Internet map data-based passenger flow prediction method, device, equipment and medium

Abstract

The application relates to an Internet map data-based passenger flow prediction method, device, equipment and medium, and relates to the technical field of intelligent traffic data processing. The method comprises the following steps: obtaining dynamic monitoring data and historical reference data of a target region based on an Internet map, wherein the dynamic monitoring data comprises user trajectory positions and real-time moving directions, and the historical reference data comprises synchronous passenger flow distribution and environmental influence parameters; based on the dynamic monitoring data and the historical reference data, time periodicity features, space transfer features and environmental correlation features are extracted to form a space-time feature matrix; and the space-time feature matrix is input into a dynamic weight prediction model to output passenger flow prediction results containing specific regions and time periods. By integrating dynamic monitoring data and historical reference data related to an Internet map, and by inputting fused time, space and environmental features into a dynamic weight prediction model, high-precision passenger flow prediction is realized, and the prediction results can directly support subsequent regulation and control strategies.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation data processing technology, and in particular to a method, apparatus, device and medium for predicting passenger flow based on Internet map data. Background Technology

[0002] Traditional passenger flow forecasting methods often rely on single data sources (such as statistics from fixed turnstiles), which suffer from limited data coverage and an inability to reflect real-time user movement trends. Furthermore, they fail to fully integrate the spatial attributes of internet maps and real-time navigation data, resulting in fragmented data sources and a disconnect between forecast results and actual user travel behavior. Therefore, there is an urgent need for a solution that uses internet maps as the core platform, integrates real-time dynamic data and historical reference data, and achieves high-precision forecasting through multi-dimensional feature fusion. Summary of the Invention

[0003] This invention provides a passenger flow prediction method based on Internet map data. First, it relies on Internet maps to obtain multi-source input data. Second, it combines the spatial attributes of Internet maps to extract and fuse multi-dimensional features. Finally, it outputs prediction results through a dynamic model and simultaneously links Internet maps to implement strategies, thus solving the shortcomings of traditional methods such as single data and disconnection from user travel behavior.

[0004] In a first aspect, the present invention provides a method for predicting passenger flow based on Internet map data, the method comprising:

[0005] Dynamic monitoring data and historical baseline data of the target area are obtained based on Internet maps. The dynamic monitoring data includes user trajectory location and real-time movement direction, and the historical baseline data includes passenger flow distribution and environmental impact parameters during the same period.

[0006] Based on the dynamic monitoring data and historical benchmark data, time periodic features, spatial transfer features and environmental correlation features are extracted and fused to form a spatiotemporal feature matrix;

[0007] The spatiotemporal feature matrix is ​​input into the dynamic weight prediction model, which outputs passenger flow prediction results that include specific regions and time periods.

[0008] Secondly, the present invention provides a passenger flow prediction device based on Internet map data, comprising:

[0009] The acquisition unit is used to acquire dynamic monitoring data and historical benchmark data of the target area. The dynamic monitoring data includes user trajectory location and real-time movement direction, and the historical benchmark data includes passenger flow distribution and environmental impact parameters during the same period.

[0010] The extraction unit is used to extract time periodic features, spatial transfer features and environmental correlation features based on the dynamic monitoring data and historical benchmark data, and fuse them to form a spatiotemporal feature matrix;

[0011] The output unit is used to input the spatiotemporal feature matrix into the dynamic weight prediction model and output passenger flow prediction results including specific regions and time periods.

[0012] Thirdly, an electronic device is provided, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0013] Memory, used to store computer programs;

[0014] When a processor executes a program stored in memory, it implements the steps of the passenger flow prediction method based on Internet map data as described in any embodiment of the first aspect.

[0015] Fourthly, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the passenger flow prediction method based on Internet map data as described in any embodiment of the first aspect.

[0016] The technical solutions provided in the embodiments of the present invention have the following advantages compared with the prior art:

[0017] By integrating dynamic monitoring data and historical benchmark data related to internet maps, and extracting and merging time, space and environmental features, the data is input into a dynamic weight prediction model. This not only solves the problems of single data and one-sided prediction in traditional methods, but also achieves high-precision passenger flow prediction. Furthermore, the prediction results can directly support the implementation of subsequent control strategies. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart illustrating a passenger flow prediction method based on internet map data provided in an embodiment of the present invention;

[0021] Figure 2A structural block diagram of a passenger flow prediction device based on Internet map data provided in an embodiment of the present invention;

[0022] Figure 3 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Example

[0025] Figure 1 This is a flowchart illustrating a passenger flow prediction method based on internet map data, provided in an embodiment of the present invention. Specifically, this embodiment proposes a passenger flow prediction method based on internet map data; see [link to documentation]. Figure 1 The passenger flow prediction method based on Internet map data includes the following steps S101-S103.

[0026] Terminology Explanation

[0027] Target area: refers to the specific geographical area for which passenger flow forecasting is required. It is defined based on the Gaussian projection coordinate system of the Internet map. For example, a 1-kilometer polygonal area around a subway transfer station, or the boundary range of POI (Point of Interest) in the core area of ​​a commercial center. The spatial grid system of the Internet map (100m×100m map grid) provides a benchmark for the sub-region division of the target area.

