Travel behavior analysis method for sparse trajectory of private electric bicycle
By identifying point clusters in trajectory data and expanding the random forest model, combining the spatiotemporal SHAP method, analyzing the sparse trajectory of private electric bicycles, the problems of insufficient processing capabilities of sparse data and fitting capabilities and interpretability in the prior art are solved, and the trade-offs on the sparse data in the existing technology are realized. The exploration of higher fitting accuracy and spatial and temporal changes is suitable for multi-city promotion.
Patent Information
- Application Number
- CN202411068720.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-06
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-06
AI Technical Summary
In the study of private electric bicycle travel behavior analysis, the problems of insufficient processing capabilities of sparse trajectory data, fitting capabilities and interpretability trade-offs, especially in the case of space-time heterogeneity, it is difficult to effectively explore the spatiotemporal and spatial changes of drivers.
A travel behavior analysis method for sparse trajectories of private electric bicycles is proposed. By identifying point clusters in the trajectory data, the classic random forest is expanded into a space-time random forest, and the nonlinear relationship between 14 influencing factors and travel is established, and the space-time SHAP method is used. Map Shapley values to the timeline and space axis to explore the spatiotemporal and spatial effects of drivers.
It achieves higher fitting accuracy and universality, and can directly infer human travel behavior through the real-time location of the electric bicycle, explore the spatiotemporal and spatial changes of the drivers, and is suitable for promotion in multiple cities.
Smart Images

Figure CN119004028A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data analysis, and in particular to a travel behavior analysis method for sparse trajectories of private electric bicycles. Background Art
[0003] At present, scholars at home and abroad have established a variety of analytical methods for electric bicycle travel to explore the driving mechanism of electric bicycle travel. For example, Wang et al. (2024) established a rideability assessment framework for urban streets based on electric bicycle trajectories, quantifying the quantitative linear relationship between cycling demand and urban environment rideability; Zhou et al. (2022) established a method for identifying commuting demand of urban people based on electric bicycle trajectories, and used the Geographically and Temporally Weighted Regression model (GTWR) to reveal the spatiotemporal impact of built environment, transportation facilities, socioeconomic and weather conditions on electric bicycle demand.
[0004] Although a lot of research has been done on electric bicycle travel, there are still shortcomings. Existing research mainly focuses on shared electric bicycle travel, while research on non-private bicycle travel is very limited (Liu et al., 2023; Yu et al., 2022). Second, existing methods are limited by the trade-off between fitting ability and interpretability, especially when considering spatiotemporal heterogeneity. For example, linear models can well explain the relationship between electric vehicle travel and driving factors, but the fitting ability is limited (linear regression, geographically weighted regression, spatiotemporal geographically weighted regression). Nonlinear models have strong fitting ability but lack interpretability (random forest, support vector machine, etc.).
[0005] In recent years, in order to strengthen the management of private electric bicycles, some cities have installed monitoring stations to collect their trajectories. Unlike the GPS trajectories of shared electric bicycles, the trajectories collected by monitoring stations are significantly sparse, thus protecting personal privacy. The collection of private electric bicycle trajectories has led to a large number of studies.
[0006] To this end, this application proposes a travel behavior analysis method for sparse trajectories of private electric bicycles, which can reveal phenomena that are difficult to discover with traditional methods, and provides a novel data mining method for in-depth understanding of electric vehicle travel patterns and behavioral characteristics. Summary of the invention
[0007] In order to make up for the deficiencies in the prior art, the present invention provides a travel behavior analysis method for sparse trajectories of private electric bicycles.
[0008] A travel behavior analysis method for sparse trajectories of private electric bicycles, characterized by comprising the following steps:
[0009] S1, identify point clusters in trajectory data, which include potential stay areas and confirmed stay areas, and extract the travel behavior of private electric bicycles based on the temporal relationship of point clusters;
[0010] S2, extending the classic random forest, proposed a spatiotemporal random forest to establish a nonlinear relationship between 14 influencing factors and private electric bicycle travel;
[0011] S3, extending the classic SHAP method, proposes spatiotemporal SHAP, maps the Shapley value to the time axis and space axis, and explores the spatiotemporal variation effects of driving factors.
[0012] Furthermore, in order to better implement the present invention, the data in S1 includes private electric bicycle trajectory data, traffic network data, point of interest data, Amap congestion data, weather data, housing price data and population data.
