the passenger flow entering the area of all sections, k represents the number of sections, i represents the section, β i represents the coefficient, a i highIndicates the coefficient of the actual entry of high-frequency passenger flow into the area, high i represents the high-frequency passenger flow into the area, b i high is the intercept term of the high-frequency passenger flow entering the area, a i low is the coefficient of the actual entry of low-frequency passenger flow into the area, low i represents the low-frequency passenger flow into the area, b i low is the intercept term of the actual entry of low-frequency passenger flow into the area, and β0 represents a constant.
[0288] S3. Use individual real-time travel data to obtain the real-time passenger flow corresponding to each road section, and input the real-time passenger flow into the inbound passenger flow prediction model to obtain the short-term passenger flow in the service area.
[0289] It should be noted that, according to the actual operation, the passenger flow into the service area in the next 30 minutes at time t0 is predicted. It is only necessary to detect the time t0, and the continuous driving time of the corresponding road section is within (t min -30,t max ) can be predicted based on the above-mentioned inbound passenger flow prediction model.
[0290] In summary, with the help of the above technical solution of the present invention, by utilizing individual historical operator data, it is possible to deeply analyze the individual's historical travel characteristics, laying a solid foundation for building an inbound passenger flow prediction model, thereby more keenly capturing passenger flow dynamics, enhancing the accuracy of short-term passenger flow prediction, and providing strong support for service area management. Through the prediction of real-time passenger flow, service area managers can respond quickly to changes in passenger flow, adjust resource allocation in real time, and ensure efficient use of resources in the service area, which not only improves the smoothness and convenience of the service, but also significantly enhances the overall satisfaction and experience value of passengers. Through the acquisition and analysis of individual real-time travel data, the service area can grasp the passenger flow situation of each section in real time, so as to adjust the traffic management strategy in time and improve the overall operational efficiency of the service area.
[0291] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A short-term prediction method for passenger flow into a service area based on operator data, characterized in that: The method comprises the following steps: S1. Analyze individual historical travel characteristics based on the acquired individual historical operator data, wherein the individual historical travel characteristics include individual actual path, individual time information and individual entry frequency; S2. Analyze group travel data using individual historical travel characteristics, and build an inbound passenger flow prediction model based on the group travel data analysis results; S3. Using individual real-time travel data, obtain the real-time passenger flow corresponding to each road section, and input the real-time passenger flow into the inbound passenger flow prediction model to obtain the short-term passenger flow in the service area; The method of analyzing group travel data by using individual historical travel characteristics and constructing an inbound passenger flow prediction model based on the group travel data analysis results includes the following steps: S21, integrating individual actual paths, individual time information and individual entry frequency to obtain group travel data; S22, analyzing group travel data to obtain the continuous driving time distribution of the inbound passenger flow and the road section area affected by the inbound passenger flow; S23, based on the continuous driving time distribution of the inbound passenger flow and the road section area affected by the inbound passenger flow, calculate the inbound passenger flow of each road section and the inbound passenger flow of all the influencing road sections, and build an inbound passenger flow prediction model; Analyzing group travel data to obtain the continuous travel time distribution of inbound passenger flow includes the following steps: Divide the whole day into preset time periods, combine group travel data, filter the inbound passenger flow, and calculate the continuous travel time of the inbound passenger flow; The calculated continuous travel time of the in-zone passenger flow is segmented using a preset time unit to obtain a number of tuple structures, and all tuple structures are sorted from high to low according to the passenger flow to obtain a passenger flow list; Calculate the minimum and maximum values in the passenger flow list to obtain the continuous driving time distribution of the in-area passenger flow; Analyzing group travel data to obtain the road section area affected by the inbound passenger flow includes the following steps: Filter out the passenger flow data entering the zone within the q period from the collected historical data; Determine the road section location of the passenger flow data within a preset time before entering the zone, and use the road section location as the road section area that affects the passenger flow entering the zone; The inbound passenger flow prediction model is a double-layer linear regression model, in which the first layer obtains high-frequency inbound passenger flow and low-frequency inbound passenger flow for each road section; The second layer obtains the coefficient of each road section for all road sections within the affected road section area.
2. The method for short-term prediction of passenger flow in a service area based on operator data according to claim 1 is characterized in that: The step of analyzing individual historical travel characteristics based on the acquired individual historical operator data, wherein the individual historical travel characteristics include individual actual path, individual time information and individual entry frequency, comprises the following steps: S11, converting the acquired individual historical operator data into a trajectory point set, and cleaning the trajectory points in the trajectory point set; S12, based on the constructed road network structure, calculate the probability of mapping the cleaned trajectory points to each road section in the road network structure, select the road section with the highest probability, and sort the selected road sections in time according to the time sequence of the trajectory points; S13, using the shortest path algorithm to process the time-sorted road sections to obtain individual actual paths; S14, using a linear difference method to calculate the individual time information of the trajectory point arriving at each node along the individual actual path; S15, matching the position of the cleaned trajectory point with the known service area position, identifying the trajectory point of the individual entering the service area, and obtaining the individual entry frequency.
