An Internet rental bicycle user preference type prediction method, system, device and storage medium based on order data

Abstract

The present invention relates to the technical field of user behavior analysis of Internet rental bicycles, and specifically discloses a method, system, device and storage medium for predicting user preference types of Internet rental bicycles based on order data. The prediction method includes the following steps: constructing a user preference type prediction database, data preprocessing, identifying users with shared rides, classifying user preference types, identifying users' common locations, extracting influencing factors of user preference types, constructing and predicting a user preference type prediction model. Through in-depth mining of actual shared bicycle and shared electric bicycle order data, the present invention comprehensively considers the influencing factors in two aspects of built environment characteristics and shared infrastructure attributes, simplifies the identification and analysis process, makes the results accurate and reliable, can truly reflect users' travel preferences, and finally constructs a user preference classification prediction model.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet rental user behavior analysis, and in particular to a method, system, device and storage medium for predicting user preference types of Internet rental bicycles based on order data. Background Art

[0002] With the rapid development of the sharing economy, Internet rental bicycles, as a green and environmentally friendly way of travel, provide convenient rental services, enabling users to flexibly travel short distances in the city and meet their daily travel needs without owning a vehicle. Users can easily find nearby shared vehicles through smartphone applications and quickly complete the rental and return process, which greatly improves the convenience and efficiency of urban travel and effectively alleviates urban traffic pressure. However, with the increasing number of shared bicycles and shared electric motorcycles, users show obvious preference differences when choosing shared travel tools. How to accurately predict the user's preference type and then optimize the deployment and management of shared bicycles and shared electric motorcycles has become an important issue in current transportation planning and market operations. Summary of the invention

[0003] In order to solve the problems existing in the prior art, the present invention provides a method, system, device and storage medium for predicting the user preference type of Internet rental bicycles based on order data. The method uses the order data, built environment data and electronic fence data of shared bicycles and shared electric motorcycles to construct a classification prediction model, comprehensively considers the influencing factors of the built environment characteristics and shared infrastructure, explores the behavioral characteristics of users when using shared travel tools, and analyzes and predicts the user's preference type, thereby solving the problems mentioned in the above background technology.

[0004] To achieve the above object, the present invention provides the following technical solution: a method for predicting user preference types of Internet rental bicycles based on order data, comprising the following steps:

[0005] S1. Build a user preference type prediction database, including the order data of the two main vehicle types of Internet rental bicycles, namely shared bicycles and shared electric motorcycles, as well as built environment data and electronic fence data;

[0006] S2. Data preprocessing: Clean the shared bicycle and shared electric motorcycle data in the database of step S1 to remove invalid and abnormal data;

[0007] S3. Identification of users with shared bike rides: Based on the user ID, identify the travel records of all shared bikes and shared e-bikes under the same ID number. Calculate the number of times and days each user uses shared bikes and shared e-bikes respectively according to the vehicle type and borrowing time, and statistically analyze the frequency distribution of users' use of shared bikes and shared e-bikes;

[0008] S4. Classification of user preference types: According to the frequency distribution of users' use of shared bikes and shared e-bikes, classify users into three types: shared bike-dominated, balanced, and shared e-bike-dominated;

[0009] S5. Identification of users' common locations: Use the HDBSCAN clustering algorithm to identify users' common locations, that is, the common borrowing and returning locations of users;

[0010] S6. Extraction of influencing factors for user preference types: Based on the users' common locations identified in step S5, establish a buffer zone and extract the surrounding built environment characteristics, including the number of various types of POIs in the city, slope, road network density, distances from the common location to the nearest public transportation and the nearest CBD, and the attributes of shared infrastructure;

[0011] S7. Construction and prediction of user preference type prediction models: Construct a multinomial Logit (MNL) model as a user preference type prediction model and train it. Input the data to be measured and use the trained model to predict the user's preference type and output the prediction results.

[0012] Preferably, the built environment data in step S1 includes: digital elevation model (DEM) data, road network data, public transportation data, and facility network data; among them, the facility network data includes various types of points of interest (POIs) such as shopping, leisure and entertainment, and commercial services, including information on geographical locations and facility types, and is processed through ArcGIS; the electronic fence data includes: fence name, central point address, area, and longitude and latitude coordinates of the central point and regional vertices.

[0013] Preferably, in step S2, it specifically includes the following:

[0014] S21. Eliminate the records with incomplete order data of shared bikes and shared e-bikes; among them, a complete order data includes information such as user ID, vehicle ID, vehicle type, borrowing date, borrowing time, borrowing longitude, borrowing latitude, returning date, returning time, returning longitude, and returning latitude;

[0015] S22. Eliminate the data with the geographical locations of borrowing and parking of shared bikes and shared e-bikes deviating from the predicted range;

[0016] S23. Exclude records with a one-way travel duration less than 1 minute or greater than 2 hours. Here, the travel duration is obtained by subtracting the pick-up time from the return time in the same order.

[0017] S24. Exclude records with a one-way travel distance less than 100 meters. Here, the travel distance is obtained by calculating the pick-up and return longitude and latitude in the same order.

[0018] Preferably, in step S4, it specifically includes the following:

[0019] Calculate the proportion of the user's shared bicycle travel times. The proportion of shared bicycle travel times is obtained by dividing the number of shared bicycle travel times by the total number of travel times. According to the proportion of shared bicycle travel times, users are equally divided into three types. The specific division includes:

[0020] Users with a relatively high proportion of shared bicycle usage, that is, users with a proportion of shared bicycle travel times ≥ 2 / 3, are classified as "shared bicycle dominant type";

[0021] Users with a relatively balanced proportion of shared bicycle and shared electric bicycle usage, that is, users with 1 / 3 < proportion of shared bicycle travel times < 2 / 3, are classified as "balanced type";

[0022] Users with a relatively low proportion of shared bicycle usage, that is, users with a proportion of shared bicycle travel times ≤ 1 / 3, are classified as "shared electric bicycle dominant type".

