Bus user travel purpose prediction method and equipment based on multi-source data

Abstract

The invention relates to a method and equipment for predicting the travel purpose of a bus user based on multi-source data, and the method comprises the steps: building a feature project based on the joint analysis of the multi-source data of urban residents, and the multi-source data comprise travel survey data, bus card swiping data and interest point data; according to whether the home address and the work place address in the record are both filled in, dividing the record into a first type of travel data and a second type of travel data, and respectively constructing a first model and a second model by using a machine learning model; identifying a home station address and a working site station address of each user in the bus card swiping data, and dividing the bus card swiping data into a first type and a second type according to whether the home station address and the working site station address of the user can be identified; and using the first model and the second model to predict the travel purpose for the classified card swiping data. Compared with the prior art, the method has the advantages of high prediction precision, high efficiency and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of public transport user travel purpose prediction, and in particular to a public transport user travel purpose prediction method based on multi-source data. Background Art

[0002] In the field of transportation, data-driven research methods are rapidly emerging, and optimizing public transportation services through big data has become a new trend. However, public transportation card swiping data often lacks information about individual travel purposes, making it impossible to understand the motivations and reasons behind user travel behavior, limiting public transportation operators' in-depth understanding of user behavior.

[0003] At the same time, the resident travel survey data contains rich information on the spatiotemporal characteristics of individual activities and travel purpose, which is of great significance for analyzing the relationship between public transportation users' travel behavior and travel purpose. These survey data often require a lot of manpower and material resources, resulting in a small sample size. Although card swiping data can be collected automatically, it is impossible to directly obtain the user's travel purpose. If the advantages of these two types of data can be combined, a large-scale card swiping data set containing users' travel purposes can be constructed, thus laying the foundation for studying the relationship between users' travel needs and travel motivations. Current research has not effectively integrated these two types of data, which has led to challenges in inferring the travel purpose of regular public transportation users.

[0004] After searching, Chinese invention patent publication number CN111968018A discloses a method and device for identifying the purpose of subway travel, which can improve the accuracy of travel purpose identification. The method includes: obtaining subway travel samples, and obtaining land use characteristic information around subway stations; dividing the subway travel samples into a training set and a test set; using the travel characteristic information in the training set and the corresponding land use characteristic information around the starting and ending subway stations as input, and the travel purpose as a predictor variable, training the classifier, and testing the trained classifier based on the test set to obtain the tested classifier; inputting the target travel characteristic information corresponding to each target travel record in the bus smart card data and the corresponding land use characteristic information around the starting and ending subway stations into the tested classifier to obtain the travel purpose identification result of each target travel record. The existing patent has the problem of not classifying travel data, resulting in low recognition accuracy.

[0005] How to accurately predict the travel purposes of public transportation users has become a technical problem that needs to be solved. Summary of the invention

[0006] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a method for predicting the travel destination of public transportation users based on multi-source data.

[0007] The purpose of the present invention can be achieved by the following technical solutions:

[0008] According to one aspect of the present invention, a method for predicting travel destinations of public transportation users based on multi-source data is provided, the method comprising the following steps:

[0009] Step S1, establishing feature engineering based on joint analysis of multi-source data of urban residents, where the multi-source data includes travel survey data, bus card swiping data and point of interest data;

[0010] Step S2, dividing the records of the travel survey data in which the travel mode is conventional public transportation into the first type of travel data and the second type of travel data according to whether both the home address and the workplace address are filled in the records, and using the machine learning model to construct the first model and the second model respectively;

[0011] Step S3, identifying the home station address and the workplace station address of each user in the bus card swiping data, and classifying the data into the first type of card swiping data and the second type of card swiping data according to whether the home station address and the workplace station address of the user can be identified;

[0012] Step S4, predicting the travel purpose using the first model and the second model respectively for the card swiping data classified in step S3.

[0013] Preferably, the process of establishing feature engineering includes: according to the travel starting point and travel end point of the travel survey data, and the boarding and alighting stations of the bus card swiping data, respectively calculating the proportion of different types of points of interest within the surrounding distance threshold range, as well as the cosine value of the departure time, the cosine value of the arrival time and the travel time as the characteristics of the user's travel.