[0013] Furthermore, in order to better realize the present invention, the stay areas in the trajectory data in S1 are divided into two categories, one is the determined stay area, and the other is the potential stay area; when the electric vehicle is parked near the monitoring pile, the real-time trajectory points of the electric vehicle will be uploaded to the database at a certain frequency, and the trajectory points will form a point cluster near the monitoring pile, that is, the determined stay area; when the electric vehicle is not parked near the monitoring pile, the real-time trajectory points of the electric vehicle will be lost in the database, and when the trajectory points are monitored again, the first monitored area may be the area where the electric vehicle is parked, that is, the potential stay area; for the potential stay area, only the first point or the last point of the trajectory segment needs to be extracted, and for the determined stay area, it is extracted by the spatiotemporal clustering method. When the stay area in the trajectory points is identified, the travel of the electric bicycle can be generated according to the temporal relationship between the stay areas, including from the determined stay area to the determined stay area, from the determined stay area to the potential stay area, from the potential stay area to the determined stay area, and from the potential stay area to the potential stay area.
[0014] Further, in order to better realize the present invention, the spatiotemporal random forest method is as follows: the spatiotemporal random forest introduces the spatiotemporal neighborhood on the basis of the classic random forest, and uses the internal driving factors and external driving factors in the spatiotemporal neighborhood to jointly infer the future travel frequency, wherein the internal driving factor refers to the historical travel frequency in the target grid, and the external driving factor refers to the urban built environment, spatial location and time window information in the target grid; the spatiotemporal random forest model further improves the model's ability to capture spatiotemporal relationships by integrating external driving factors and internal driving factors, thereby improving the modeling accuracy of urban travel; this method needs to fuse the driving factors of the target grid itself with the driving factors in the neighborhood to ensure a one-to-one correspondence between the driving factors and the travel frequency, and the specific fusion method is as follows:
[0015]
[0016] Where: and They represent the kth driving factors of the target grid i before and after fusion, respectively, where the first driving factor is Fig.12 The fourteenth driving factor is Historical Trip Frequency in Fig.12 HousePrice in Ω i represents the set of spatiotemporal neighbors of the target grid i; Represents the distance between the target grid i and the neighborhood grid j. From the above formula, we can see that for the 5th to 13th driving factors, the fused driving factor is the sum of the driving factors in the neighborhood; for the 14th driving factor, the fused driving factor is the weighted mean of the driving factors in the neighborhood.
[0017] Further, in order to better implement the present invention, S3 is specifically as follows: a three-dimensional cube is established with the spatial grid index as the S axis, the time window as the T axis, and the driving factor as the F axis. For any driving factor in the three-dimensional cube, the Shapley value of the corresponding driving factor is calculated based on the game theory principle, and the Shapley value is assigned to the corresponding spatial position and time window through spatial mapping and time mapping. When all driving factors in the three-dimensional cube are mapped, the changes of the driving factors with different time periods and different spatial positions can be analyzed. The Shapley value of the driving factor in the three-dimensional cube is calculated by the following formula:
[0018]
[0019] Where: STRF represents the spatiotemporal random forest model; represents the kth driving factor of target grid i; N represents the total number of time windows × spatial grids, which can be decomposed into N S ×NT ; Represents the driving factor in a three-dimensional cube, given by Decompose to obtain; represents the Shapley value of the driving factor of the sth grid in the tth time window; when When it is greater than 0, it means that the driving factor has a positive effect on electric vehicle travel in the sth grid in the tth time window; When it is less than 0, it means that the driving factor has a negative effect on electric vehicle travel in the sth grid in the tth time window.