[0023] Preferably, in step S5, the method of using HDBSCAN clustering to identify the user's frequently visited places specifically includes:

[0024] S51. Input data set: Input the pick-up and return longitude and latitude of all shared bicycle users.

[0025] S52. Coordinate system conversion: Convert the longitude and latitude of the user's pick-up and return from the geographic coordinate system to the projected coordinate system.

[0026] S53. Group by user ID and divide the user's travel records into different clusters.

[0027] S54. Set parameters: Define the corresponding minimum number of sample points (min_cluster_size) required to form a cluster according to the user travel times interval.

[0028] S55. Calculate the core distance and density estimation: Calculate the core distance of each pick-up and return point of the user, and define the core distance c k (x) = d(x, N k (x)), which is the distance from the current point x to the k-th nearest point N k (x), and the mutual reachable distance d k (a, b) expression is:

[0029] d k (a, b) = max{c k (a), c k (b), d(a, b)}

[0030] Among them, d(a, b) is the original distance between two points a and b;

[0031] S56. Construct a density adjacency graph: Based on the mutual reachability distance, construct a weighted adjacency graph, where the weight of the edge is the mutual reachability distance between point pairs;

[0032] S57. Generate a minimum spanning tree: The Prim algorithm is used internally in the algorithm. Using the sample points as the original points, construct a minimum spanning tree with the mutual reachability distance between other points and the original points as the weights;

[0033] S58. Establish a cluster hierarchy: The algorithm traverses and reorders the edges of the minimum spanning tree with the mutual reachability distance as the weight, and classifies each edge into a new cluster;

[0034] S59. Extract stable clusters: According to the persistence analysis and the min_cluster_size parameter, extract the most stable clusters as the final result;

[0035] S510. Determine the common locations: The clustering result may obtain more than one cluster. Each cluster represents a location where users often borrow and return bicycles. Select the cluster with the largest number of bicycle borrowing and returning points in the clustering clusters, that is, the most representative cluster, as the common location of the user.

[0036] Preferably, in step S6, it specifically includes the following:

[0037] S61. Establish a buffer: Based on the user's common location obtained in step S5, use ArcGIS software to establish a buffer for extracting influencing factors;

[0038] S62. Extract the quantity of various POIs: Extract the quantities of public service facilities, leisure service facilities, commercial service facilities, residential facilities, and office facilities within the buffer;

[0039] S63. Slope: Extract the average slope within the buffer range around the user's common location;

[0040] S64. Road network density: Extract the road length within the buffer range. The ratio of this road length to the buffer area is the road network density;

[0041] S65. Calculate the nearest distance: Calculate the distances between the common location and the nearest bus stop, subway station, and the nearest CBD;

[0042] S66. Extraction of shared infrastructure attributes: including the vehicle supply in the buffer zone and the number of virtual fences; the extraction of the vehicle supply in the buffer zone includes: randomly selecting a typical working day from the data preprocessed in step S2 as a representative, calculating the distances between the pick-up points and return points of all orders on this day and the user's common locations, screening out the order records located in the buffer zone and numbering them;

[0043] Then divide this day into several time periods, and according to the pick-up time and the numbers, count the pick-up quantities of shared bicycles and shared e-bikes in each buffer zone within each time period;

[0044] For the same buffer zone, the pick-up quantity in the next time period is the available quantity of vehicles in the previous time period; by calculating the average value of the pick-up quantities in all time periods of a day, obtain the average supply quantities of shared bicycles and shared e-bikes.

[0045] Preferably, in step S7, it specifically includes the following:

[0046] S71. Model construction: Taking the user preference types, namely shared bicycle-dominated, balanced, and shared e-bike-dominated, as the dependent variables, and using the built environment characteristics and shared infrastructure attributes in step S6 as the explanatory variables, establish a multinomial Logit (MNL) model;

[0047] S72. Model training: Determine the relationship between the explanatory variables and the user preference types through learning, and use the maximum likelihood estimation method to estimate the parameters of the MNL model;

[0048] S73. Model prediction: According to the estimated model parameters, input new built environment characteristics and shared infrastructure attribute data to predict the user's preference type;

[0049] S74. Output of prediction results: Through the fitted MNL model, calculate the probabilities of the user belonging to each preference type, and select the preference type with the highest probability as the final prediction result.

[0050] On the other hand, to achieve the above object, the present invention also provides the following technical solution: An internet rental bicycle user preference type prediction system based on order data, the system includes the following modules:

[0051] Database construction module, constructing a user preference type prediction database, including order data of two main vehicle types of internet rental bicycles, namely shared bicycles and shared e-bikes, as well as built environment data and virtual fence data;

[0052] Data preprocessing module, cleaning the shared bicycle and shared e-bike data in the database, and removing invalid data and abnormal data;

[0053] The user identification module identifies all the travel records of shared bicycles and shared e-bicycles under the same ID number based on the user ID, calculates the usage times and days of shared bicycles and shared e-bicycles for each user according to the vehicle type and borrowing time, and statistically analyzes the frequency distribution of users' use of shared bicycles and shared e-bicycles;

[0054] The user preference type classification module classifies users into three categories: shared bicycle-dominated type, balanced type, and shared e-bicycle-dominated type according to the frequency distribution of users' use of shared bicycles and shared e-bicycles;

[0055] The user's common location identification module uses the HDBSCAN clustering algorithm to identify the user's common borrowing and returning locations, that is, the locations where the user commonly borrows and returns vehicles;

[0056] The influencing factor extraction module, based on the identified common locations of users, establishes a buffer zone and extracts the surrounding built environment features, including the number of various types of POIs in the city, slope, road network density, the distances from the common location to the nearest public transportation and the nearest CBD, and the attributes of shared infrastructure;

[0057] The prediction model construction and prediction module constructs a multinomial Logit (MNL) model as the user preference type prediction model and trains it. Input the data to be measured and use the trained model to predict the user's preference type and output the prediction result.