[0014] Preferably, if the home address and the work address of a certain travel survey data are not missing, it is classified as the first category of travel data, otherwise it is classified as the second category of travel data.

[0015] Preferably, the process of constructing the first model includes: filtering out records of the first type of travel data whose travel purpose is commuting or going home, dividing the remaining data into a training set and a test set, and using a machine learning model to train and generate the first model;

[0016] The construction process of the second model includes: selecting all second-category travel data, dividing them into a training set and a test set, and using a machine learning model to train and generate a second model.

[0017] More preferably, the machine learning model is an XGBoost model.

[0018] Preferably, the identification of the home site address and workplace site address of each user in the bus card swiping data in step S3 is specifically as follows: in the bus card swiping data for a period of time, the departure site with the largest number of first trips for the user and a proportion of the number of trips to all days exceeding a first threshold is identified as the user's home site address, and the departure site with the largest number of last trips for many days for the user and a proportion of the number of trips to all days exceeding a second threshold is identified as the user's workplace site address.

[0019] Preferably, the process of dividing the card swiping data into the first category and the second category includes:

[0020] If both the home stop address and the workplace stop address can be identified in a bus travel record, it is classified as the first type of card swiping data, otherwise it is classified as the second type of card swiping data.

[0021] Preferably, the step S4 comprises:

[0022] For the first type of card swiping data, if the distance between the alighting station and the home station address is ≤ the third threshold, it is identified as the purpose of "going home"; if the distance between the alighting station and the workplace station address is ≤ the fourth threshold, it is identified as the purpose of "commuting"; for the remaining first type of card swiping data with unidentified travel purposes, the first model is used to predict the travel purpose;

[0023] For the second type of card swiping data, the second model is used to predict the travel purpose.

[0024] Preferably, the method further comprises evaluating the prediction results based on multi-dimensional evaluation indicators, including:

[0025] For travel survey data, the prediction results are evaluated in terms of accuracy, precision, and recall;

[0026] For bus card swiping data, the Jensen-Shannon divergence of the prediction results is calculated for evaluation.

[0027] According to another aspect of the present invention, an electronic device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the method described above is implemented when the processor executes the program.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] 1) The present invention uses multi-source data joint analysis to establish a travel feature project for public transportation users, and analyzes the relationship between travel spatiotemporal characteristics and travel purposes; the travel survey data and public transportation card swiping data are classified according to whether the user's home address and workplace address are filled in or recognized, and for different types of travel data, the machine learning model is used to build and train their own prediction models, and the trained different types of prediction models are applied to different types of card swiping data to predict the travel purposes of public transportation users, and the prediction accuracy is more accurate than the previous model without partitioning categories.

[0030] 2) The present invention classifies both travel survey data and bus card swiping data, and then uses different models to predict different types of bus card swiping data. The fine classification enables the output of more accurate travel purpose predictions, which helps to explore the deep mechanisms and motivations behind bus users' travel behaviors and obtain a more scientific and accurate analysis of the user travel behavior mechanism. This is of great significance for accurately grasping user travel needs and promoting bus companies to efficiently and reasonably allocate resources.

[0031] 3) For the first type of card swiping data, the present invention first adopts rules to identify the travel purposes of commuting and going home, and then uses the first model to output the prediction results of other travel purposes. The hierarchical prediction enhances the accuracy and efficiency of the prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a flow chart of the method for predicting the travel purpose of public transportation users in the present invention;

[0033] Figure 2 This is a schematic diagram of the process of identifying the travel purpose of the bus card swiping data in the present invention;

[0034] Figure 3 The confusion matrix of the prediction results of the first model and travel purpose in the present invention;

[0035] Figure 4 The confusion matrix of the travel destination prediction results of the second model in the present invention;

[0036] Figure 5 This is a schematic diagram of the proportion distribution of travel survey data and bus card swiping data based on travel purpose in the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0038] The bus card swiping data does not contain the user's travel purpose information. The existing methods fail to effectively integrate multi-source data. The present invention accurately grasps user needs by establishing appropriate feature engineering for inferring travel purposes, which can be used to improve the service level of bus companies.