[0020] The beneficial effects of the present invention are:
[0021] Compared with the travel analysis method for shared electric bicycle trajectories, the method of the present invention has better universality. First, the present method can directly infer human travel behavior through the real-time location of electric bicycles, which reduces the dependence of travel behavior extraction on order data and is conducive to promotion in multiple cities. Secondly, the present invention also establishes a nonlinear driving mechanism modeling method that considers spatial correlation and heterogeneity. Compared with the existing modeling methods, the present invention not only achieves higher fitting accuracy, but also can explore the spatiotemporal variation effects of driving factors. Finally, the present invention can explore the spatiotemporal variation effects of driving factors in nonlinear models. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] Figure 1 A schematic diagram of a research area of an embodiment of the present invention;
[0023] Figure 2 A schematic diagram of the research framework for analyzing the spatiotemporal travel patterns of electric bicycles in the present invention;
[0024] Figure 3 Schematic diagram of extracting the number of electric bicycle trips for the present invention; (a) from a determined stop point to a determined stop point, (b) from a determined stop point to a potential stop point, (c) from a potential stop point to a determined stop point, (d) from a potential stop point to a potential stop point;
[0025] Figure 4 This is a schematic diagram of the spatiotemporal random forest of the present invention;
[0026] Figure 5 The spatial mapping and temporal mapping diagram of the Shapley value of the present invention;
[0027] Figure 6 Statistical distribution diagram of travel distance and travel duration at different time points of the present invention; (a) travel distance on working days, (b) travel distance on non-working days, (c) travel duration on working days, (d) travel duration on non-working days;
[0028] Figure 7 The relative travel frequency diagram of working days and non-working days of the present invention;
[0029] Figure 8 The spatiotemporal distribution diagram of the travel frequency of the present invention; (a) spatial distribution of travel frequency on working days, (b) spatial distribution of travel frequency on non-working days, (c) temporal distribution of travel frequency on working days and non-working days;
[0030] Fig. 9 Comparison results of the spatiotemporal random forest of the present invention and the baseline; (a) linear regression, (b) XGBoost, (c) random forest, (d) spatial random forest, (e) temporal random forest, (f) spatiotemporal random forest;
[0031] Fig.10 The driving factors of the present invention have an impact weight and a time heterogeneity diagram; (a) SHAP value of each driving factor; (b) SHAP value of commercial POI density in the time dimension;
[0032] Fig.11 The driving factors influence weights and their spatial heterogeneity diagrams of the present invention; (a) SHAP values of all driving factors, (b) number of commercial POIs in the spatial dimension, (c) SHAP value of commercial POI density in the spatial dimension;
[0033] Fig.12 is a table of driving factors in the spatiotemporal random forest of the present invention;
[0034] Fig.13 This is a table of fitting results (mean ± standard deviation) of the spatiotemporal random forest of the present invention and the baseline. DETAILED DESCRIPTION
[0035] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Generally, the components of the embodiments of the present invention described and shown in the accompanying drawings here can be arranged and designed in various different configurations.
[0036] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present invention.
[0037] Figure 1-Figure 13 This is a specific embodiment of the present invention, which is a travel behavior analysis method for sparse trajectories of private electric bicycles.
[0038] like Figure 1 As shown, this embodiment takes the area within the third ring road of Wuhan as the study area, and divides the study area into regular grids of 200 meters × 200 meters.
[0039] The data set used in this embodiment mainly includes private electric bicycle trajectory data, traffic network data, point of interest data, Amap congestion data, weather data, housing price data and population data.
[0040] The trajectory data of private electric bicycles is collected through monitoring stakes in the study area. When a private electric bicycle moves into the detection range of the monitoring stake, the real-time location information of the person will be uploaded to the database. The time span of the data set used in this embodiment is from December 28, 2020 to January 5, 2021, a total of 9 days.
[0041] In addition to the private electric bicycle trajectory data, this embodiment uses web crawler technology to obtain the traffic network data, point of interest data, Amap congestion data, housing price data and population data of the same period, so as to construct the driving factors of electric bicycle travel (which will be discussed further below). The traffic network data, point of interest data and Amap congestion data are from Amap open platform, the housing price data is from Fangtianxia.com, and the population data is from WorldPop grid.
[0042] like Figure 2 As shown, the travel behavior analysis framework proposed in this embodiment is divided into three parts: extraction of private electric bicycle travel based on sparse trajectories, travel mechanism modeling in the spatiotemporal domain, and analysis of driving factors that take into account spatiotemporal heterogeneity. First, this embodiment identifies potential / determined stay areas in sparse trajectories, and extracts the travel behavior of private electric bicycles based on the temporal relationship of the stay areas. Secondly, this embodiment extends the classic random forest model and proposes a spatiotemporal random forest method, which establishes a nonlinear relationship between driving factors and electric bicycle travel characteristics in the spatiotemporal domain. Finally, this embodiment extends the classic SHAP method, mapping the Shapley value to the time axis and space axis to explore the changes in driving factors in time and space.