[0028] Dynamic monitoring data refers to data collected in real time from internet maps that reflects the user's movement status. Its core purpose is to capture the current trend of passenger flow changes. Among them, "user trajectory location" is the user's real-time coordinates obtained by the internet map APP through GPS / BeiDou positioning (precisely matching the grid number of the internet map, such as "internet map grid G123: 116.402°E, 39.898°N"), "real-time movement direction" is determined by combining the real-time navigation path of the internet map (such as "the user plans the path 'Metro Exit B → Transfer Passage → Line 4 Platform' in the internet map, and the current movement direction is towards the transfer passage"), and also includes the real-time path congestion data of the internet map (such as "the internet map shows that the current congestion level of the 'Exit B to Transfer Passage' path is 'moderate'").

[0029] Historical baseline data: refers to historical data related to the target area over a past period, including historical data dimensions of the internet map; among them, "Simultaneous passenger flow distribution" combines the user trajectory heat map of the same period in the history of the internet map (e.g., "In the past 30 working days, the average trajectory heat value of the transfer station area in the internet map during the morning peak of 7:30-8:00 is 85 (out of 100)"); "Environmental impact parameters" adds historical navigation request data of the internet map (e.g., "In the past rainy morning peak, the number of internet map navigation requests related to transfer stations was 18% higher than that on sunny days") and POI activity data (e.g., "During the same period in the history of shopping mall promotion activities, the number of navigation requests around the shopping mall in the internet map increased by 35%").

[0030] Time-periodic characteristics: The regular characteristics of passenger flow changes over time are extracted from historical benchmark data, and new time-dimensional related data from the Internet map are added (such as "In the morning rush hour of the past 30 working days, the search volume of the keyword 'transfer station' in the Internet map peaked between 7:20 and 7:25, and the passenger flow density increased by 15% in the following 10 minutes"), which further improves the predictability of time characteristics.

[0031] Spatial transfer characteristics: Based on the spatial attributes extracted from dynamic monitoring data, historical benchmark data and internet maps, the characteristics are regular. Among them, "sub-region transfer patterns" are combined with the path topology of the internet map (e.g., "the shortest path between Exit B and the transfer passage in the internet map contains 2 turning points, and the user selection rate of this path reaches 92%), and "density distribution association" is combined with the POI location of the internet map (e.g., "the area around the convenience store POI next to the transfer passage in the internet map has a 20% higher passenger flow density than other areas").

[0032] Environmental correlation characteristics: Combining environmental impact parameters from historical baseline data and the regularity characteristics extracted from scene data of internet maps, a new "internet map scene correlation" is added (such as "when the internet map shows that the surrounding roads are congested, the passenger flow at the subway transfer station increases by 22% compared to when the roads are clear" and "during the 'large concert' POI event in the internet map, the passenger flow around the venue increases by 50%").

[0033] Spatiotemporal Feature Matrix: A structured data set formed by aligning three types of features according to the "Internet map grid - time segment" dimension; the spatial dimension is indexed by the grid number of the Internet map (e.g., rows represent Internet map grids G101-G150), the temporal dimension is based on the navigation time granularity of the Internet map (e.g., 5 minutes / segment), and the cell data includes fused features such as "Internet map grid heat value + path selection rate + navigation request volume".

[0034] Dynamic weight prediction model: Add weight adaptation logic for Internet map data. When the real-time trajectory update frequency of Internet map is >1500 records / minute (i.e., real-time data is sufficient), the weight ratio of "Internet map path transfer features" will be increased (e.g., from 60% to 65%). When the integrity of historical navigation request data of Internet map is >90% (i.e., historical data is reliable), the weight ratio of "Internet map scene association features" will be increased.

[0035] Specific area: refers to the subdivided unit within the target area based on POI or grid division on the Internet map (such as "grid G128 of passage A of the transfer station in the Internet map" or "grid G205-G208 corresponding to the atrium on the first floor of the shopping mall in the Internet map"), to ensure that the prediction results correspond one-to-one with the spatial identifiers on the Internet map.

[0036] Passenger flow forecast results: In addition to quantitative data, a new format adapted to Internet maps has been added to output structured data of "Internet map grid - time period - predicted passenger flow density", which can be directly imported into the heat map rendering module of Internet maps (e.g., "Internet map G128 grid 8:00-8:05 predicted passenger flow density 550 people / 100 square meters, corresponding to heat map level 'red (congested)'").

[0037] Internet map navigation linkage: A new feature refers to converting passenger flow forecast results into navigation strategies on the internet map (such as congestion warnings and route recommendations), and pushing them to users who plan relevant routes within the target area, forming a closed loop of "prediction-control-feedback".

[0038] S101, Based on the Internet map, obtain dynamic monitoring data and historical benchmark data of the target area. The dynamic monitoring data includes user trajectory location and real-time movement direction, and the historical benchmark data includes passenger flow distribution and environmental impact parameters during the same period.

[0039] In one embodiment, the step of acquiring dynamic monitoring data of the target area includes:

[0040] The location of the user's trajectory is collected through the positioning module of the internet map software, which includes satellite positioning and base station positioning; the real-time movement direction of the user is extracted by combining the real-time navigation function of the internet map, and real-time path congestion data of the internet map is collected simultaneously. The congestion data includes the congestion level and the average traffic speed of the corresponding path, and the congestion level is divided according to preset rules.

