Urban riding traffic flow prediction method and system
By using space-weighted Tyson polygon buffer and limit gradient enhancement algorithm in the prediction of shared bicycle travel, the problem that existing technology is difficult to predict shared bicycle traffic during the stable epidemic is solved, and more efficient and accurate prediction results are achieved.
Patent Information
- Application Number
- CN202510017352.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-06
- Publication Date
- 2025-06-03
AI Technical Summary
The existing methods of shared bicycle travel prediction are difficult to effectively deal with nonlinear relationships and complex patterns, especially in the fluctuations of the epidemic period, and it is impossible to accurately predict the traffic flow of shared bicycles.
The space-weighted Tyson polygon buffer is used instead of the traditional fishing net segmentation method, to obtain structured data, retain spatial heterogeneity and autocorrelation information, and to fit the data using the extreme gradient boosting (XGBoost) algorithm to realize the prediction of shared bicycle travel volume with geospatial characteristics.
It improves the accuracy and speed of traffic flow prediction of shared bicycles, and can better cope with nonlinear relationships and complex patterns, especially during the fluctuation period of the epidemic.
Smart Images
Figure CN120087779A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of shared bicycle travel traffic flow prediction, and more particularly to a method and system for predicting urban cycling traffic flow. Background Art
[0002] During the pandemic, due to public health considerations, the government restricted the use of centralized public transportation systems for a certain period of time. As the main alternative mode of travel, the usage rate and frequency of shared bicycles have changed significantly compared to the past. The experience during the pandemic has shown that the shared bicycle system has great potential to enhance the resilience of the urban transportation system in disruptive events. Paying attention to the overall impact mechanism of shared bicycle travel can greatly improve the urban resilience in the post-pandemic era and enhance the rationality of urban transportation system design in the era of normalized epidemic prevention. Existing shared bicycle travel prediction methods are mostly based on traditional statistical regression methods or traditional machine learning. These two types of methods often cannot well handle non-linear relationships and complex patterns in prediction, especially for shared bicycle traffic flow with significant spatial heterogeneity. They cannot pay attention to the local effects of features and cannot reflect the spatial fluctuations of feature impacts. In addition, existing shared bicycle predictions related to the epidemic mostly target the entire pre-epidemic, during-epidemic, and post-epidemic periods, the epidemic lockdown period, and the period of gradual opening of the epidemic, but lack predictions for the fluctuating state during the stable period of the epidemic. This corresponds to the peak of fluctuations in the post-pandemic normalization stage after policy opening, which is of great significance for the present and future coexisting with the epidemic.
[0003] Therefore, it is a technical problem to be solved to provide a method for predicting shared bicycle traffic flow applicable to the stable period of the epidemic. Summary of the Invention
[0004] The purpose of the present invention is to overcome the above-mentioned defects of the existing technologies and provide a method and system for predicting urban cycling traffic flow. During the acquisition process of structured data, the Thiessen polygon buffer considering spatial weighting is used to replace the traditional fishing net segmentation method, so that both spatial heterogeneity and spatial autocorrelation information are retained in the structured data. On this basis, the structured data obtained by the above method is fitted with the Extreme Gradient Boosting (XGBoost) algorithm, realizing the prediction of shared bicycle travel volume considering geographical spatial features.
[0005] The purpose of the present invention can be achieved by the following technical solutions:
[0006] According to the first aspect of the present invention, a method for predicting urban cycling traffic flow is provided. The method uses an Extreme Gradient Boosting model to predict the shared bicycle traffic flow during the stable period of the epidemic, including:
[0007] Preprocess the station information of shared bicycle stations in the target urban space, and construct a Thiessen polygon buffer with a preset size based on the preprocessed information;
[0008] Collect meteorological factors, epidemic factors, and urban space factors in the Thiessen polygon buffer, and process them to obtain meteorological indicators, epidemic variables, and urban space factor indicators;
[0009] Construct a data set based on the meteorological indicators, epidemic variables, and urban space factor indicators, and input it into the extreme gradient boosting model to output the predicted traffic flow of shared bicycle trips.