[0058] On the other hand, to achieve the above object, the present invention also provides the following technical solution: an electronic device, the electronic device includes: a processor; and a memory for storing one or more programs;

[0059] When the one or more programs are executed by the processor, the processor is caused to execute the above-mentioned method for predicting the user preference type of Internet rental bicycles based on order data.

[0060] On the other hand, to achieve the above object, the present invention also provides the following technical solution: a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned method for predicting the user preference type of Internet rental bicycles based on order data is implemented.

[0061] The beneficial effects of the present invention are:

[0062] 1) By constructing a complete user preference type prediction database, the present invention comprehensively considers multi-dimensional data sources, ensuring the accuracy and comprehensiveness of the prediction;

[0063] 2) The present invention uses the HDBSCAN clustering algorithm to identify the common borrowing and returning locations of users and analyzes the influencing factors of user preference types based on this;

[0064] 3) The present invention constructs an MNL model to analyze and predict the preference types of users. This model fully considers the relationship between user behavior characteristics and environmental factors, realizes accurate prediction and effective management of user behavior, and ensures a high degree of fit between the delivery strategy and user needs. Description of the Drawings

[0065] Figure 1 It is a schematic flow chart of the steps of the method for predicting the preference types of users of Internet rental bicycles based on order data according to the present invention;

[0066] Figure 2 It is a schematic flow chart of the HDBSCAN clustering process;

[0067] Figure 3 It is a schematic diagram of the modules of the system for predicting the preference types of users of Internet rental bicycles based on order data according to the present invention;

[0068] Figure 4 It is a schematic diagram of the structure of an electronic device in an embodiment of the present invention;

[0069] In the figure, 110 - database construction module; 120 - data preprocessing module; 130 - user identification module; 140 - user preference type division module; 150 - user common location identification module; 160 - influencing factor extraction module; 170 - prediction model construction and prediction module; 210 - processor; 220 - storage. Detailed Embodiments

[0070] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. 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.

[0071] Please refer to Figure 1 , the present invention provides a technical solution: a method for predicting the preference types of users of Internet rental bicycles based on order data, including the following steps:

[0072] S1. Construct a database for predicting user preference types, including order data of two main vehicle types of Internet rental bicycles, namely shared bicycles and shared electric bicycles, as well as built environment data and electronic fence data.

[0073] The built environment data includes: digital elevation model DEM data, road network data, public transportation data and facility network data; the facility network data includes multiple categories of points of interest POIs such as shopping, leisure and entertainment, and commercial services, which contain information on geographical location and facility types, and are processed through ArcGIS; the electronic fence data includes: fence name, center point address, area, and latitude and longitude coordinates of the center point and area vertices.

[0074] This embodiment operates the present invention in combination with the actual shared bicycle and shared electric motorcycle order data in Kunming. Using the shared bicycle and shared electric motorcycle order data of Kunming for 31 days from April 20 to May 20, 2022, a total of 3.051 million 1,459 items were collected, including 2.701 million 1,280 shared electric motorcycle order data. Covering five main urban areas including Guandu District, Wuhua District, Panlong District, Xishan District and Chenggong District, these areas gather a variety of land use functions such as residential, work areas, schools, large commercial areas, and transportation hubs. The number of shared riding users is large, indicating that the data is representative. Up to now, the area has 125,000 shared bicycles and shared electric motorcycles, and 15,000 non-motor vehicle parking spaces. These measures make shared riding a convenient, environmentally friendly and healthy travel option, which helps to reduce traffic congestion and air pollution, while promoting the sustainable development of the city.

[0075] The shared bike user preference type prediction database includes Kunming’s shared bike and shared electric bike order data, built environment data, and electronic fence data. The built environment data mainly includes road network data, public transportation data, and various types of facility network data. The facility network data comes from the shopping, leisure and entertainment, commercial services, and other multi-category points of interest (POI) of the Amap open map platform, including geographic location, facility type, and other information, and is processed through ArcGIS. The road network data comes from the OSM database, and the road network is corrected using ArcGIS software combined with the prediction range. Public transportation data includes bus stop information. The specific content of the electronic fence data is shown in Table 1. Specifically, the electronic fence data includes: fence name, center point address, area, and latitude and longitude coordinates of the center point and area vertices.

[0076] Table 1 Electronic fence information (partial)

[0077]

[0078]

[0079] S2. Data preprocessing: Clean the shared bicycle and shared electric motorcycle data in the database of step S1 to remove invalid and abnormal data.

[0080] The details include:

[0081] S21. Eliminate the records with incomplete order data of shared bicycles and shared e-bicycles. Among them, a complete order data includes information such as user ID, vehicle ID, vehicle type, borrowing date, borrowing time, borrowing longitude, borrowing latitude, returning date, returning time, returning longitude, and returning latitude.

[0082] S22. Eliminate the data of the geographical locations of borrowing and parking of shared bicycles and shared e-bicycles that deviate from the predicted range.

