A method, device and storage medium for predicting the location of a shared bicycle

Abstract

The present invention relates to a shared bicycle location prediction method, device, and storage medium. This method, based on migration trends, incorporates the migration trend matrix between stations and their associated clusters as eigenvalues ​​into station location information. It then uses a fuzzy C-means (FCM) clustering algorithm to iteratively cluster shared bicycle datasets, thereby outputting shared bicycle movement prediction results. Compared with existing technologies, this method offers advantages such as high accuracy and close adherence to practical applications.

Description

Technical Field

[0001] The present invention relates to the field of urban transportation, and in particular to a method, device and storage medium for predicting the position of a shared bicycle. Background Art

[0002] Urban traffic flow forecasting is the process of predicting the number of passengers entering (inflow) or leaving (outflow) each area of ​​a city. Effectively forecasting urban traffic flow is critical to many business applications, such as mobile marketing, route optimization, preparing for a major shift in transportation, or slowing the spread of epidemics such as COVID-19. Take mobile marketing as an example; mobile technology allows businesses to obtain real-time location information of passengers and provide personalized promotional information tailored to the passenger's location and time, such as sending coupons to potential customers at a nearby theater. Understanding and predicting how passengers move around a city, especially leveraging regularities and contextual patterns, helps take the right actions at the right place and time, which is crucial to the success of mobile marketing, including higher redemption probabilities, faster redemption behavior, and higher transaction amounts. Many methods have been proposed to predict urban traffic flow, ranging from time series methods and deep learning methods to deep learning methods designed specifically for urban traffic forecasting.

[0003] However, existing technologies fail to account for the hierarchical nature of human mobility in cities. Each person is unique, and their lives transcend the boundaries of their community. In a recent Nature paper, Alessandretti et al. observed that human mobility (human movement) exhibits a hierarchical pattern. In the process of cross-regional movement, humans migrate in a characteristic hierarchy of neighborhoods, regions, cities, and countries. Migration trends between different hierarchies guide future migration between regions, and migration forms can be divided into intra-layer (local) migration and inter-layer (global) migration.

[0004] Based on the hierarchical structure of human mobility, this hierarchy can be combined to predict urban traffic flow. First, a hierarchical iterative fuzzy C-means clustering algorithm is proposed to group locations, taking into account the migration trends between adjacent locations and distant locations, but with significant correlation. For example, if only spatial proximity is considered, a restaurant may be similar to a nearby residential area. However, if information is aggregated at a higher level (for example, clustering locations that customers usually visit after dining), one may observe that restaurants are highly similar to bars, which may be farther away than residential areas. This correlation exists because people usually go to bars for drinks after dinner. Even if bars and restaurants are far apart, they are still closely connected.

[0005] like Figure 2As shown in Figure 2, the mobility patterns are divided into three levels: the lowest level (spatial proximity), the middle level (road network), and the highest level (personnel mobility flow). People tend to visit nearby locations (e.g., buying food from a shopping mall within walking distance), and these locations (e.g., residential areas and shopping malls) can be clustered based on their proximity information. Figure 2 We can use the P loop in to show such clusters (the three clusters are A, B, and C). There are also two independent points (D and E) that are far away from each other and do not belong to any cluster. Most commonly used clustering algorithms, such as K-means, can only cluster locations based on their proximity. Next, consider clusters that are connected to each other by major roads in a city. Road networks, especially freeways and highways, facilitate travel between major clusters of locations; therefore, urban traffic flow is more likely to follow such a road network. Using Figure 2 We use the Q lines in the graph to show the highways or freeways connecting the clusters we obtained in the first step. Next, we group them together using Q loops to show higher-level clusters. Finally, we need to explore hidden human mobility patterns. As in the example shown in the previous paragraph, in terms of urban traffic flow, a group of restaurants and a group of bars are often strongly correlated because people often go to drink after dinner, even if they are far away and not connected to the city's main roads, such as Figure 2 As shown, the R line shows this high-level hidden human movement flow pattern.

[0006] Analyzing urban traffic flow exhibits a hierarchical structure, not just spatially and temporally, but also at a higher level, focusing on migration trends. However, existing technologies for predicting shared bike locations fail to consider this migration trend, resulting in limitations in the analysis used. Summary of the Invention

[0007] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a highly accurate and practical shared bicycle location prediction method, device and storage medium.

[0008] The purpose of the present invention can be achieved by the following technical solutions:

[0009] According to a first aspect of the present invention, a method, device, and storage medium for predicting the location of a shared bicycle are provided. The method incorporates a migration trend matrix between a station and its associated clusters as an eigenvalue into the station location information based on migration trends, and uses a fuzzy C-means clustering (FCM) algorithm to iteratively cluster a shared bicycle dataset, thereby outputting a shared bicycle movement prediction result.