[0010] As a preferred technical solution, the preprocessing includes:
[0011] Based on the station information, use the Gaussian kernel density interpolation method to calculate the kernel density data of shared bicycle stations in the target urban space. Its expression is:
[0012]
[0013] Among them, i represents the number of the station, and i ∈ [1, n]; n represents the total number of stations; τ represents the bandwidth parameter, which controls the width of the kernel function and determines the influence range of each station; k represents the Gaussian kernel function, and its formula is s i represents the location of the i-th station; s represents the location of the target point; Λ represents the kernel density estimation parameter;
[0014] Normalize the kernel density data. Its expression is:
[0015]
[0016] Among them, M represents the normalized density value; max represents the maximum value in the kernel density data; min represents the minimum value in the kernel density data;
[0017] Construct a cost grid equation, and perform map algebra calculations based on the normalized kernel density data.
[0018] As a preferred technical solution, the preset size is 250 meters.
[0019] As a preferred technical solution, the meteorological indicators include: average temperature, wind speed, average precipitation, total precipitation, average minimum temperature, average maximum temperature.
[0020] As a preferred technical solution, the epidemic factors include the epidemic infection rates and the shared bicycle travel volume data in different periods; the epidemic variables are obtained by processing the epidemic infection rates and the shared bicycle travel volume data in different periods.
[0021] As a preferred technical solution, the method for obtaining the urban spatial factor indicators is as follows: based on the urban spatial factors, use spatial syntax to generate the urban spatial factor indicators for each land use type; the land use types include factories, commerce, apartments, offices, public areas, and transportation; the urban spatial factor indicators include gravity values, integration values, proximity values, and straightness values.
[0022] As a preferred technical solution, the method for constructing the data set is as follows:
[0023] Use the urban network POI analysis method to integrate the gravity values of the urban spatial factor indicators into the Thiessen scale of the Thiessen polygon buffer.
[0024] Use the Kriging interpolation method to interpolate and calibrate the spatial indicators, epidemic variables, and urban spatial factor indicators.
[0025] As a preferred technical solution, the training method of the extreme gradient boosting model is as follows:
[0026] Construct a historical data set and divide it into a training set and a test set according to a preset ratio.
[0027] Calculate the mean value of each type of data in the training set as the predicted value of the model input, and initialize the model parameters.
[0028] Based on the training set, perform iterative training until the iterative stop condition is met. In each iterative training, calculate the residual between the predicted value and the target value, fit a new regression tree for residual fitting, and optimize the model parameters based on the fitted residuals.
[0029] Use the test set to evaluate the model performance and select the model with the best performance.
[0030] As a preferred technical solution, the iterative stop conditions include:
[0031] The number of iterations meets the maximum number of iterations; or, the difference between the residual of the previous iterative training and the residual of the next iterative calculation is less than a preset value; or, the performance of the model deteriorates on the test set.
[0032] According to the second aspect of the present invention, a system for predicting urban cycling traffic flow is provided, and the system is used to implement the above method.
[0033] Compared with the prior art, the present invention uses Thiessen polygons with spatial weighting to replace the traditional fishing net segmentation method to construct a buffer zone, so that the obtained structured data retains both spatial heterogeneity and spatial autocorrelation information. The extreme gradient boosting model is used to process the collected structured data, which solves the problem that the traditional statistical regression method cannot well handle non-linear relationships, and also solves the problem that previous machine learning models ignore the local effects of features and spatial fluctuations, resulting in fast speed and high prediction accuracy for shared bicycle traffic flow prediction. BRIEF DESCRIPTION OF THE DRAWINGS
[0034] Figure 1 is the overall framework diagram of the present invention;
[0035] Figure 2 is the method flow chart of the present invention;
[0036] Figure 3 is the schematic diagram of the target city space in the embodiment of the present invention;
[0037] Figure 4 is the schematic diagram of epidemic data in the embodiment of the present invention;
[0038] Figure 5 is the schematic diagram of the trip distribution data of shared bicycles in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0039] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.
[0040] Obviously, the accompanying drawings in the following description are only some examples or embodiments of the present application. For those of ordinary skill in the art, without creative efforts, the present application can also be applied to other similar scenarios based on these drawings. In addition, it can also be understood that although the efforts made in this development process may be complex and lengthy, for those of ordinary skill in the art related to the content disclosed in the present application, some design, manufacturing or production changes based on the technical content disclosed in the present application are only conventional technical means and should not be understood that the content disclosed in the present application is insufficient.