3. The method for short-term prediction of passenger flow in a service area based on operator data according to claim 2 is characterized in that: The method of using the acquired individual historical operator data as a track point set and cleaning the track points in the track point set includes the following steps: S111, converting the acquired individual historical operator data into a trajectory point set, and arranging the trajectory points in the trajectory point set in ascending order according to arrival time; S112, initializing the set, taking the first trajectory point after sorting as the starting point of the set, and searching for trajectory points according to the preset arrival time difference threshold and spatial distance threshold as the search radius; S113, based on each trajectory point in the trajectory point set, calculate the time difference and straight-line spatial distance between the next trajectory point and the current trajectory point, and determine whether the time difference and straight-line spatial distance meet the set time difference threshold and spatial distance threshold conditions; if both conditions are met, the next trajectory point is included in the set, and the search is continued with the next trajectory point as the starting point; if any condition is not met, the current trajectory point is used as the starting point of the new set, and step S112 is repeated until both conditions are met; S114, repeating steps S112 to S113 until all trajectory points are assigned to corresponding sets; S115. Based on the trajectory points in each set, the longitude, latitude, arrival time and departure time of the new trajectory point are calculated using calculation formulas respectively, the new trajectory point is determined, and the new trajectory point replaces the trajectory point set to form a cleaned trajectory point.
4. The method for short-term prediction of passenger flow in a service area based on operator data according to claim 3 is characterized in that: The new trajectory point longitude calculation formula is: Where, Lng a Indicates the longitude of the ath trajectory point in the set; Lng represents the longitude of the trajectory point in the replacement set; n represents the number of trajectory points in the replacement set; The new trajectory point latitude calculation formula is: In the formula, Lat a Represents the latitude of the ath trajectory point in the set; Lat represents the latitude of the trajectory point in the replacement set; n represents the number of trajectory points in the replacement set; The calculation formula for the new trajectory point arrival time is: t sa represents the arrival time of the ath trajectory point in the set; t s represents the arrival time of the trajectory points in the replacement set; a means replacing the ath trajectory point in the set; n represents the number of trajectory points in the replacement set; The calculation formula for the new trajectory point departure time is: t ea represents the departure time of the ath trajectory point in the set; t e represents the departure time of the trajectory point in the replacement set; a means replacing the ath trajectory point in the set; n represents the number of trajectory points in the replacement set.
5. The method for short-term prediction of passenger flow in a service area based on operator data according to claim 2 is characterized in that: The probability calculation formula of mapping the cleaned trajectory points to each road section in the road network structure is: In the formula, represents the matching probability between trajectory point p and road segment q; q represents the road segment in the road network structure; p represents the trajectory point after cleaning; Q represents the set of all candidate road segments; σ d represents the standard deviation of the distance; σ θ Indicates the standard deviation of the direction; d(p,q) represents the vertical distance between trajectory point p and road segment q; θ(p,q) represents the direction difference between trajectory point p and road segment q; q' generally refers to a road section in the road network structure; d(p,q') represents the vertical distance between trajectory point p and road segment q'; θ(p,q') represents the direction difference between trajectory point p and road segment q'.
6. The method for short-term prediction of passenger flow in a service area based on operator data according to claim 2, characterized in that: The process of using the shortest path algorithm to process the time-sorted road sections to obtain individual actual paths includes the following steps: S131, based on the time-sorted road sections, using the shortest path algorithm to obtain k driving paths between trajectory points; S132, collecting the trajectory points of several users who travel through k driving paths in the same period, and calculating the average speed between the trajectory points of several users; S133, obtaining the actual distance of the driving path, combining the average speed, calculating the average time of passing the driving path, and estimating the passing time of each path based on the time difference between the trajectory points; S134. Compare the difference between the average duration and the passing time of each path to obtain the individual actual path.
7. The method for short-term prediction of passenger flow in a service area based on operator data according to claim 6, characterized in that: The comparing the difference between the average duration and the path duration to obtain the individual actual path includes: If the difference between the average duration and the passing time of each path is different, the path with the smaller difference is selected as the individual actual path; If the difference between the average duration and the passing time of each path is the same, the path with the shorter distance is selected as the actual path; If no user's trajectory point appears on the path within the same time period, the path with the shorter distance is selected as the actual path.