[0010] The method specifically comprises the following steps:

[0011] Step S1: Obtain the shared bicycle dataset and perform data preprocessing;

[0012] Step S2: Perform fuzzy C-means clustering (FCM) based on the shared bicycle stations to obtain relevant clusters;

[0013] Step S3: Generate a dual-channel matrix between clusters every hour based on the inflow and outflow data;

[0014] Step S4: Fusing the dual-channel matrix into the spatiotemporal residual network model to predict the inter-regional shared bicycle movement results. If the prediction error is greater than the preset threshold, go to step S5; otherwise, go to step S6.

[0015] Step S5: Based on the relevant clusters obtained in step S2, obtain the migration trend matrix between the current station and the relevant clusters, and add the reduced-dimensional matrix as the eigenvalue to the geographic location information of the station. Then go to step S2 to perform fuzzy C-means clustering iteration until the result is stable and the iteration ends.

[0016] Step S6: Output the shared bicycle movement prediction result.

[0017] Preferably, the shared bicycle dataset in step S1 includes trip interval time, trip start time, trip end time, trip start geographical location and trip end geographical location information.

[0018] Preferably, the initial fuzzy C-means clustering FCM in step S2 includes the following sub-steps:

[0019] Step S21: Randomly initialize weights;

[0020] Step S22: Calculating the centroid in the fuzzy C-means clustering FCM algorithm based on the weighted average of the membership degrees;

[0021] Step S23: Update the fuzzy pseudo partitions, including weights and memberships.

[0022] Preferably, the membership is a probability function, which represents the strength of each data sample to each cluster; the membership update satisfies: the closer the sample point is to the centroid, the higher the membership.

[0023] Preferably, in the fuzzy C-means clustering algorithm in step S2, each data sample belongs to and only belongs to one cluster.

[0024] Preferably, the inflow and outflow dual-channel matrix in step S3 refers to the image RGB channel model and is used as the input of the spatiotemporal residual network to predict the movement ratio of shared bicycles.

[0025] Preferably, the prediction error in step S4 is RMSE root mean square error.

[0026] Preferably, the dimension reduction process in step S5 uses the Fibonacci paradigm to reduce the dimension of the migration trend matrix.

[0027] According to a second aspect of the present invention, an electronic device is provided, comprising a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the above method when executing the program.

[0028] According to a third aspect of the present invention, a computer-readable storage medium stores a computer program, which implements the above method when executed by a processor.

[0029] Compared with the prior art, the present invention has the following advantages:

[0030] The traditional prediction model is based on a grid generation matrix, and the similarity of neighboring areas is relatively large. However, after trend clustering, the present invention still has similarities in farther areas. This similarity is trend similarity, and the prediction accuracy is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 This is a flow chart of the shared bicycle spatiotemporal residual prediction method of the present invention;

[0032] Figure 2 Analyze the intention of the present invention level;

[0033] Figure 3 This is a schematic diagram of the technical framework of the present invention;

[0034] Figure 4 Schematic diagram of the hierarchical iterative fuzzy C-means clustering algorithm of the present invention;

[0035] Figure 5 Schematic diagram of the dual-channel model fusion spatiotemporal residual network model of the present invention. DETAILED DESCRIPTION

[0036] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0037] This study, based on observations, reveals a hierarchical structure in analyzing urban traffic flow. This hierarchy encompasses not only spatial and temporal aspects but also higher-level migration trends. This method has broad commercial application value. Using shared bicycles as an example, this study can be extended to human mobility patterns. Using this method, we can capture interregional mobility patterns and accurately predict how people move around a city. This can help marketers make the right decisions, taking the right actions at the right place and time, which is crucial to successful mobile marketing.

[0038] like Figure 1 As shown in the figure, the present invention proposes a shared bicycle location prediction method. This method is based on the migration trend, incorporates the migration trend matrix between the station and the related cluster as the eigenvalue into the station location information, and uses the fuzzy C-means clustering FCM algorithm to iteratively cluster the shared bicycle data set to output the shared bicycle movement prediction result. Figure 3 A schematic diagram of the technical framework of the present invention is given. The method specifically includes the following steps:

[0039] Step S1: Obtain a shared bicycle dataset, including trip interval time, trip start time, trip end time, trip start location, and trip end location information, and perform data preprocessing;

[0040] Step S2: Perform fuzzy C-means clustering (FCM) based on the shared bicycle stations to obtain relevant clusters, specifically:

[0041] The initial fuzzy C-means clustering FCM in step S2 includes the following sub-steps:

[0042] Step S21: Randomly initialize weights;

[0043] Step S22: Calculating the centroid in the fuzzy C-means clustering FCM algorithm based on the weighted average of the membership degrees;

[0044] Step S23: Update the fuzzy pseudo partition, including weights and membership, where the membership is a probability function that represents the strength of each data sample to each cluster. The update condition satisfies that the closer the sample point is to the centroid, the higher the membership;

[0045] In the fuzzy C-means clustering algorithm, each data sample belongs to and only belongs to one cluster.