[0041] The present invention provides a method for predicting urban cycling traffic flow, which introduces a weighted Thiessen polygon to replace the traditional buffer construction method. Through geostatistical interpolation calibration and spatial statistics, meteorological factors, epidemic factors, and urban spatial factors are statistically included as input data for a machine learning model to more accurately predict the shared bicycle travel volume during the epidemic fluctuation period. Therefore, due to the high efficiency, scalability, and excellent performance of Extreme Gradient Boosting (XGBoost) when performing tasks, especially on structured data, the present invention selects XGBoost as the shared bicycle traffic flow prediction model.
[0042] And to verify the feasibility and superiority of the method provided by the present invention, the present invention compares this method with traditional methods to verify the model performance. The overall process of the present invention is as detailed in Figure 1 shown, where the process of predicting the shared bicycle traffic flow using the method provided by the present application is as Figure 2 shown, including:
[0043] S1. Perform data preprocessing on the site information of shared bicycle stations in the target urban space, and construct a Thiessen polygon buffer with a preset size based on the preprocessed information.
[0044] S11. Although the first law of geography states that inverse distance weighting can be used as an algorithm considering spatial correlation, it is only suitable for discussing the radiation range of commercial facilities and does not consider the distribution scale of individual stations. Therefore, the present invention uses the Gaussian kernel density interpolation method to avoid this problem. Specifically, based on the above-mentioned site information, the Gaussian kernel density interpolation method is used to calculate the kernel density data of shared bicycle stations in the target urban space, and its expression is:
[0045]
[0046] where i represents the number of the station, and i ∈ [1, n]; n represents the total number of stations; τ represents the bandwidth parameter, which controls the width of the kernel function and determines the influence range of each station; k represents the Gaussian kernel function, and its formula is s i represents the position of the i-th station; s represents the position of the target point; Λ represents the kernel density estimation parameter.
[0047] The kernel density represents the aggregation degree of shared bicycle stations in the urban space through the kernel function. The higher the aggregation degree per unit area, the smaller the service range of a single station.
[0048] S12. Normalize the kernel density data, and its expression is:
[0049]
[0050] Among them, M represents the normalized density value; max represents the maximum value in the kernel density data; min represents the minimum value in the kernel density data.
[0051] S13. Construct a cost grid equation and perform map algebra calculation based on the normalized kernel density data.
[0052] S14. Construct a Thiessen polygon buffer based on the results of map algebra calculation.
[0053] The starting and ending points of the shared bicycle trips at the parking spots are restricted by the location of each parking spot, but most related studies use the service area within a certain radius around the bicycle station as an influencing factor for predicting the traffic flow of shared bicycles. Due to the uneven distribution of bicycle stations, when the distance is too close, the correlation between the independent variables of adjacent bicycle stations is easily ignored, and there is overlap in the service range. Through verification, it is found that a radius of 250m is suitable for modeling shared bicycle demand, and the application of Thiessen polygons can improve regression performance compared to using circular buffers alone. At the pedestrian scale, users tend to choose the bicycle station closest to the departure or destination, regardless of the distance to other walking stations. Selecting Thiessen polygons as the division method can ensure that each point in the service area is closest to all other stations. Therefore, the present invention establishes a 250-meter Thiessen polygon buffer.
[0054] S2. Collecting meteorological factors, epidemiological factors and urban space factors in the Thiessen polygon buffer zone, and processing them to obtain meteorological indicators, epidemiological variables and urban space factor indicators.
[0055] For meteorological factors, six indicators, including average temperature, wind speed, average precipitation, total precipitation, average minimum temperature and average maximum temperature, were selected as meteorological indicators; according to actual epidemiological factors, epidemiological infection rates and shared bicycle travel volume data in different periods were extracted and processed to obtain epidemiological variables; according to urban spatial factors, spatial syntax was used to generate urban spatial factor indicators for each land use type, where land use types include factories, commercial, apartments, offices, public and transportation, and urban spatial factor indicators include gravity value, degree of penetration, proximity and linearity. Detailed data and their sources are shown in Table 1:
[0056] Table 1 Data sheet
[0057]
[0058]
[0059] Among them, the calculation method of each urban space factor index is shown in Table 2:
[0060]
[0061] In the table, Jobs j represents the number of job positions in employment zone j; Time ij represents the travel time between residential area i and employment zone j; v represents the gamma coefficient; N represents the total number of nodes in the network; k represents a node; n jk (i) represents the number of paths passing through node i in the shortest path from node j to node k; n jk represents the number of shortest paths between nodes j and k; d ij represents the shortest distance between nodes i and j; represents the Euclidean distance between nodes i and j.