[0041] In a specific embodiment, real-time route congestion data refers to route traffic status data calculated in real time by internet maps using user trajectories, traffic events, and other data. This data serves as an auxiliary basis for judging the current density of passenger flow. Congestion level refers to a classification description of route traffic status (e.g., mild, moderate, severe), used to quantify the degree of congestion. Preset rules refer to pre-defined standards for classifying congestion levels (e.g., "mild congestion: average travel speed 10-20 km / h; moderate congestion: 5-10 km / h; severe congestion: <5 km / h").

[0042] In this embodiment, taking "Dynamic monitoring data collection of the morning peak (7:30-8:00) at the XX transfer station of Metro Line 3 in a certain city" as an example: The location module of the internet map APP on the user's mobile phone (which supports both BeiDou satellite positioning and mobile base station positioning) is used to collect the user's trajectory location within 1 kilometer of the transfer station. The preset positioning accuracy standard is set to "error ≤ 10 meters" to ensure that the user's location can accurately match a preset 100m × 100m grid in the internet map (e.g., user A's location matches grid G15, user B's matches grid G16); combined with the real-time navigation function of the internet map, the user's real-time movement direction is extracted (e.g., "user A's current direction of movement"). The navigation path for user B is 'Exit A of XX Station → Transfer Passage → Line 4 Platform', with the direction of movement being the transfer passage; the navigation path for user B is 'Exit B of XX Station → Line 3 Platform', with the direction of movement being the Line 3 Platform'. Real-time path congestion data from the internet map is collected simultaneously. The average travel speed for the path from 'Exit A of XX Station to the transfer passage' is 15 km / h, which is classified as mild congestion according to the preset rule ("10-20 km / h is mild congestion"). The average travel speed for the path from 'Exit B of XX Station to Line 3 Platform' is 8 km / h, which is classified as moderate congestion. These data together constitute the dynamic monitoring data for this period, providing a real-time basis for subsequent feature extraction.

[0043] In one embodiment, the step of obtaining the historical benchmark data includes:

[0044] Select periodic data with the same time attributes as the prediction period. The time attributes include weekday / weekend attributes and peak / off-peak time attributes. The selected periodic data has a preset period length. The internet map-related data in the periodic data includes historical user trajectory heatmaps, navigation request volumes, and POI activity records for the same period. The POI activity records include the time and location information of commercial and public activities. The environmental impact parameters are supplemented through the historical weather interface and traffic event interface of the internet map. The specific correlation logic is that when a specific environmental identifier and the corresponding area navigation request volume change occur simultaneously in the historical data of the internet map, the correlation is included in the basis for extracting environmental correlation features.

[0045] In a specific embodiment, the concurrent periodic data refers to historical data with the same time attributes as the prediction period, serving as the core sample for extracting historical patterns (e.g., to predict "Friday morning rush hour," all past Friday morning rush hour data are selected). Preset period length: The pre-set collection duration of the concurrent periodic data (e.g., 30 days, 90 days) must cover different environmental conditions (rainy / sunny days, holidays / ordinary days) to ensure data representativeness. User trajectory heatmap: Visual data in internet maps that intuitively displays the density of historical user trajectories during the same period through color depth; darker colors represent denser passenger flow, providing a direct basis for quantifying historical passenger flow distribution. Changes in corresponding area navigation requests: The difference between the navigation request volume in the target area and that in ordinary environments when a specific environmental marker appears (e.g., navigation requests increase by 15% on rainy days compared to sunny days), reflecting the impact of the environment on travel demand.

[0046] In this embodiment, taking "predicting the passenger flow in a certain city's XX business district from 10:00 to 12:00 on Saturdays" as an example: First, the time attribute of the prediction period is determined to be "weekend + midday". Data from all Saturdays from 10:00 to 12:00 within the past 30 days (preset period length) are selected. This data covers 6 rainy days and 24 sunny days (different environmental conditions). The internet map-related data in the same period data includes: user trajectory heatmap (showing that the core area of ​​the business district was the darkest in color from 11:00 to 12:00 on past Saturdays, corresponding to the highest passenger flow density), navigation request volume (average 12,000 times / 2 hours, reaching 14,000 times / 2 hours on rainy days), and POI activity. The data collection included activity records (eight Saturdays from 10:00 to 12:00 in the past 30 days where food events were held in the "first-floor dining area of ​​XX business district"); weather data for these 30 days was obtained through the historical weather interface of the internet map; and records of "road construction around the business district" were obtained through the traffic event interface. It was found that when the "rainy day" marker appeared in the historical data of the internet map, the number of navigation requests to the business district increased by about 18% compared to sunny days; and when the "food event in the dining area" marker appeared, the number of navigation requests increased by 25% compared to when there were no events. The correlation between these two "specific environmental markers and changes in navigation requests" was incorporated into the subsequent extraction of environmental correlation features to provide historical pattern support for predicting Saturday customer flow.

[0047] S102, Based on the dynamic monitoring data and historical benchmark data, extract time periodic features, spatial transfer features and environmental correlation features, and fuse them to form a spatiotemporal feature matrix.

[0048] In one embodiment, the step of extracting the time periodicity feature includes:

[0049] The historical data of the Internet map is analyzed using a sliding window method with a preset duration to extract the time lag relationship between the peak of Internet map search volume and the peak of passenger flow density; the time lag relationship of the preset quantity period is statistically calculated, and the calculation result is used as the core parameter of the time periodicity feature. The core parameter is used to predict the time when the passenger flow peak occurs in the target period.