[0083] S23. Eliminate the records with a one-way travel duration less than 1 minute and greater than 2 hours. Among them, the travel duration is obtained by subtracting the borrowing time from the returning time in the same order.

[0084] S24. Eliminate the records with a one-way travel distance less than 100 meters. Among them, the travel distance is obtained by calculating the borrowing longitude and latitude and the returning longitude and latitude in the same order.

[0085] After cleaning the order data of shared bicycles and shared e-bicycles, there are 2,303,289 remaining, accounting for about 75.48% of the total data. Table 2 shows some examples of the travel records of shared bicycles and shared e-bicycles.

[0086] Table 2 Travel Data Record Table of Shared Bicycles and Shared E-bicycles (Partial)

[0087]

[0088] Table 2 Travel Data Record Table of Shared Bicycles and Shared E-bicycles (Partial) (Continued)

[0089]

[0090]

[0091] S3. Identification of the same users for shared rides: Based on the user ID, identify all the travel records of shared bicycles and shared e-bicycles under the same ID number, and calculate the number of times and days that each user uses shared bicycles and shared e-bicycles respectively according to the vehicle type and borrowing time. In order to ensure that the users of shared bicycles and shared e-bicycles are frequent and long-term user groups, delete the travel data of users with fewer travel times and travel days. Statistically analyze the frequency distribution of users' use of shared bicycles and shared e-bicycles.

[0092] In this embodiment, in order to ensure that the users of shared bicycles and shared e-bicycles are frequent and long-term user groups, delete the travel data of users with a total number of rides less than 12 times and a total number of days less than 3 days. After screening according to the above screening conditions, 8,609 same users are identified, with a total of 175,160 travel records, among which there are 114,626 travel records of shared e-bicycles, accounting for 67.22%.

[0093] S4. Classification of user preference types: According to the frequency distribution of users' use of shared bicycles and shared electric bicycles, users are divided into three categories: shared bicycle-dominated, balanced, and shared electric bicycle-dominated.

[0094] Specifically, it includes the following:

[0095] Calculate the proportion of the number of shared bicycle trips of users. The proportion of shared bicycle trips is obtained by dividing the number of shared bicycle trips by the total number of trips; according to the proportion of shared bicycle trips, users are equally divided into three types. The specific classification includes:

[0096] Users with a relatively high proportion of shared bicycle use, that is, users with a proportion of shared bicycle trips ≥ 2 / 3, are classified as "shared bicycle-dominated";

[0097] Users with a relatively balanced proportion of shared bicycle and shared electric bicycle use, that is, users with 1 / 3 < proportion of shared bicycle trips < 2 / 3, are classified as "balanced";

[0098] Users with a relatively low proportion of shared bicycle use, that is, users with a proportion of shared bicycle trips ≤ 1 / 3, are classified as "shared electric bicycle-dominated".

[0099] In this embodiment, users with a proportion of trips in the range of (0.66 - 1.0) are recorded as shared bicycle-dominated, with a total of 2052, accounting for 23.84% of the total users; users with a proportion in the range of (0.33 - 0.66) are recorded as balanced, with a total of 1449, accounting for 16.83% of the total users; users with a proportion in the range of (0 - 0.33) are recorded as shared electric bicycle-dominated, with a total of 5108, accounting for 59.33% of the total users.

[0100] S5. Identification of users' frequently used places: The HDBSCAN clustering algorithm is used to identify users' frequently used places, that is, the frequently used borrowing and returning locations of users.

[0101] The method of using HDBSCAN clustering to identify users' frequently used places is as Figure 2 shown, and specifically includes:

[0102] S51. Input data set: Input the longitude and latitude of the borrowing and returning locations of all shared cycling users;

[0103] S52. Coordinate system conversion: Convert the longitude and latitude of users' borrowing and returning locations from the geographic coordinate system to the projected coordinate system;

[0104] S53. Group by user ID and divide the travel records of users into different clusters;

[0105] S54. Set parameters: Define the corresponding minimum number of sample points (min_cluster_size) required to form clusters according to the user travel time interval;

[0106] S55. Calculate the core distance and density estimation: Calculate the core distance of each pick-up and drop-off point of the user, and define the core distance c k (x) = d(x, N k (x)), which is the distance from the current point x to the k-th nearest point N k (x), and the mutual reachability distance d k between two points (a, b) is expressed as:

[0107] d k (a, b) = max{c k (a), c k (b), d(a, b)}

[0108] where d(a, b) is the original distance between points a and b;

[0109] S56. Construct a density adjacency graph: Based on the mutual reachability distance, construct a weighted adjacency graph, where the weight of the edge is the mutual reachability distance between point pairs;

[0110] S57. Generate a minimum spanning tree: The Prim algorithm is used internally in the algorithm. Using the sample points as the original points, construct a minimum spanning tree with the mutual reachability distances between other points and the original points as weights;

[0111] S58. Establish a cluster hierarchy: Using the mutual reachability distance as the weight, traverse and reorder the edges of the minimum spanning tree, and classify each edge into a new cluster;

[0112] S59. Extract stable clusters: According to the persistence analysis and the min_cluster_size parameter, extract the most stable clusters as the final result;

[0113] S510. Determine the common locations: The clustering result may obtain more than one cluster. Each cluster represents a location where the user often picks up and drops off the vehicle. Select the cluster with the largest number of pick-up and drop-off points in the clustering clusters, that is, the most representative cluster, as the common location of the user.

[0114] Specifically, we divide the travel frequency interval into 6 intervals: less than 24 times, 24 to 48 times, 48 to 96 times, 96 to 192 times, 192 to 384 times, and greater than or equal to 384 times. The corresponding minimum clustering size values are 3, 4, 6, 8, 10, and 12 respectively.