[0062] S3. Construct a dataset based on meteorological indicators, epidemic variables, and urban spatial factor indicators, and input it into the Extreme Gradient Boosting model to output the prediction of the traffic flow of shared bicycle trips.
[0063] Among them, the method for constructing the dataset is as follows:
[0064] S31. Use the urban network POI analysis method to integrate the gravity value of the urban spatial factor indicators into the Thiessen scale of the Thiessen polygon buffer.
[0065] S32. Use the Kriging interpolation method to interpolate and calibrate the spatial indicators, epidemic variables, and urban spatial factor indicators to provide standard variables for the Extreme Gradient Boosting model.
[0066] In the present invention, the training method of the Extreme Gradient Boosting model is as follows:
[0067] A1. Construct a historical dataset and divide it into a training set and a test set according to a preset ratio;
[0068] A2. Calculate the mean of each type of data in the training set as the predicted value of the model input, and initialize the model parameters;
[0069] A3. Perform iterative training based on the training set until the iterative stop condition is met, that is, the number of iterations meets the maximum number of iterations; or, the difference between the residual of the previous iterative training and the residual of the next iterative calculation is less than a preset value; or, the performance of the model on the test set deteriorates.
[0070] Calculate the residual between the predicted value and the target value in each iterative training, fit a new regression tree for residual fitting, and optimize the model parameters based on the fitted residual;
[0071] Evaluate the performance of the model using the test set and select the model with the optimal performance.
[0072] In the process of obtaining structured data, the present invention uses a Thiessen polygon including spatial weighting to replace the traditional fishing net segmentation method for buffer construction, retaining both spatial heterogeneity and spatial autocorrelation information in the structured data. On this basis, the XGBoost algorithm is used to fit the structured data obtained by the above method, realizing the prediction of shared bicycle trips considering geographical spatial features, which is more accurate.
[0073] The present invention also provides an urban cycling traffic flow prediction system for implementing the above method. Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working process described can refer to the corresponding process in the foregoing method embodiment and will not be elaborated herein.
[0074] To verify the feasibility and superiority of the method provided in this application, in this embodiment, taking New York City from November 1, 2020 to February 28, 2021 as an example, Manhattan, Brooklyn and Queens with shared bicycle stations are collected as test areas as Figure 3 shown, and a total of 1,219 samples of Thiessen polygons and weather, epidemic, urban space, and shared bicycle trip volume data are collected.
[0075] The collected weather, epidemic and urban space data are used as the data set, among which the epidemic data is as Figure 4 shown; the shared bicycle trip volume data is used as the output result, and its data distribution is as Figure 5 shown. Performance verification is carried out using respectively: Decision Tree (DT), Random Forest (RF), Extreme Gradient Boosting (ADA), Gradient Boosting Decision Tree (GBDT), Extreme Tree (ET), Category Boosting (CATBOOST), K-Nearest Neighbor (KNN), Back Propagation (BP), Support Vector Regression (SVR), Light Gradient Boosting Machine (LIGHTGBM) and Extreme Gradient Boosting (XGBOOST).
[0076] The results are shown in Table 3:
[0077] Table 2 Comparison results of machine learning model performance
[0078]
[0079] It can be seen that the XGBoost model performs the most prominently in these comparisons, with the highest R 2 value (0.912) and the lowest RMSE (5.696). This shows that XGBoost is superior to other models in terms of explaining data variation and prediction accuracy. For other models, such as MGWR, although they also show relatively high R 2 values, their RMSE is significantly higher than that of XGBoost.