[0050] In specific embodiments, the peak search volume on internet maps refers to the moment when the number of requests for keywords (such as "XX subway station" or "XX shopping mall") for a target area on internet maps reaches its highest value during the same historical period. This reflects the peak of user attention to the target area and is usually earlier than the actual peak passenger flow. Peak passenger flow density refers to the moment when the number of passengers per unit space within the target area reaches its highest value during the same historical period. This is a core indicator of the time-periodic characteristics (unit: people / 100 square meters). Time lag relationship refers to the time difference between the peak search volume on internet maps and the peak passenger flow density (e.g., search volume peaks at 7:20, passenger flow density peaks at 7:30, with a lag of 10 minutes). This reflects the time pattern of users' "search-travel-arrival" and is key to predicting the peak passenger flow moment.

[0051] In this embodiment, taking "extracting the time periodic characteristics of the morning peak (7:00-9:00) of a certain city's XX subway station" as an example: a sliding window method with a preset duration of 10 minutes is used to analyze the historical data of the Internet map for the past 12 working days (preset quantity of the same period). Within each window, the search volume of the keyword "XX subway station" and the passenger flow density of the corresponding grid are counted. It is found that the peak of Internet map search volume on each working day occurs around 7:20 (e.g., 7:21 on July 1st and 7:19 on July 8th), and the corresponding peak of passenger flow density occurs around 7:30 (e.g., 7:31 on July 1st and 7:29 on July 8th). The average of the time lag relationship (8-12 minutes) of these 12 periods is taken to obtain a statistical result of 10 minutes, which is used as the core parameter of the time periodic characteristics. When predicting "next Monday's morning peak", if the Internet map search volume is detected to peak at 7:20, the peak of passenger flow density can be predicted around 7:30 based on the core parameter, providing a time basis for advance diversion.

[0052] In one embodiment, the step of extracting the spatial transfer features includes:

[0053] A spatial relationship graph of the target area is constructed based on the path topology data of the Internet map. The path topology data includes road connection relationships and passage restrictions. The Internet map grid is used as the node of the spatial relationship graph, and the user transfer frequency between grids is used as the edge weight of the spatial relationship graph. The spatial transfer probability between grids is calculated by a preset probability model, and the spatial transfer probability is corrected by combining the types of POIs on the Internet map. The types of POIs include service facilities and traffic nodes.

[0054] In a specific embodiment, path topology data refers to structured data in an internet map describing the connections between roads, passages, and paths within a target area. This includes "which paths are interconnected" (road connection relationships) and "which paths are prohibited / restricted" (passage restrictions), forming the basis for constructing spatial associations. Spatial association graph: A graphical model constructed based on path topology data, describing the relationships between different spatial units within the target area, used to quantify passenger flow transfer patterns in the spatial dimension. Road connection relationships: Information in the path topology data describing the connectivity between roads and passages (e.g., "the subway entrance A passage is directly connected to the transfer passage," "the escalator on the first floor of the shopping mall is connected to the passage in the second-floor retail area"), determining the range of paths that passenger flow can transfer. Nodes: Elements in the spatial association graph representing subdivided spatial units within the target area, specifically referring to internet map grids (e.g., G1, G2), each node corresponding to an actual spatial area. Edge weights: The weight value corresponding to the line segment connecting two nodes in the spatial association graph, specifically referring to the frequency of user transfers between two internet map grids (e.g., the transfer frequency from grid G1 to G2 is 500 times / hour). A higher weight indicates a closer passenger flow association between the two grids.

[0055] In this embodiment, taking "extracting the spatial transfer characteristics of a subway station in a certain city during the morning rush hour" as an example: First, a spatial association graph is constructed based on the path topology data of an internet map. The road connection relationships in the path topology data show that "metro entrance A grid G1 is directly connected to transfer passage grid G2" and "transfer passage grid G2 is connected to Line 4 platform grid G3". The passage passage restriction shows that "during the morning rush hour, G1→G2→G3 only allows one-way passage". G1, G2, and G3 are used as nodes in the spatial association graph. The frequency of user transfers from G1 to G2 during past morning rush hours is 800 times / hour, and the frequency of transfers from G2 to G3 is 750 times / hour. The transfer frequency is used as the edge weight between nodes. A Marshall method is used. The Kov chain model (preset probability model) calculates the spatial transfer probability: the probability of G1→G2 = 800 times / total exit frequency of G1 950 times ≈ 84%, and the probability of G2→G3 = 750 times / total inflow frequency of G2 800 times ≈ 94%. Further, the probability is corrected by combining the type of POI on the Internet map: there is one "convenience store POI" (service facility) in grid G2. Statistics show that 10% of users will stay in the convenience store. Therefore, the transfer probability of G2→G3 is corrected to 94%×(1-10%)≈85%. Finally, the spatial transfer characteristics of "G1→G2 probability 84% and G2→G3 probability 85%" are obtained, which provides a spatial pattern basis for predicting the flow of people between grids.

[0056] S103, input the spatiotemporal feature matrix into the dynamic weight prediction model, and output the passenger flow prediction results including specific regions and time periods.

[0057] In one embodiment, the weight adjustment step of the dynamic weight prediction model includes:

[0058] The system presets a real-time trajectory update frequency threshold for the internet map and a historical navigation request data integrity threshold. When the real-time trajectory update frequency exceeds the update frequency threshold, the weight ratio of spatial transfer features is adjusted to a preset range, wherein the weight of internet map path transfer probability accounts for a preset proportion of the total weight of spatial features. When the integrity of historical navigation request data exceeds the integrity threshold, the weight ratio of time periodic features is adjusted to a preset range, wherein the weight of the lag relationship between internet map search volume and passenger flow density accounts for a preset proportion of the total weight of time features. The dynamic weight prediction model adopts a fusion architecture combining temporal and spatial networks to process time features and spatial features respectively.