[0043] The extraction of electric vehicle travel is a key link in the travel behavior analysis framework. Since private electric bicycles are only tracked near monitoring stations, the trajectories of private electric bicycles are relatively sparse. In addition, unlike the trajectories of shared electric bicycles, the trajectories of private electric bicycles do not have order data to assist in the extraction of travel behaviors. In view of the above-mentioned differences, this embodiment proposes a travel behavior extraction method based on the trajectories of private electric bicycles. Specifically, this embodiment first identifies the stay areas in the trajectory data, and then generates the travel behavior of private electric bicycles based on the temporal relationship between these stay areas.
[0044] like Figure 3 As shown, the stay areas in the trajectory data are mainly divided into two categories, one is the determined stay area, and the other is the potential stay area. When the electric vehicle is parked near the monitoring pile, the real-time trajectory points of the electric vehicle will be uploaded to the database at a certain frequency, and the trajectory points will form a point cluster near the monitoring pile, that is, the determined stay area. When the electric vehicle is not parked near the monitoring pile, the real-time trajectory points of the electric vehicle will be lost in the database. When the trajectory point is monitored again, the area monitored for the first time may be the area where the electric vehicle is parked, that is, the potential stay area. For the potential stay area, this embodiment only needs to extract the first point or the last point of the trajectory segment. For the determined stay area, this embodiment extracts it through the method of spatiotemporal clustering. When the stay area in the trajectory points is identified, the travel of the electric bicycle can be generated according to the temporal relationship between the stay areas, including the time series relationship from determining the stay area to determining the stay area ( Figure 3 a) From confirmed stay area to potential stay area ( Figure 3 b) From potential stay area to confirmed stay area ( Figure 3 c) From potential stop area to potential stop area ( Figure 3 d).
[0045] The random forest model is one of the traditional machine learning models. Due to its powerful nonlinear fitting ability, it has been successfully applied to modeling the driving mechanism of urban travel. Although the random forest model has achieved satisfactory modeling accuracy, it still has shortcomings. Specifically, the current modeling methods are mostly simple applications of the random forest model, and do not consider the spatiotemporal relationship contained in the travel data. The independent and identically distributed sample assumption of the classic random forest model is difficult to fully model the complex internal mechanisms and mechanisms of action in travel data. In view of this, this embodiment expands the classic random forest model and proposes a spatiotemporal random forest method.
[0046] like Figure 4 As shown in the figure, the spatiotemporal random forest introduces the spatiotemporal neighborhood based on the classic random forest, and uses the internal driving factors and external driving factors in the spatiotemporal neighborhood to infer the future travel frequency. Among them, the internal driving factor refers to the historical travel frequency in the target grid, and the external driving factor refers to the information such as the urban built environment, spatial location and time window in the target grid. The spatiotemporal random forest model further improves the model's ability to capture spatiotemporal relationships by integrating external and internal driving factors, thereby improving the modeling accuracy of urban travel. Fig.12 The 14 driving factors in the spatiotemporal random forest are shown in detail. It should be noted that this embodiment needs to fuse the driving factors of the target grid itself with the driving factors in the neighborhood to ensure a one-to-one correspondence between the driving factors and the travel frequency. Specifically, the fusion method is shown in the following formula.
[0047]
[0048] Where: and They represent the kth driving factors of the target grid i before and after fusion, respectively, where the first driving factor is Fig.12 The fourteenth driving factor is Historical Trip Frequency in Fig.12 HousePrice in Ω i represents the set of spatiotemporal neighbors of the target grid i; Represents the distance between the target grid i and the neighborhood grid j. From the above formula, we can see that for the 5th to 13th driving factors, the fused driving factor is the sum of the driving factors in the neighborhood; for the 14th driving factor, the fused driving factor is the weighted mean of the driving factors in the neighborhood.