[0080] In addition to comparing the performance with different machine learning models, the method of the present embodiment is also compared with the traditional spatial map model in terms of performance. The results are shown in Table 4:
[0081] Table 3 Performance result diagram of the traditional spatial map model
[0082]
[0083] It can be seen that the method of the present application has the highest R 2 value (0.912) and the lowest RMSE (5.696). This shows that it is superior to other models in terms of explaining data variation and prediction accuracy. Other models, such as MGWR, although also showing a relatively high R 2 value, its RMSE is significantly higher than that of this method.
[0084] Therefore, the method provided by the present invention has feasibility and superiority.
[0085] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims
1. A method for predicting urban cycling traffic flow, characterized in that: The method uses the extreme gradient boosting model to predict the shared bicycle traffic flow during the stable epidemic period, including: Preprocess the station information of the shared bicycle stations in the target city space, and build a Thiessen polygon buffer of a preset size based on the preprocessed information; Collecting meteorological factors, epidemiological factors and urban space factors in the Thiessen polygon buffer zone, and processing them to obtain meteorological indicators, epidemiological variables and urban space factor indicators; A data set is constructed based on the meteorological indicators, epidemiological variables and urban spatial factor indicators, and the extreme gradient boosting model is input to output the shared bicycle travel traffic flow prediction.
2. The method for predicting urban cycling traffic flow according to claim 1, characterized in that: The pretreatment comprises: Based on the station information, the Gaussian kernel density interpolation method is used to calculate the kernel density data of shared bicycle stations in the target city space, and its expression is: Where i represents the site number, and i∈[1,n]; n represents the total number of sites; τ represents the bandwidth parameter, which controls the width of the kernel function and determines the influence range of each site; k represents the Gaussian kernel function, whose formula is s i represents the location of the i-th site; s represents the location of the target point; Λ represents the kernel density estimation parameter; The kernel density data is normalized, and the expression is: Among them, M represents the normalized density value; max represents the maximum value in the kernel density data; min represents the minimum value in the kernel density data; Construct the cost grid equation and perform map algebra calculations based on the normalized kernel density data.
3. The method for predicting urban cycling traffic flow according to claim 1, characterized in that: The preset size is 250 meters.
4. The method for predicting urban cycling traffic flow according to claim 1, characterized in that: The meteorological indicators include: average temperature, wind speed, average precipitation, total precipitation, average minimum temperature, and average maximum temperature.
5. The method for predicting urban cycling traffic flow according to claim 1, characterized in that: The epidemic factors include epidemic infection rates and shared bicycle travel volume data in different periods; the epidemic variables are obtained by processing the epidemic infection rates and shared bicycle travel volume data in different periods.
6. The method for predicting urban cycling traffic flow according to claim 1, characterized in that: The method for obtaining the urban space factor indicators is: based on the urban space factors, using space syntax to generate urban space factor indicators for each land use type; the land use types include factories, businesses, apartments, offices, public and transportation; the urban space factor indicators include gravity value, penetration, proximity and linearity.
7. The method for predicting urban cycling traffic flow according to claim 6, characterized in that: The method for constructing the data set is: Using the urban network POI analysis method, the gravity value of the urban space factor index is integrated into the Thiessen scale of the Thiessen polygon buffer zone; The spatial indicators, epidemic variables and urban spatial factor indicators are interpolated and calibrated using the Kriging interpolation method.
8. The method for predicting urban cycling traffic flow according to claim 1, characterized in that: The training method of the extreme gradient boosting model is: Construct a historical data set and divide it into a training set and a test set according to a preset ratio; Calculate the mean of each type of data in the training set as the predicted value of the model input and initialize the model parameters; Iterative training is performed based on the training set until an iteration stop condition is met, residuals between predicted values and target values are calculated in each iterative training, a new regression tree is fitted for residual fitting, and model parameters are optimized based on the fitted residuals; The test set is used to evaluate the model performance and select the model with the best performance.
9. The method for predicting urban cycling traffic flow according to claim 8, characterized in that: The iteration stopping conditions include: The number of iterations meets the maximum number of iterations; or, the difference between the residual of the previous iteration training and the residual calculated in the next iteration is less than the preset value; or, the performance of the model on the test set deteriorates.
10. A city cycling traffic flow prediction system, characterized in that: The system is used to implement the method described in any one of claims 1 to 9.