[0039] Example 1

[0040] This embodiment relates to a method for predicting the travel destination of public transportation users based on multi-source data. Figure 1 , the method comprising:

[0041] Step S1, obtaining the travel survey data D of urban residents survey , bus card swiping data D card And point of interest data, including: for the travel starting points and destinations of the travel survey data, and the boarding and alighting points of the bus card swiping data, feature engineering is established according to the proportion of different types of points of interest and the departure and arrival times (or boarding and alighting times) within the distance threshold range around the starting and ending points (or boarding and alighting points).

[0042] Use electronic map (such as Baidu Map, Amap, etc.) API to obtain POI data, including six categories of POI: catering, companies, leisure and entertainment, science and education, business and residential, and medical care;

[0043] For travel survey data D survey , obtain the coordinates of the starting point and the end point of each trip survey data. Filter all points of interest in the point of interest data that are less than 600 meters away from the corresponding station as set A, calculate the proportion of all different types of points of interest in set A, and calculate the cosine value of the departure time and arrival time and the travel time as the characteristics of the trip;

[0044] For bus card swipe data D card , get the boarding station coordinates and the alighting station coordinates of each card swiping data. Filter all the points of interest whose distance from the corresponding station is less than 600 meters in the point of interest data as set B, calculate the proportion of all different types of points of interest in set B, and calculate the cosine value of the departure time and arrival time and the travel time as the features of the ride record.

[0045] Step S2, from the acquired travel survey data, filter out records where the travel mode is conventional public transportation, where conventional public transportation includes urban public transportation such as buses and subways. According to whether the home address and workplace address of the user to whom the travel survey data belongs are filled in, divide all travel records into first and second categories of travel data, and train first and second category travel purpose classification machine learning models respectively, corresponding to the first model and the second model respectively, including:

[0046] Step S2-1, travel survey data Dsurvey Classification is performed. If the home address and the work address of a travel data are not missing, it is classified as the first type of travel data, otherwise it is classified as the second type of travel data;

[0047] Step S2-2, filter out the records whose travel purpose is commuting or going home in the first category of travel data, take the remaining data (the travel purpose in the first category of travel data is neither commuting nor going home) and divide it into training sets and test sets, and use machine learning training to generate the first model; select all the second category of travel data, divide it into training sets and test sets, and use machine learning training to generate the second model.

[0048] Step S3: for the acquired bus card swiping data, identify the home station address and the workplace station address of each user, and classify all card swiping records into the first category and the second category according to whether the home station address and the workplace station address of the user to whom the card swiping data belongs can be identified, including:

[0049] Step S31, identifying each user's home station address and work station address: filtering bus card swiping data D card For a longer period of time (e.g., one month), the home and work stations of each user are identified. The identification rule is: the departure station with the most first trips in multiple days and the proportion of the number of trips to all days is higher than the threshold, which is identified as the user's home station address. home The departure station with the most last trips in multiple days and a ratio higher than the threshold for all days is identified as the user's workplace station address s work ;

[0050] Step S32, classifying the processing result of step S31: if both the home site address and the work site address of the resident on a certain trip can be identified, it is classified as the first type of card swiping data, otherwise it is classified as the second type of card swiping data;

[0051] Step S33: For the first type of card swiping data, if the alighting station is al and home site addresses home Satisfy dis(s al ,s home )≤600 meters, it is identified as the "return home" destination. If the alighting station is s al and work site address work Satisfy dis(s al ,s work )≤600 meters, it is identified as "commuting". For the other two types of unidentified card swiping data, the two models trained in step S2 are used to identify the travel purpose, such as Figure 2 .

[0052] Step S4: Apply the first model and the second model to the first and second types of card swiping data respectively, and evaluate the effectiveness of the inference model using multi-dimensional evaluation indicators such as accuracy, precision, recall, and Jensen-Shannon divergence.

[0053] In the travel survey data survey On the test set, the accuracy (Acc), precision (Pre), and recall (Rec) of the final model are calculated using the following formula:

[0054]

[0055] Among them, TP, FP, FN, and TN are the number of true positives, true negatives, false positives, and false negatives, respectively, and K is the total number of categories.

[0056] Data D on bus card swipe card The Jensen-Shannon divergence of the final inference model result is calculated using the following formula:

[0057]

[0058] Where p is the true distribution of travel purpose in the residents’ travel survey data, q is the estimated distribution of travel purpose in the card swiping data, and D KL (a||b) is the Kullback-Leibler divergence between distribution a and distribution b.