[0049] Although this embodiment establishes a nonlinear relationship between driving factors and electric vehicle travel in the spatiotemporal domain, it is not clear how these driving factors affect the output of the model. SHapley Additive exPlanation (SHAP) is a method for explaining the output of machine learning models. It helps users understand why the model makes decisions by measuring the contribution of each feature to the model output. However, the current SHAP method is mainly used to reveal the weight of the influence of driving factors on the model output, and does not reflect the changes in driving factors with different time periods and different spatial locations. Therefore, this embodiment expands the traditional SHAP method, maps the output of the SHAP method to the time axis and space axis, and then reveals the impact of spatiotemporal heterogeneity on driving factors.
[0050] like Figure 5 As shown, this embodiment establishes a three-dimensional cube with the spatial grid index as the S axis, the time window as the T axis, and the driving factor as the F axis. For any driving factor in the three-dimensional cube, this embodiment can calculate the Shapley value of the corresponding driving factor based on the principle of game theory, and assign the Shapley value to the corresponding spatial position and time window through spatial mapping and time mapping. When all driving factors in the three-dimensional cube are mapped, this embodiment can analyze the changes of driving factors with different time periods and different spatial positions. The Shapley value of the driving factor in the three-dimensional cube can be calculated by the following formula.
[0051]
[0052] Where: STRF represents the spatiotemporal random forest model; represents the kth driving factor of target grid i; N represents the total number of time windows × spatial grids, which can be decomposed into N S×N T ; Represents the driving factor in a three-dimensional cube, given by Decompose to obtain; represents the Shapley value of the driving factor of the sth grid in the tth time window; when When it is greater than 0, it means that the driving factor has a positive effect on electric vehicle travel in the sth grid in the tth time window; When it is less than 0, it means that the driving factor has a negative effect on electric vehicle travel in the sth grid in the tth time window.
[0053] Experimental results analysis:
[0054] First, we explore the differences in travel distance and duration between working days and non-working days from a macro perspective. Then, we analyze the differences in travel frequency at different time locations and spatial locations from a micro perspective.
[0055] Figure 6 The statistical distribution of travel distance and trip duration on weekdays and non-workdays is shown. The results show that the travel distance shows an obvious logarithmic log distribution on weekdays and non-workdays, and the travel duration shows an obvious Hill distribution on weekdays and non-workdays. In addition to the statistical distribution, there are slight differences in travel distance and travel duration on weekdays and non-workdays. Compared with electric bicycle trips on weekdays, the travel distance and travel duration of electric bicycle trips on non-workdays are slightly shorter. For example, the proportion of trips less than 2 kilometers increased from 59.7% on weekdays to 60.3% on non-workdays, and the proportion of trips less than 10 minutes increased from 51.2% on weekdays to 52.7% on non-workdays. In addition, compared with the studies on shared electric vehicle travel and shared bicycles in the existing literature, the travel distance and travel duration of private electric vehicle trips are slightly longer. On weekdays, 48.7% of electric vehicle trips have travel distances concentrated between 1 km and 4 km, and 45.8% of electric vehicle trips have travel times concentrated between 5 minutes and 20 minutes. On non-working days, 49.1% of electric vehicle trips are between 1 km and 4 km, and 45.3% of electric vehicle trips are between 5 minutes and 20 minutes. This shows that private electric bicycle travel not only solves the problem of first / last kilometer travel, but also serves the daily commuting needs of the public.
[0056] Figure 7The difference between travel frequency on weekdays and non-workdays is shown. The results show that travel frequency presents a clear bimodal distribution on weekdays, with the morning peak of travel being 7:30-9:00 and the evening peak of travel being 16:00-18:30. Compared with weekdays, there is no obvious morning peak / evening peak in travel frequency on non-workdays. The results presented by travel frequency are consistent with our common sense. In addition, Figure 8 The differences in travel frequency at different time locations and spatial locations are shown at a micro scale. The results show that electric bicycle travel shows obvious heterogeneity in spatial regions and time windows. On a spatial scale, electric bicycle travel hotspots are mainly concentrated in the city center. On a temporal scale, electric bicycle travel hotspots are concentrated in the morning and evening peaks on weekdays, and travel hotspots are concentrated throughout the day on non-working days. Figure 7 The temporal and spatial heterogeneity presented also indirectly proves the necessity of establishing the STRF model.