[0059] In specific embodiments, the weighting percentage of spatial transfer features is as follows: Spatial transfer features account for the proportion of the total input features of the dynamic weight prediction model (e.g., 60%, 70%). A higher weight indicates that the model relies more on spatial features to predict passenger flow. The weighting percentage of time-periodic features is as follows: Time-periodic features account for the proportion of the total input features of the dynamic weight prediction model (e.g., 55%-65%). A higher weight indicates that the model relies more on historical time patterns to predict passenger flow. The weighting of the lag relationship between internet map search volume and passenger flow density is as follows: The proportion of this sub-feature, "search volume-passenger flow density lag relationship," in the total weighting of time-periodic features (e.g., 45%-55%). Because this lag relationship can predict peak passenger flow times, a higher proportion is needed to highlight its importance. Temporal networks are neural networks that process time-dimensional features (e.g., LSTM long short-term memory networks, GRU gated recurrent units), which are adept at capturing long-term dependencies in time series (e.g., passenger flow fluctuations over time) and are used to process time-periodic features. Spatial networks: Neural networks that process spatial dimensional features (such as GCN graph convolutional networks and GAT graph attention networks), which are good at capturing the node relationships in graph structure data (such as the passenger flow transfer patterns between grids) and are used to process spatial transfer features.

[0060] In this embodiment, taking "weight adjustment of dynamic weight prediction model for morning peak hours at a subway transfer station in a certain city" as an example: First, two core thresholds are preset: the real-time trajectory update frequency threshold for the internet map is 1500 records / minute, and the historical navigation request data integrity threshold for the internet map is 90%. When monitoring at 7:00 AM during the morning peak, the real-time trajectory update frequency for the internet map is 1800 records / minute (higher than the threshold), indicating sufficient real-time data. The weight ratio of spatial transfer features is adjusted to a preset range of 60%-70%, of which the weight of "internet map path transfer probability" (e.g., G1→G2 probability 84%) accounts for 55% of the total weight of spatial features (preset ratio), ensuring that the model prioritizes real-time spatial flow patterns. If, during a certain morning peak, the real-time trajectory update frequency is only 10 records / minute due to signal problems... If the number of requests per minute is 00 (below the threshold) and the integrity of historical navigation request data is 92% (above the threshold), then the weight of time-periodic features will be adjusted to a preset range of 55%-65%. Among them, the weight of "the lag relationship between internet map search volume and passenger flow density" (such as passenger flow peaking 10 minutes after search volume peaks) accounts for 50% of the total weight of time features (preset ratio), ensuring that the model relies on reliable historical time patterns. This dynamic weight prediction model adopts a fusion architecture of "LSTM temporal network + GCN spatial network". The LSTM network processes time-periodic features (such as passenger flow peak time prediction), and the GCN network processes spatial transfer features (such as inter-grid transfer probability). Finally, the prediction results are output by integrating the two types of features through dynamic weights, taking into account the prediction accuracy under different data conditions.

[0061] In one embodiment, the passenger flow prediction result includes a unique ID of the internet map grid, the prediction period, passenger flow density, internet map heat map level, and congestion warning level; the heat map level is divided according to a preset density standard; the congestion warning level is divided according to a preset risk standard, and the warning level is used to trigger different prompt styles on the internet map.

[0062] In a specific embodiment, the unique ID of the internet map grid is a unique identifier for each grid of a preset size in the internet map (e.g., G1, G15, G208). This identifier is unique and accurately corresponds to the actual spatial area (e.g., G15 corresponds to the "100m x 100m area of ​​Metro Exit A"). It is the core identifier linking the prediction result to the actual space. The prediction time period is the specific time interval corresponding to the prediction result, accurate to the minute (e.g., 7:30-7:40, 10:00-10:15) to ensure the prediction result is time-specific and can be used for time-segmented adjustments. Passenger flow density is the number of passengers per unit space in the area corresponding to the unique ID of the internet map grid within the prediction time period. This is the core quantitative indicator for passenger flow prediction, typically expressed as "people / 100 square meters" or "people / 50 square meters," directly reflecting the density of passenger flow. The internet map heatmap level converts passenger flow density into a color classification standard for the internet map heatmap. The predicted passenger flow density is visually displayed through color depth (e.g., red represents high density, yellow represents medium density, and green represents low density), providing a way to intuitively present the prediction results to users or operators. Preset density standards: Pre-defined rules for classifying heat map levels on internet maps (e.g., "Red: Passenger density > 500 people / 100 square meters; Yellow: 300-500 people / 100 square meters; Green: < 300 people / 100 square meters"). These standards need to be set in conjunction with the carrying capacity of the target area (e.g., a high threshold for a subway passage with high carrying capacity; a low threshold for a shopping mall corridor with low carrying capacity). Congestion warning levels: Passenger flow risk levels (e.g., Level 1, Level 2, Level 3) are determined based on a comparison of passenger density and the carrying capacity of the target area. These levels trigger control measures of varying intensities, with higher levels indicating higher congestion risk. Preset risk standards: Pre-defined rules for classifying congestion warning levels (e.g., "Level 1 warning: Passenger density > 120% of carrying capacity; Level 2 warning: 100%-120% of carrying capacity; Level 3 warning: < 100% of carrying capacity"). Carrying capacity needs to be calculated based on the type of area (subway passage, shopping mall, venue) (e.g., a subway passage with a carrying capacity of 600 people / 100 square meters). Different display styles of prompts on internet maps: The differentiated display format when internet maps push congestion warnings to users or operators must match the congestion warning level to ensure the effectiveness of the warning information transmission (e.g., a strong pop-up prompt for a level 1 warning, a banner prompt for a level 2 warning, and no prompt for a level 3 warning).