[0115] S6. Extraction of influencing factors of user preference types: Based on the common locations of users identified in step S5, establish a buffer zone, and extract the surrounding built environment characteristics, including the number of various types of POIs in the city, slope, road network density, the distances from the common location to the nearest public transportation and the nearest CBD, and the attributes of shared infrastructure.

[0116] Specifically, it includes the following:

[0117] S61. Establish a buffer zone: Based on the user's frequently used locations obtained in step S5, use ArcGIS software to establish a 500-meter buffer zone for extracting influencing factors.

[0118] S62. Extract the quantities of various POIs: Extract the quantities of public service facilities, leisure service facilities, commercial service facilities, residential facilities, and office facilities within the buffer zone.

[0119] S63. Slope: Extract the average slope within the buffer zone around the user's frequently used locations.

[0120] S64. Road network density: Extract the road length within the buffer zone, and the ratio of this road length to the buffer zone area is the road network density.

[0121] S65. Calculate the nearest distances: Calculate the distances between the frequently used locations and the nearest bus stops, subway stations, and the nearest CBDs.

[0122] S66. Extraction of shared infrastructure attributes: Include the quantity of vehicle supply and the number of electronic fences within the buffer zone; the extraction of the vehicle supply within the buffer zone includes: randomly select a typical working day from the data preprocessed in step S2 as a representative, calculate the distances between the pick-up points and drop-off points of all orders from 00:00 to 23:59 on this day and the user's frequently used locations, screen out the order records located within the buffer zone and number them, such as retaining the order records less than 500 meters and numbering them.

[0123] Then divide this day into several time periods, such as 124 intervals at 10-minute intervals. According to the pick-up time and numbering, count the pick-up quantities of shared bicycles and shared electric bicycles in each buffer zone within each time period.

[0124] For the same buffer zone, the pick-up quantity in the next time period is the available quantity of vehicles in the previous time period; by calculating the average value of the pick-up quantities in all time periods of a day, obtain the average supply quantities of shared bicycles and shared electric bicycles. Thus, analyze the influence of the vehicle supply on user selection. In addition, the number of electronic fences around the frequently used locations is also considered.

[0125] S7. Construction and prediction of the user preference type prediction model: Construct a multinomial Logit (MNL) model as the user preference type prediction model and train it, input the data to be measured, use the trained model to predict the user's preference type, and output the prediction result.

[0126] Specifically, it includes the following:

[0127] S71. Build a model: Using the user preference type (shared bicycle-dominated, balanced, shared electric bicycle-dominated) as the dependent variable, and taking the built environment characteristics and shared infrastructure attributes in step S6 as explanatory variables, establish a multinomial Logit (MNL) model;

[0128] When building the model, select the "balanced" users as the control group. First, the entire dataset is divided. 80% of the data is used for the training set to build the model, and 20% of the data is used for the test set to evaluate the model performance. Next, a multicollinearity test is conducted on each explanatory variable in the training set. The results show that the variance inflation factor (VIF) of all explanatory variables is less than 7.5, indicating that there is no serious multicollinearity problem among the variables, and the model analysis can continue.

[0129] S72. Train the model: Determine the relationship between the explanatory variables and the user preference type through learning, and use the maximum likelihood estimation method to estimate the parameters of the MNL model;

[0130] On the training set, the model parameters are estimated by the maximum likelihood estimation method. Table 3 shows the final MNL model results. The goodness of fit of the model is evaluated on the training set through the pseudo R 2 index. The results show that the overall goodness of fit of the model is good, and it can effectively capture the relationship between users' travel preferences and their influencing factors. In addition, to verify the robustness and generalization ability of the model, model prediction verification is carried out on the training set and the test set respectively. Table 4 shows the prediction accuracies of both. The results show that the generalization performance of the model is good and the prediction effect is stable, further proving the model's prediction ability on unseen data.

[0131] Table 3 Model Results

[0132]

[0133]

[0134] Table 4 Prediction Accuracies of the Training Set and the Test Set

[0135]

[0136] S73, Model prediction: According to the estimated model parameters (fitted MNL model parameters), input new built environment characteristics and shared infrastructure attribute data to predict the user's preferred type; among them, the built environment characteristic data is extracted through the urban geographic information system (GIS), covering road network data, public transportation data and various types of facility network data, etc. The data can be obtained from the urban planning department or open GIS platform; for the deployment planning of Internet rental bicycles in new areas, the shared infrastructure attribute data can come from shared travel service providers or government departments, covering information such as the expected total deployment and the planned number of electronic fences. These multi-source data provide a solid data foundation for model prediction, thereby effectively predicting whether users prefer shared bicycles, shared electric motorcycles, or a balanced use of both.

[0137] S74. Output prediction results: Calculate the probability that the user belongs to each preference type through the fitted MNL model, and select the preference type with the highest probability as the final prediction result.

[0138] This prediction method based on multi-source data can help understand user needs more accurately and provide support for subsequent traffic planning and market operation strategies.

[0139] Through the above steps, the method of the present invention has shown strong practicality, accuracy and scalability in predicting the travel preference types of Internet rental bicycle users. This not only helps to gain a deeper understanding of users' travel habits and needs, and helps operators optimize the placement and quantity of shared bicycles and shared electric motorcycles, but also provides scientific decision-making support for urban traffic planning, improves resource allocation efficiency, and further improves user experience.