[0046] Step S3: Generate a two-channel matrix between clusters every hour based on the inflow and outflow data. This matrix refers to the image RGB channel model and is used as the input of the spatiotemporal residual network to predict the shared bicycle movement ratio.

[0047] Step S4: Fusing the dual-channel matrix into the spatiotemporal residual network model to predict the inter-regional shared bicycle movement results. If the prediction error is greater than the preset threshold, go to step S5; otherwise, go to step S6.

[0048] Step S5: Based on the relevant clusters obtained in step S2, obtain the migration trend matrix between the current station and the relevant clusters, and add the migration trend matrix after Fibonacci normal form dimension reduction as the eigenvalue to the geographical location information of the station. Then go to step S2 to perform fuzzy C-means clustering iteration until the result is stable and the iteration ends;

[0049] Step S6: Output the shared bicycle movement prediction result.

[0050] The method is specifically:

[0051] The dataset is a shared bicycle trip record, which contains information such as the trip interval, trip start and end time, and the trip start and end geographic locations (latitude and longitude). The verification dataset of this invention is a New York shared bicycle dataset. Therefore, it is different from the domestic shared bicycle system. The start and end locations are fixed shared bicycle stations, and the number of stations is limited.

[0052] The fuzzy C-means clustering algorithm can not only cluster geographic locations (latitude and longitude), but also add additional feature dimensions for clustering.

[0053] The fuzzy C-means clustering algorithm is different from other clustering algorithms such as AP clustering, Gaussian mixture model clustering, K-means clustering, and spectral clustering in that each data sample belongs to one and only one cluster. The membership function in fuzzy C-means provides probabilities, which essentially represent the strength of each data sample for each cluster.

[0054] Since the clustering algorithm can only process low-dimensional arrays, the clustering feature dimension is increased according to step 5. Here, the Fibonacci paradigm is used to reduce the dimension of the migration trend matrix to obtain new clusters, and iterate in sequence until the stable iteration terminates;

[0055] The present invention refers to the image RGB channel model and establishes an inflow and outflow dual-channel model as the input of the spatiotemporal residual network to predict the movement ratio of shared bicycles;

[0056] The step S2 is specifically as follows:

[0057] Step S21: Random initialization, i.e., random selection of weights. The number of clusters needs to be set in the first step.

[0058] Step S22: Calculate the centroid in FCM, which is based on the weighted average of the membership degrees;

[0059] Step S23: Update the fuzzy pseudo partition (update weights or membership). The closer the sample point is to the centroid, the higher the membership, and vice versa.

[0060] For the detailed settings of the hierarchical iterative clustering algorithm, such as Figure 4 As shown. Given a position X={x1,x2,...,x m In the first step, fuzzy C-means clustering is based entirely on the geographical information of each location:

[0061]

[0062] Among them, (x i.lat , x i.lon ) are the longitude and latitude information of the i-th position respectively;

[0063] After the first iteration, fuzzy C-means clusters locations based on their geographical information and migration trends:

[0064]

[0065] in is the Fibonacci paradigm of the migration trend matrix at time t;

[0066] Iterate in sequence until the stable iteration terminates. The stable iteration judgment condition is: when the prediction error is less than the preset threshold; the prediction error is the RMSE root mean square error. When the root mean square error fluctuation is stable at 0.01, the iteration is stable and the algorithm terminates.

[0067] The specific process of steps S3 and S4: generating a dual-channel matrix between clusters every hour based on inflow and outflow, fusing the obtained dual-channel matrix with the spatiotemporal residual network model, and predicting the movement of shared bicycles between regions is as follows:

[0068] After obtaining the output matrix from step S2, a fusion model is used to convert the inflow and outflow across clusters into a two-channel image-like matrix as the input of the spatiotemporal residual network. Figure 5 As shown, a 2-channel matrix is ​​used to store the inflow and outflow of migration between clusters, just like the RGB channels of an image (RGB images have three channels: R, G, B). Next, an hourly two-channel matrix is ​​obtained, which contains the hourly urban traffic flow between each cluster.

[0069] After implementing the fusion model, the hourly urban traffic dual-channel matrix is ​​used as input to predict urban traffic using a spatiotemporal residual network. The spatiotemporal residual network consists of three main components: trend, cycle, and proximity. These three components share the same network architecture: a CNN followed by a series of residual units. This structure captures the spatial dependencies between adjacent and non-adjacent regions.