[0057] like Fig. 9 As shown, this embodiment uses linear regression, XGBoost, and random forest models as benchmark experiments to analyze the advantages of the Spatiotemporal Random Forest model in fitting accuracy. At the same time, the fitting accuracy of the Spatial RandomForest model and the Temporal Random Forest model is also analyzed to reflect the rationality of the model design. The results show that the fitting accuracy of the XGBoost and Random Forest models is significantly better than that of linear regression, which indicates that there is a nonlinear relationship between electric bicycle travel and influencing factors. Secondly, the fitting accuracy of the Spatial Random Forest model and the Temporal Random Forest model is higher than that of the Random Forest model, which shows the rationality of introducing time or spatial relationships into the classic random forest model. Finally, the fitting accuracy of the Spatiotemporal RandomForest model is better than that of the Spatial Random Forest model and the Temporal Random Forest model, which shows the necessity of modeling spatiotemporal relationships at the same time. In addition, this embodiment analyzes the stability of the fitting accuracy of the SpatiotemporalRandom Forest model under different random seeds, and the results are shown in Fig.13 The results show that the Spatiotemporal Random Forest model not only has good prediction accuracy, but also has good stability, which further proves the advantages of the Spatiotemporal Random Forest model.
[0058] Fig.10 (a) shows the influence weight of driving factors on electric bicycle travel. The results show that the number of electric bicycle trips will increase with the increase of historical travel frequency. It is worth noting that the travel frequency of electric bicycles is significantly affected by factors such as commercial POI density, medical POI density, and housing prices. Taking commercial POI density as an example, the travel frequency of electric bicycles will increase with the increase of POI density. In addition to analyzing the influence weight of driving factors on the travel frequency of electric vehicles, the framework proposed in this embodiment can further analyze the temporal heterogeneity of the influence weights. Fig.10 (b) shows the impact of commercial POI density on electric bicycle travel in different time windows. The results show that the impact weight of commercial POI density on electric bicycle travel changes over time and is consistent with Figure 7 The travel frequency of electric bicycles in the study shows a high degree of consistency. This means that the impact of driving factors on electric bicycle travel during peak travel periods is greater than the impact of driving factors on electric bicycle travel during non-peak travel periods.
[0059] In addition to the temporal heterogeneity of the influencing weights, this embodiment can further analyze the differences in the influencing weights at different spatial locations. It is worth noting that the method proposed in this embodiment can reveal phenomena that are difficult to discover with traditional methods. Taking population density as an example, both high-density and low-density areas may promote the travel of electric bicycles, which leads to Fig.11 (a) It is difficult to directly reveal the correlation between population density and electric bicycle travel. Fig.11 (a) Fig.11 (b) and Fig.11 (c) can not only find the spatiotemporal pattern of population density factors, but also clarify the distribution of influence weights (Shapely values) at different spatial locations. In addition, this embodiment also finds that the influence weights are clustered in spatial locations. The results show that although the importance of spatial location in Figure 10 (a) is ranked low, spatial location is extremely important for electric bicycle travel.
[0060] This embodiment establishes a travel behavior analysis method for private electric bicycle trajectories, and verifies the effectiveness of this method by taking the private electric bicycle trajectories in Wuhan as an example. First, this embodiment identifies point clusters (potential / determined stay areas) in the trajectory data, and extracts the travel behavior of private electric bicycles based on the temporal relationship of the point clusters. Secondly, this embodiment proposes a spatiotemporal random forest method to establish a nonlinear relationship between 14 influencing factors and private electric bicycle travel. Finally, this embodiment uses an expanded traditional SHAP model to reveal the spatiotemporal variation effects of driving factors. The method proposed in this embodiment can reveal phenomena that are difficult to discover using traditional methods, and provides a novel data mining method for in-depth understanding of electric vehicle travel patterns and behavioral characteristics.
[0061] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Other modifications or equivalent substitutions made to the technical solution of the present invention by ordinary technicians in the field should be included in the scope of the claims of the present invention as long as they do not depart from the spirit and scope of the technical solution of the present invention.
Claims
1. A travel behavior analysis method for sparse trajectories of private electric bicycles, characterized in that: The following steps are involved: S1, identify point clusters in trajectory data, which include potential stay areas and confirmed stay areas, and extract the travel behavior of private electric bicycles based on the temporal relationship of point clusters; S2, extending the classic random forest to establish a spatiotemporal forest to establish the nonlinear relationship between 14 influencing factors and private electric bicycle travel; S3, expands the classic SHAP method to establish the spatiotemporal SHAP method, maps the Shapley value to the time axis and space axis, and explores the spatiotemporal variation effects of driving factors.