[0140] Based on the same inventive concept as the above method embodiment, the present application embodiment also provides an Internet rental bicycle user preference type prediction system based on order data, which can implement the functions provided by the above method embodiment, such as Figure 3 As shown, the system includes the following modules:

[0141] The database construction module 110 constructs a user preference type prediction database, including order data of two main types of Internet rental bicycles, namely, shared bicycles and shared electric motorcycles, as well as built environment data and electronic fence data;

[0142] The data preprocessing module 120 cleans the shared bicycle and shared electric motorcycle data in the database and removes invalid data and abnormal data;

[0143] The user identification module 130 identifies the travel records of all shared bicycles and shared electric motorcycles under the same ID number based on the user ID, calculates the number of times and days each user uses the shared bicycle and shared electric motorcycle according to the vehicle type and borrowing time, and statistically analyzes the frequency distribution of users using shared bicycles and shared electric motorcycles;

[0144] The user preference type classification module 140 divides the users into three types: shared bicycle dominant type, balanced type and shared motorcycle dominant type according to the frequency distribution of users using shared bicycles and shared motorcycles;

[0145] The user frequently used location identification module 150 uses the HDBSCAN clustering algorithm to identify the user frequently used locations, i.e. the user frequently uses the location to borrow and return the car;

[0146] The influencing factor extraction module 160 establishes a buffer zone based on the identified user-frequently used places, and extracts the surrounding built environment characteristics, including the number of various POIs in the city, slope, road network density, distances between the frequently used places and the nearest public transportation and the nearest CBD, and shared infrastructure attributes;

[0147] The prediction model construction and prediction module 170 constructs a multinomial Logit (MNL) model as a user preference type prediction model and trains it, inputs the test data, uses the trained model to predict the user's preference type, and outputs the prediction result.

[0148] Based on the same inventive concept as the above method embodiment, the present application embodiment also provides an electronic device, such as Figure 4 As shown, the device includes: a processor 210; and a memory 220, for storing one or more programs;

[0149] When the one or more programs are executed by the processor 210, the processor executes the method for predicting the user preference type of Internet rental bicycles based on order data.

[0150] The method for predicting the user preference type of Internet rental bicycles based on order data specifically includes the following steps:

[0151] Build a user preference type prediction database, including order data of two main types of Internet rental bicycles, namely shared bicycles and shared electric motorcycles, as well as built environment data and electronic fence data;

[0152] Data preprocessing: Clean the shared bicycle and shared electric motorcycle data in the database and remove invalid and abnormal data;

[0153] Identification of the same user in shared riding: Based on the user ID, the travel records of all shared bicycles and shared electric motorcycles with the same ID number are identified. The number of times and days each user used shared bicycles and shared electric motorcycles are calculated according to the vehicle type and borrowing time, and the frequency distribution of users' use of shared bicycles and shared electric motorcycles is statistically analyzed;

[0154] Classification of user preference types: Based on the frequency distribution of users using shared bicycles and shared electric motorcycles, users are divided into three categories: shared bicycle dominant, balanced, and shared electric motorcycle dominant;

[0155] Identification of user-frequented locations: HDBSCAN clustering algorithm is used to identify user-frequented locations, i.e., the locations where users frequently borrow and return bicycles;

[0156] Extraction of factors influencing user preference types: Based on the identified user-frequently used locations, a buffer zone is established to extract the surrounding built environment characteristics, including the number of various POIs in the city, slope, road network density, the distance between the frequently used locations and the nearest public transportation and the nearest CBD, and shared infrastructure attributes;

[0157] Construction and prediction of user preference type prediction model: Construct a multinomial Logit (MNL) model as a user preference type prediction model and train it. Input the test data and use the trained model to predict the user's preference type, and output the prediction results.

[0158] Based on the same inventive concept as the above-mentioned method embodiment, the embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by the processor 210, the method for predicting the user preference type of Internet rental bicycles based on order data is implemented.

[0159] The method for predicting the user preference type of Internet rental bicycles based on order data specifically includes the following steps:

[0160] Build a user preference type prediction database, including order data of two main types of Internet rental bicycles, namely shared bicycles and shared electric motorcycles, as well as built environment data and electronic fence data;

[0161] Data preprocessing: Clean the shared bicycle and shared electric motorcycle data in the database and remove invalid and abnormal data;

[0162] Identification of the same user in shared riding: Based on the user ID, the travel records of all shared bicycles and shared electric motorcycles with the same ID number are identified. The number of times and days each user used shared bicycles and shared electric motorcycles are calculated according to the vehicle type and borrowing time, and the frequency distribution of users' use of shared bicycles and shared electric motorcycles is statistically analyzed;

[0163] Classification of user preference types: According to the frequency distribution of users' use of shared bicycles and shared electric bicycles, users are divided into three categories: shared bicycle-dominated, balanced, and shared electric bicycle-dominated.

[0164] Identification of users' common locations: The HDBSCAN clustering algorithm is used to identify users' common locations, that is, the locations where users commonly borrow and return vehicles.

[0165] Extraction of influencing factors of user preference types: Based on the identified common locations of users, a buffer zone is established, and the surrounding built environment features are extracted, including the number of various types of POIs in the city, slope, road network density, distances from the common location to the nearest public transportation and the nearest CBD, and the attributes of shared infrastructure.

[0166] Construction and prediction of user preference type prediction model: A multinomial Logit (MNL) model is constructed as the user preference type prediction model and trained. The input test data is used to predict the user's preference type by the trained model, and the prediction results are output.

[0167] In several embodiments provided in the embodiments of the present invention, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device and method embodiments described above are only illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions, and operations of devices, methods, and computer program products according to multiple embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order from that marked in the accompanying drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, as well as the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0168] In addition, the functional modules in each embodiment of the present invention can be integrated together to form an independent part, or each module can exist separately, or two or more modules can be integrated to form an independent part.