[0070] Step S5: Based on the clusters obtained in step S2, a migration trend matrix between the current station and the cluster is obtained, the matrix is ​​reduced in dimension using a paradigm, and the matrix is ​​added as an eigenvalue to the geographic location information of the station to perform iterative fuzzy C-means clustering. Steps S2, S3, and S4 are repeated until the result is stable and the iteration is completed. The specific process is as follows:

[0071] The Fibonacci pattern is as follows:

[0072]

[0073] Among them, A ij is the migration trend matrix, ||A|| F Fibonacci pattern of the migration trend matrix.

[0074] The validation dataset for this invention is a New York shared bike dataset, which starts and ends at fixed shared bike stations with a limited number of stations. The following table shows the comparison results between the method of this invention and the existing technology.

[0075] Table 1

[0076]

[0077] It can be found from the table that compared with the existing algorithm, the root mean square error RMSE of the prediction of the present invention is 4.43, and the prediction accuracy is greatly improved.

[0078] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the described module can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0079] The electronic device of the present invention includes a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) or loaded from a storage unit into a random access memory (RAM). In the RAM, various programs and data required for device operation can also be stored. The CPU, ROM, and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0080] Many components in a device are connected to the I / O interface, including: input units, such as a keyboard and mouse; output units, such as various types of displays and speakers; storage units, such as magnetic disks and optical disks; and communication units, such as network cards, modems, and wireless communication transceivers. The communication unit allows the device to exchange information / data with other devices via computer networks such as the Internet and / or various telecommunication networks.

[0081] The processing unit performs the various methods and processes described above, such as methods S1 to S6. For example, in some embodiments, methods S1 to S6 may be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed onto the device via a ROM and / or a communication unit. When the computer program is loaded into the RAM and executed by the CPU, one or more steps of methods S1 to S6 described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute methods S1 to S6 by any other appropriate means (e.g., by means of firmware).

[0082] The functions described above herein may be performed, at least in part, by one or more hardware logic components. For example, and without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), and the like.

[0083] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. Such program code can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0084] In the context of the present invention, machine-readable medium can be a tangible medium that can contain or store a program for use with an instruction execution system, device or equipment or used in combination with an instruction execution system, device or equipment. Machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. More specific examples of machine-readable storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0085] 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 such modifications or substitutions are intended to be within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A method for predicting the location of a shared bicycle, characterized in that: Based on the migration trend, this method incorporates the migration trend matrix between stations and related clusters as the eigenvalue into the station location information, and uses the fuzzy C-means clustering (FCM) algorithm to iteratively cluster the shared bicycle dataset to output the shared bicycle movement prediction results. The method specifically comprises the following steps: Step S1: Obtain the shared bicycle dataset and perform data preprocessing; Step S2: Perform fuzzy C-means clustering (FCM) based on the shared bicycle stations to obtain relevant clusters; Step S3: Generate a dual-channel matrix between clusters every hour based on the inflow and outflow data; Step S4: Fusing the dual-channel matrix into the spatiotemporal residual network model to predict the inter-regional shared bicycle movement results. If the prediction error is greater than the preset threshold, go to step S5; otherwise, go to step S6. Step S5: Based on the relevant clusters obtained in step S2, obtain the migration trend matrix between the current station and the relevant clusters, and add the reduced-dimensional matrix as the eigenvalue to the geographic location information of the station. Then go to step S2 to perform fuzzy C-means clustering iteration until the result is stable and the iteration ends. Step S6: Output the shared bicycle movement prediction result; The initial fuzzy C-means clustering FCM in step S2 includes the following sub-steps: Step S21: Randomly initialize weights; Step S22: Calculating the centroid in the fuzzy C-means clustering FCM algorithm based on the weighted average of the membership degrees; Step S23: Update the fuzzy pseudo partition, including weight and membership; The membership is a probability function that represents the strength of each data sample to each cluster; the membership update satisfies: the closer the sample point is to the centroid, the higher the membership; The dimension reduction process in step S5 uses the Fibonacci paradigm to reduce the dimension of the migration trend matrix. The Fibonacci paradigm is as follows: Among them, A ij is the migration trend matrix, ||A|| F Fibonacci pattern of the migration trend matrix; The inflow and outflow dual-channel matrix in step S3 refers to the image RGB channel model and is used as the input of the spatiotemporal residual network to predict the movement ratio of shared bicycles.

2. A shared bicycle location prediction method according to claim 1, characterized in that: The shared bicycle dataset in step S1 includes trip interval time, trip start time, trip end time, trip start location and trip end location information.

3. A shared bicycle location prediction method according to claim 1, characterized in that: In the fuzzy C-means clustering algorithm in step S2, each data sample belongs to and only belongs to one cluster.

4. A shared bicycle location prediction method according to claim 1, characterized in that: The prediction error in step S4 is the RMSE root mean square error.

5. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the program, the method according to claims 1 to 4 is implemented.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images