[0063] In this embodiment, taking "passenger flow forecasting at the interchange station of Metro Line 2 and Line 4 in a certain city during the morning peak (predicted period: 7:30-8:00 the next day)" as an example, the process of the method is as follows:

[0064] First, the target area was defined as "a 1-kilometer radius around the transfer station defined by the Gaussian projection coordinate system of the internet map (including 300 100m x 100m internet map grids from G101 to G400)". Two types of data were obtained from the internet map: one was dynamic monitoring data (collected through the internet map app, such as "User A's internet map coordinates at 7:00 are grid G128, real-time navigation path is 'Exit B → transfer passage → Line 4 platform', and the direction of movement is transfer"). The congestion level is 'mild', as shown by the real-time online map display of the current trajectory of grid G128 at a frequency of 1800 updates per minute. Secondly, there is historical baseline data (online map data from the morning rush hour of the past 30 working days, 7:30-8:00 AM, such as "the average daily passenger flow heat value of grid G128 (transfer corridor) is 88, and the average daily navigation requests on the online map during the same period are 15,000; among them, the navigation requests on 10 rainy days are 18% higher than on sunny days, and the passenger flow density of grid G128 is 12% higher than on sunny days").

[0065] Next, based on the above data (including internet map features), three types of features were extracted: time periodicity features (from historical internet map data, it was found that "during the morning rush hour of the past 30 working days, after the search volume for 'transfer station' on the internet map peaked at 7:20, the passenger flow in grid G128 increased by 15% within 10 minutes, with the peak passenger flow period being from 7:45 to 7:55") and spatial transfer features (combined with the path topology of the internet map, it was calculated that "92% of users entering the station from grid G128 (Exit B) chose the shortest path recommended by the internet map, 'Exit B → Transfer Passage,' and transferred to grid G135 (the core area of ​​the transfer passage)." ), Environmental correlation characteristics (combining weather correlation data from the Internet map, it is determined that "if the next day is a rainy day, the passenger flow density of grid G135 needs to be increased by 12% based on the historical average of 450 people / 100 square meters"); then, the three types of features are aligned according to the dimension of "Internet map grid (G101-G400) - 10-minute time segment (7:30-7:40, etc.)" and fused to form a spatiotemporal feature matrix (such as the feature value of "G135 grid - 7:45-7:55" includes "time peak increase of 15%, path transfer rate of 92%, rainy day increase of 12%, Internet map historical heat value of 88");

[0066] Finally, the spatiotemporal feature matrix is ​​input into the dynamic weight prediction model (because the real-time trajectory update frequency of the internet map is 1800 records / minute > the threshold of 1500, the model increases the weight of "internet map path transfer feature" to 65%), and outputs the passenger flow prediction results adapted to the internet map: "The next day from 7:30 to 8:00, the predicted passenger flow density of the internet map grid G135 (core area of ​​the transfer passage) from 7:45 to 7:55 is 550 people / 100 square meters (corresponding to 'red congestion' on the internet map heat map), and the passenger flow density of the G128 grid (Exit B) is 7." The system predicts a passenger flow density of 380 people per 100 square meters from 7:35 to 7:45. Simultaneously, it links with the internet map navigation module to push real-time alerts to users planning to travel through the G135 grid between 7:30 and 8:00 the following day: "The core area of ​​the transfer corridor is congested from 7:45 to 7:55. It is recommended to adjust the departure time to before 7:40 or choose the G129 grid (alternate route) for an alternative route." Furthermore, it collects user feedback on route adjustments through the internet map (e.g., if 30% of users adopt the alternative route suggestion), which is used to optimize the feature weights of the next round of dynamic weight prediction model.

[0067] The embodiments of the present invention can achieve the following advantages:

[0068] Using internet maps as the core carrier, three major improvements are achieved: First, the data source is more accurate, relying on the positioning and navigation data of internet maps to solve the problem of traditional data fragmentation; second, feature extraction is more realistic, combining the spatial topology, POI and scene data of internet maps to make spatiotemporal features more match users' travel behavior; third, the strategy implementation is more efficient, the prediction results can be directly linked to internet map navigation to form a closed loop of "data collection-feature extraction-prediction-control-feedback", which significantly improves the practicality of passenger flow prediction and the real-time nature of control, and is suitable for various public scenarios that need to be linked with users' travel behavior.

[0069] See Figure 2 This invention also provides a passenger flow prediction device 400 based on Internet map data, which includes an acquisition unit 401, an extraction unit 402, and an output unit 403.

[0070] The acquisition unit is used to acquire dynamic monitoring data and historical benchmark data of the target area. The dynamic monitoring data includes user trajectory location and real-time movement direction, and the historical benchmark data includes passenger flow distribution and environmental impact parameters during the same period.