[0169] When the above-mentioned function is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, an electronic device, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs. It should be noted that in this article, the terms "include", "comprise", or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such a process, method, article, or device. Without further limitations, an element defined by the statement "including a..." does not exclude the existence of another identical element in the process, method, article, or device including the said element.

[0170] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The singular forms "a", "the", and "said" used in the embodiments of the present invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0171] It should be understood that the term "and / or" used herein is only a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this article generally represents an "or" relationship between the associated objects before and after.

[0172] Depending on the context, the word "if" as used herein can be interpreted as "when", "while", "in response to determining", or "in response to detecting". Similarly, depending on the context, the phrase "if determined" or "if detected (stated condition or event)" can be interpreted as "when determined", "in response to determining", "when detecting (stated condition or event)", or "in response to detecting (stated condition or event)".

[0173] The "first / second" mentioned in the embodiments is only used to distinguish similar objects and does not represent a specific order for the objects. Understandably, the "first / second" can be interchanged in a specific order or sequence when permitted. It should be understood that the objects distinguished by the "first / second" can be interchanged under appropriate circumstances so that the embodiments described herein can be implemented in an order other than those illustrated or described herein.

[0174] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for predicting user preference types of Internet rental bicycles based on order data. It is characterized in that The steps include: S1. Build a user preference type prediction database, including the order data of two types of Internet rental bicycles, namely shared bicycles and shared electric motorcycles, as well as built environment data and electronic fence data; S2. Data preprocessing: Clean the shared bicycle and shared electric motorcycle data in the database of step S1 to remove invalid and abnormal data; S3. Identification of the same user in shared riding: Based on the user ID, the travel records of all shared bicycles and shared electric motorcycles with the same ID number are identified. The number of times and days each user uses shared bicycles and shared electric motorcycles are calculated according to the vehicle type and borrowing time, and the frequency distribution of users using shared bicycles and shared electric motorcycles is statistically analyzed; S4. Classification of user preference types: Based on the frequency distribution of users using shared bicycles and shared electric motorcycles, users are divided into three categories: shared bicycle dominant type, balanced type, and shared electric motorcycle dominant type; S5. Identification of user-frequently used locations: using the HDBSCAN clustering algorithm to identify user-frequently used locations, i.e., locations where users frequently borrow and return vehicles; S6. Extraction of factors influencing user preference types: Based on the frequently used places of users identified in step S5, a buffer zone is established to extract the surrounding built environment characteristics, including the number of urban POIs, slope, road network density, the distance between the frequently used places and the nearest public transportation and the nearest CBD, and shared infrastructure attributes; S7. Construction and prediction of user preference type prediction model: Construct a multinomial Logit (MNL) model as a user preference type prediction model and train it. Input the test data and use the trained model to predict the user's preference type, and output the prediction result.

2. The method for predicting user preference types of Internet rental bicycles based on order data according to claim 1, Features: The built environment data in step S1 includes: digital elevation model DEM data, road network data, public transportation data and facility network data; the facility network data includes multiple categories of points of interest POI such as shopping, leisure and entertainment, and commercial services. The data contains both geographic location information and facility type information, and finally the data is processed through ArcGIS; the electronic fence data includes: fence name, center point address, area, and latitude and longitude coordinates of the center point and area vertices.

3. The method for predicting user preference types of Internet rental bicycles based on order data according to claim 1, Features: In step S2, the specific steps include: S21. Eliminate incomplete records of shared bicycle and shared electric motorcycle order data; wherein a complete order data includes information of user ID, vehicle ID, vehicle type, borrowing date, borrowing time, borrowing longitude, borrowing latitude, return date, return time, return longitude, and return latitude; S22. Eliminate the data where the geographical locations of shared bicycles and shared electric motorcycles borrowed and parked deviate from the predicted range; S23. Exclude records with a one-way travel duration less than 1 minute or greater than 2 hours. Here, the travel duration is obtained by subtracting the car rental time from the car return time in the same order. S24. Exclude records with a one-way travel distance less than 100 meters. Here, the travel distance is obtained by calculating the latitude and longitude of the car rental and return in the same order.

4. The method for predicting the user preference type of Internet rental bicycles based on order data according to claim 1, characterized in that: In step S4, it specifically includes the following: Calculate the proportion of the number of shared bicycle trips of the user. The proportion of the number of shared bicycle trips is obtained by dividing the number of shared bicycle trips by the total number of trips. According to the proportion of the number of shared bicycle trips, the users are equally divided into three types. The specific division includes: Users with a relatively high proportion of shared bicycle usage, that is, users with a proportion of shared bicycle trips ≥ 2 / 3 are classified as "shared bicycle dominant type"; Users with a relatively balanced proportion of shared bicycle and shared electric bicycle usage, that is, users with 1 / 3 < proportion of shared bicycle trips < 2 / 3 are classified as "balanced type"; Users with a relatively low proportion of shared bicycle usage, that is, users with a proportion of shared bicycle trips ≤ 1 / 3 are classified as "shared electric bicycle dominant type".