2. The travel behavior analysis method for sparse trajectories of private electric bicycles according to claim 1 is characterized by: The data in S1 includes private electric bicycle trajectory data, traffic network data, point of interest data, Amap congestion data, weather data, housing price data and population data.
3. The travel behavior analysis method for sparse trajectories of private electric bicycles according to claim 1 is characterized by: The stay areas in the trajectory data in S1 are divided into two categories: one is the confirmed stay area, and the other is the potential stay area. When an electric vehicle is parked near a monitoring pile, the real-time trajectory points of the electric vehicle will be uploaded to the database at a certain frequency, and the trajectory points will form a point cluster near the monitoring pile, i.e., the confirmed stay area. When the electric vehicle is not parked near the monitoring post, the real-time trajectory point of the electric vehicle will be lost in the database. When the trajectory point is monitored again, the first monitored area may be the area where the electric vehicle is parked, that is, the potential staying area. For the potential staying area, it is only necessary to extract the first point or the last point of the trajectory segment. For the determined staying area, it is extracted by the spatiotemporal clustering method. When the staying area in the trajectory point is identified, the travel of the electric bicycle can be generated according to the temporal relationship between the staying areas, including from the determined staying area to the determined staying area, from the determined staying area to the potential staying area, from the potential staying area to the determined staying area, and from the potential staying area to the potential staying area.
4. The travel behavior analysis method for sparse trajectories of private electric bicycles according to claim 1 is characterized by: The spatiotemporal random forest method is as follows: the spatiotemporal random forest introduces a spatiotemporal neighborhood based on the classic random forest, and uses internal driving factors and external driving factors in the spatiotemporal neighborhood to jointly infer future travel frequencies, wherein the internal driving factor refers to the historical travel frequency in the target grid, and the external driving factor refers to the urban built environment, spatial location and time window information in the target grid; the spatiotemporal random forest model further improves the model's ability to capture spatiotemporal relationships by integrating external driving factors and internal driving factors, thereby improving the modeling accuracy of urban travel; This method needs to fuse the driving factors of the target grid itself with the driving factors in the neighborhood to ensure a one-to-one correspondence between the driving factors and the travel frequency. The specific fusion method is as follows: Where: and Represents the target grid Before and after fusion driving factors, of which the first driving factor is Historical Trip Frequency and the fourteenth driving factor is House Price; represents the set of spatiotemporal neighbors of target grid i; Represents the distance between the target grid and the neighborhood grid. From the above formula, we can see that for the 5th to 13th driving factors, the fused driving factor is the sum of the driving factors in the neighborhood. For the 14th driving factor, the fused driving factor is the weighted mean of the driving factors in the neighborhood.
5. The travel behavior analysis method for sparse trajectories of private electric bicycles according to claim 1 is characterized by: The S3 is specifically as follows: a three-dimensional cube is established with the spatial grid index as the S axis, the time window as the T axis, and the driving factor as the F axis. For any driving factor in the three-dimensional cube, the Shapley value of the corresponding driving factor is calculated based on the game theory principle, and the Shapley value is assigned to the corresponding spatial position and time window through spatial mapping and time mapping. When all driving factors in the three-dimensional cube are mapped, the changes of the driving factors with different time periods and different spatial positions can be analyzed. The Shapley value of the driving factor in the three-dimensional cube is calculated by the following formula: Where: STRF represents the spatiotemporal random forest model; The target grid driving factors; N represents the total number of time windows × spatial grids, which can be decomposed into ; represents the driving factor in the three-dimensional cube, obtained by decomposition; represents the Shapley value of the driving factor of the s-th grid in the t-th time window; When it is greater than 0, it means that the driving factor has a positive effect on electric vehicle travel in the sth grid in the tth time window; when it is less than 0, it means that the driving factor has a negative effect on electric vehicle travel in the sth grid in the tth time window.
Citation Information
Patent Citations
Spatial auto-regression model based method for predicting non-commuting travel using electric bicycles
CN108053053A
Human travel sparse trajectory data interpolation reconstruction method
CN110399919A
Shared bicycle use influence assessment method fusing spatial heterogeneity and nonlinearity
CN117172574A