[0071] In one embodiment, the step of acquiring dynamic monitoring data of the target area includes:

[0072] The location of the user's trajectory is collected through the positioning module of the Internet map software, and the positioning module includes satellite positioning and base station positioning;

[0073] By combining the real-time navigation function of the Internet map to extract the user's real-time movement direction, the real-time path congestion data of the Internet map is collected simultaneously. The congestion data includes the congestion level and the average traffic speed of the corresponding path. The congestion level is divided according to preset rules.

[0074] In one embodiment, the step of obtaining the historical benchmark data includes:

[0075] Select periodic data with the same time attributes as the prediction period. The time attributes include weekday / weekend attributes and peak / off-peak period attributes. The length of the selected periodic data is a preset period length.

[0076] The internet map-related data in the same periodic data includes historical user trajectory heatmaps, navigation request volume, and POI activity records for the same period. The POI activity records include the time and location information of commercial and public activities.

[0077] The environmental impact parameters are supplemented through the historical weather interface and traffic event interface of the Internet map. The specific correlation logic is that when a specific environmental identifier and the corresponding area navigation request volume change occur simultaneously in the historical data of the Internet map, the correlation is included in the basis for extracting environmental correlation features.

[0078] The extraction unit is used to extract time periodic features, spatial transfer features and environmental correlation features based on the dynamic monitoring data and historical benchmark data, and fuse them to form a spatiotemporal feature matrix.

[0079] In one embodiment, the step of extracting the time periodicity feature includes:

[0080] The historical data of Internet maps were analyzed using a sliding window method with a preset duration to extract the time lag relationship between the peak of Internet map search volume and the peak of passenger flow density.

[0081] Statistical calculations are performed on the time lag relationship of the preset quantity of the same period, and the calculation results are used as the core parameters of the time periodicity characteristics. The core parameters are used to predict the peak time of passenger flow in the target period.

[0082] In one embodiment, the step of extracting the spatial transfer features includes:

[0083] A spatial association graph of the target area is constructed based on the path topology data of the Internet map. The path topology data includes road connection relationships and passage restrictions. The Internet map grid is used as the node of the spatial association graph, and the user transfer frequency between grids is used as the edge weight of the spatial association graph.

[0084] The spatial transfer probability between grids is calculated by using a preset probability model, and the spatial transfer probability is corrected by combining the types of POIs on the Internet map, including service facilities and transportation nodes.

[0085] The output unit is used to input the spatiotemporal feature matrix into the dynamic weight prediction model and output passenger flow prediction results including specific regions and time periods.

[0086] In one embodiment, the weight adjustment step of the dynamic weight prediction model includes:

[0087] Preset thresholds for the real-time trajectory update frequency of internet maps and the integrity threshold for historical navigation request data;

[0088] When the real-time trajectory update frequency is greater than the update frequency threshold, the proportion of spatial transfer feature weights is adjusted to a preset range, wherein the Internet map path transfer probability weight accounts for a preset proportion of the total weight of spatial features.

[0089] When the integrity of historical navigation request data exceeds the integrity threshold, the weight ratio of time periodic features is adjusted to a preset range, wherein the weight of the lag relationship between internet map search volume and passenger flow density accounts for a preset proportion of the total weight of time features.

[0090] The dynamic weight prediction model adopts a fusion architecture that combines temporal and spatial networks to process temporal and spatial features respectively.

[0091] In one embodiment, the passenger flow prediction result includes a unique ID of the internet map grid, the prediction period, passenger flow density, internet map heat map level, and congestion warning level; the heat map level is divided according to a preset density standard; the congestion warning level is divided according to a preset risk standard, and the warning level is used to trigger different prompt styles on the internet map.

[0092] like Figure 3 As shown, Figure 3 This is a schematic block diagram of a computer device provided in an embodiment of this application. The computer device 500 can be a terminal or a server. The terminal can be an electronic device with communication functions, such as a smartphone, tablet, laptop, desktop computer, personal digital assistant, or wearable device. The server can be a standalone server or a server cluster composed of multiple servers.

[0093] The computer device 500 includes a processor 502, a memory, and a network interface 505 connected via a system bus 501. The memory may include a non-volatile storage medium 503 and internal memory 504.

[0094] The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. When the computer program 5032 is executed, it causes the processor 502 to execute a passenger flow prediction method based on Internet map data.

[0095] The processor 502 provides computing and control capabilities to support the operation of the entire computer device 500.

[0096] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute a passenger flow prediction method based on Internet map data.

[0097] The network interface 505 is used for network communication with other devices. Those skilled in the art will understand that the above structure is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device 500 to which the present application is applied. A specific computer device 500 may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements.

[0098] It should be understood that in the embodiments of this application, the processor 502 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0099] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program may be stored in a storage medium, which is a computer-readable storage medium. The computer program is executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.

[0100] Therefore, the present invention also provides a storage medium. This storage medium can be a computer-readable storage medium. The storage medium stores a computer program.

[0101] The storage medium is a physical, non-transient storage medium, such as a USB flash drive, external hard drive, read-only memory (ROM), magnetic disk, or optical disk, or any other physical storage medium capable of storing program code. The computer-readable storage medium can be non-volatile or volatile.

[0102] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0103] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of each unit is merely a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0104] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the device of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0105] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0106] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0107] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Since these modifications and variations fall within the scope of the claims and their equivalents, this invention also intends to include these modifications and variations.