5. The method for predicting the user preference type of Internet rental bicycles based on order data according to claim 1, characterized in that: In step S5, the method for using HDBSCAN clustering to identify the user's frequently used places specifically includes: S51. Input data set: Input the latitude and longitude of the car rental and return of all shared bicycle users. S52. Coordinate system conversion: Convert the latitude and longitude of the car rental and return of the user from the geographic coordinate system to the projected coordinate system. S53. Group by user ID and divide the travel records of the user into different clusters. S54. Set parameters: Define the minimum number of sample points min_cluster_size required to form a cluster corresponding to the user travel times interval. S55. Calculate the core distance and density estimation: Calculate the core distance of each pick-up and drop-off point of the user, and define the core distance c k (x) = d(x, N k (x)), which is the distance from the current point x to the k-th nearest point N k (x), and the mutual reachability distance d k (a, b) is expressed as: d k (a, b) = max{c k (a), c k (b), d(a, b)} where d(a, b) is the original distance between points a and b; S56. Construct a density adjacency graph: Based on the mutual reachability distance, construct a weighted adjacency graph, where the weight of the edge is the mutual reachability distance between the point pairs. S57. Generate a minimum spanning tree: The Prim algorithm is used internally in the algorithm. Using the sample points as the original points and the mutual reachability distance between other points and the original points as the weights to construct a minimum spanning tree. S58. Establish a cluster hierarchy: The algorithm traverses and reorders the edges of the minimum spanning tree with the mutual reachability distance as the weight, and classifies each edge into a new cluster. S59. Extract stable clusters: According to the persistence analysis and the min_cluster_size parameter, extract the most stable cluster as the final result. S510. Determine the frequently used place: The clustering result may obtain more than one cluster. Each cluster represents a place where the user often rents and returns the vehicle. Select the cluster with the largest number of car rental and return points in the clustering cluster, that is, the most representative cluster, as the user's frequently used place.

6. The method for predicting the user preference type of Internet rental bicycles based on order data according to claim 1, characterized in that: In step S6, it specifically includes the following: S61, establishing a buffer zone: based on the user's frequently used locations obtained in step S5, using ArcGIS software to establish a buffer zone for extracting influencing factors; S62, extracting the number of POIs: extracting the number of public service facilities, leisure service facilities, commercial service facilities, residential facilities and office facilities in the buffer zone; S63, slope: extracting the average slope value within the buffer zone around the user's frequently used area; S64, road network density: extract the road length within the buffer zone, and the ratio of the road length to the buffer zone area is the road network density; S65. Calculate the nearest distance: Calculate the distance between the frequently used place and the nearest bus stop, subway station and the nearest CBD; S66. Extraction of shared infrastructure attributes: including the supply of vehicles in the buffer zone and the number of electronic fences; The extraction of vehicle supply in the buffer zone includes: randomly selecting a typical working day as a representative from the data pre-processed in step S2, calculating the distances between the borrowing point and the return point of all orders on that day and the user's usual place, and filtering out the order records in the buffer zone and numbering them; Then divide the day into several time periods, and count the number of shared bicycles and shared electric motorcycles borrowed in each buffer zone in each time period according to the borrowing time and number; For the same buffer zone, the number of borrowed vehicles in the next time period is the number of available vehicles in the previous time period; by calculating the average number of borrowed vehicles in all time periods of the day, the average supply of shared bicycles and shared electric motorcycles can be obtained.

7. The method for predicting user preference types of Internet rental bicycles based on order data according to claim 1, Features: In step S7, the specific steps include: S71, constructing a model: taking the user preference type, i.e., shared bicycle dominant type, balanced type, and shared electric motorcycle dominant type as the dependent variable, and taking the built environment characteristics and shared infrastructure attributes in step S6 as the explanatory variables, a multinomial logit (MNL) model is established; S72, training model: determining the relationship between the explanatory variables and the user preference type through learning, and estimating the parameters of the MNL model using the maximum likelihood estimation method; S73, model prediction: based on the estimated model parameters, input new built environment characteristics and shared infrastructure attribute data to predict the user's preference type; S74. Output prediction results: Calculate the probability that the user belongs to each preference type through the fitted MNL model, and select the preference type with the highest probability as the final prediction result.

8. A prediction system according to the method for predicting user preference types of Internet rental bicycles based on order data according to any one of claims 1 to 7, Features: The system includes the following modules: A database construction module (110) constructs a user preference type prediction database, including order data of two types of Internet rental bicycles, namely shared bicycles and shared electric motorcycles, as well as built environment data and electronic fence data; A data preprocessing module (120) cleans the shared bicycle and shared electric motorcycle data in the database to remove invalid data and abnormal data; The user identification module (130) identifies the travel records of all shared bicycles and shared electric bicycles under the same ID number based on the user ID, calculates the number of times and days that each user uses shared bicycles and shared electric bicycles respectively according to the vehicle type and borrowing time, and statistically analyzes the frequency distribution of users' use of shared bicycles and shared electric bicycles; The user preference type classification module (140) classifies users into three categories: shared bicycle-dominated type, balanced type, and shared electric bicycle-dominated type according to the frequency distribution of users' use of shared bicycles and shared electric bicycles; The user's common location identification module (150) uses the HDBSCAN clustering algorithm to identify the user's common locations, that is, the common borrowing and returning locations of the user; The influencing factor extraction module (160) establishes a buffer based on the identified user's common locations and extracts the surrounding built environment features, including the number of urban POIs, slope, road network density, the distances from the common location to the nearest public transportation and the nearest CBD, and the attributes of shared infrastructure; The prediction model construction and prediction module (170) constructs a multinomial Logit (MNL) model as a user preference type prediction model and trains it, inputs the data to be measured, uses the trained model to predict the user's preference type, and outputs the prediction result.

9. An electronic device, characterized in that: The electronic device includes: a processor (210); and a memory (220) for storing one or more programs; When the one or more programs are executed by the processor (210), the processor is caused to execute the method for predicting the user preference type of internet rental bicycles based on order data according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by the processor (210), the method for predicting the user preference type of internet rental bicycles based on order data according to any one of claims 1-7 is implemented.

Images