[0108] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A passenger flow prediction method based on Internet map data, characterized in that, The method comprises: Obtain dynamic monitoring data and historical benchmark data of the target area based on an Internet map, wherein the dynamic monitoring data comprises user trajectory position and real-time moving direction, and the historical benchmark data comprises contemporaneous passenger flow distribution and environmental impact parameters; Based on the dynamic monitoring data and the historical benchmark data, extract time periodicity features, space transfer features and environmental correlation features, and fuse to form a space-time feature matrix; Input the space-time feature matrix into a dynamic weight prediction model to output passenger flow prediction results for specific areas and time periods; The weight adjustment step of the dynamic weight prediction model comprises: Pre-set an Internet map real-time trajectory update frequency threshold and a historical navigation request data integrity threshold; When the real-time trajectory update frequency is greater than the update frequency threshold, the space transfer feature weight proportion is adjusted to a pre-set range, wherein the Internet map path transfer probability weight accounts for a pre-set proportion of the total space feature weight; When the historical navigation request data integrity is greater than the integrity threshold, the time periodicity feature weight proportion is adjusted to a pre-set range, wherein the Internet map search volume-passenger flow density lag relationship weight accounts for a pre-set proportion of the total time feature weight; The prediction model adopts a fusion architecture combining time series network and space network to process time features and space features respectively.

2. The method of claim 1, wherein, The acquisition step of the dynamic monitoring data of the target area comprises: Collect user trajectory positions through the positioning module of the Internet map software, wherein the positioning module comprises satellite positioning and base station positioning; Extract the real-time moving direction of the user in combination with the real-time navigation function of the Internet map, and synchronously collect real-time path congestion data of the Internet map, wherein the congestion data comprises congestion levels and average travel speeds of corresponding paths, and the congestion levels are divided according to pre-set rules.

3. The method of claim 1, wherein, The acquisition step of the historical benchmark data comprises: Select contemporaneous period data with the same time attribute as the prediction period, wherein the time attribute comprises weekday / weekend attribute and peak / flat period attribute, and the selection length of the contemporaneous period data is a pre-set period length; The Internet map related data in the contemporaneous period data comprises historical contemporaneous user trajectory heat maps, navigation request volumes and POI activity records, wherein the POI activity records comprise time and position information of commercial activities and public activities; The environmental impact parameters are supplemented through the historical weather interface and traffic event interface of the Internet map, and the specific correlation logic is that when a specific environmental identifier and a corresponding change in navigation request volume in the Internet map historical data appear at the same time, the correlation is included in the extraction basis of the environmental correlation features.

4. The method of claim 1, wherein, The extraction step of the time periodicity features comprises: Analyze the historical data of the Internet map using a pre-set length sliding window method to extract the time lag relationship between the search volume peak and the passenger flow density peak of the Internet map; Statistically calculate the time lag relationships of a pre-set number of contemporaneous periods, and use the calculation results as the core parameters of the time periodicity features, which are used to predict the passenger flow peak occurrence time of the target period.

5. The method of claim 1, wherein, The extraction step of the space transfer features comprises: The path topology data based on the Internet map is used to construct a space correlation graph of a target area, the path topology data including road connection relationship and channel traffic restriction, an Internet map grid is taken as a node of the space correlation graph, and user transfer frequency between grids is taken as an edge weight of the space correlation graph. A spatial transfer probability between grids is calculated through a preset probability model, and the spatial transfer probability is corrected in combination with a type of POI of the Internet map, the type of the POI including a service facility and a traffic node.

6. The method of claim 1, wherein, The passenger flow prediction result includes an Internet map grid unique ID, a prediction period, a passenger flow density, an Internet map heat map level and a congestion early warning level, the heat map level is divided according to a preset density standard, the congestion early warning level is divided according to a preset risk standard, and the early warning level is used to trigger different prompt styles of the Internet map.

7. An Internet map data-based passenger flow prediction device characterized by comprising: The method comprises the following steps: An acquisition unit is configured to acquire dynamic monitoring data and historical reference data of a target area, the dynamic monitoring data including user trajectory positions and real-time moving directions, and the historical reference data including contemporaneous passenger flow distribution and environmental influence parameters; An extraction unit is configured to extract time periodicity features, spatial transfer features and environmental correlation features based on the dynamic monitoring data and the historical reference data, and fuse the features to form a time-space feature matrix; An output unit is configured to input the time-space feature matrix into a dynamic weight prediction model, and output a passenger flow prediction result including specific areas and time periods; The weight adjustment step of the dynamic weight prediction model comprises the following steps: presetting an Internet map real-time trajectory update frequency threshold and a historical navigation request data integrity threshold; when the real-time trajectory update frequency is greater than the update frequency threshold, the spatial transfer feature weight proportion is adjusted to a preset range, wherein the Internet map path transfer probability weight accounts for a preset proportion of the total spatial feature weight; when the historical navigation request data integrity is greater than the integrity threshold, the time periodicity feature weight proportion is adjusted to a preset range, wherein the Internet map search volume-passenger flow density lag relationship weight accounts for a preset proportion of the total time feature weight; The prediction model adopts a fusion architecture combining a time sequence network and a space network to process time features and space features, respectively.

8. A computer device, comprising: The system comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; The memory is configured to store a computer program; The processor is configured to execute the program stored on the memory to implement the steps of the method according to any one of claims 1-6.

9. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method according to any one of claims 1-6.

Images