[0059] The present invention utilizes multi-source data to establish a single trip feature project for public transport users, analyzes the relationship between travel spatiotemporal characteristics and travel purposes, and constructs two models of public transport users' travel purposes. This helps to explore the deep mechanisms and motivations behind public transport users' travel behaviors and obtain a more scientific and accurate analysis of the user travel behavior mechanism. This is of great significance for accurately grasping user travel needs and promoting public transport companies to efficiently and reasonably allocate resources.

[0060] Example 2

[0061] The present embodiment also relates to a method for predicting travel destinations of public transportation users based on multi-source data, and a verification test for predicting travel destinations of public transportation users based on semi-supervised learning.

[0062] In this embodiment, the selected research data include travel survey data of a city in 2020, bus card swiping data from October to December 2018, and points of interest data of the city in 2018.

[0063] The original travel survey data contains 11 travel purpose categories, which are merged to obtain four travel purpose categories: commuting, going home, entertainment and medical treatment. The travel proportions of these four purposes are 41.4%, 48.5%, 9.6% and 0.5% respectively.

[0064] For the travel survey data and bus card swiping data, feature engineering is established according to the proportion of different types of points of interest within 600 meters around the starting and ending points (or boarding and alighting stations), departure time, arrival time (or boarding and alighting time) and travel time. The selection of detailed feature variables is shown in Table 1 below.

[0065] Table 1

[0066] variable describe Boarding time The position of the boarding time in a 24-hour day (converted to cosine value) Drop-off time The position of the time of getting off in a 24-hour day (converted to cosine value) Duration The time spent on this trip Starting point_catering Proportion of food and beverage points of interest around the starting point (boarding station) Starting point_Leisure Proportion of leisure and entertainment points of interest around the starting point (boarding station) Starting point_Science and Education Proportion of science, education and culture points of interest around the starting point (boarding station) Starting point_Medical Proportion of healthcare points of interest around the starting point (boarding station) Starting point_Enterprise Proportion of corporate points of interest around the starting point (boarding station) Starting point_Residential Proportion of commercial and residential points of interest around the starting point (boarding station) End_Catering Proportion of food and beverage points of interest around the destination (get-off station) End_Leisure Proportion of leisure and entertainment points of interest around the destination (get-off station) End_Science and Education Proportion of science, education and culture points of interest around the destination (get-off station) End_Medical Proportion of medical and health care points of interest around the destination (get-off station) End_Enterprise Proportion of corporate points of interest around the destination (get-off station) End_Residential Proportion of commercial and residential points of interest around the destination (get-off station)

[0067] For the travel survey data, all travel records are divided into the first category and the second category according to whether the home address and work address of the user to whom the travel record belongs are filled in. Extract the records of the first category of data whose travel purpose is neither commuting nor going home, and train the first category of travel purpose inference model, that is, the first model; use all the second category of data to train the second category of travel purpose inference model, that is, the second model. The division ratio of the training set and the test set in the above two training processes is 80% and 20% respectively. Based on the two training sets, the XGBoost model is trained separately, and its hyperparameters are determined by grid search. The trained models are the first model and the second model respectively. The specific hyperparameter values ​​are shown in Table 2.

[0068] Table 2

[0069] Hyperparameters First Model Second Model Number of estimators 140 140 Maximum Depth 7 7 Minimum child node weight 2 8 Subsample rate 0.6 0.8 Learning Rate 0.1 0.1

[0070] For bus card swiping data, the home station address and workplace station address of each user are identified based on three months of card swiping data. The identification rule is: the departure station with the most first trips in multiple days and the proportion of the number of trips to all days is higher than 0.6, which is identified as the user's home station address. home The departure station with the most last trips in multiple days and a ratio higher than 0.6 of the number of trips in all days is identified as the user's workplace station address s work .

[0071] According to whether the home and work station addresses of the user to whom the bus card swipe data belongs can be identified, all card swipe records are divided into the first and second categories of card swipe data. If the home and work station addresses of the residents of a certain trip can be identified, it is classified as the first category of card swipe data, otherwise it is classified as the second category of card swipe data.

[0072] For the first type of card swiping data, if the alighting station is s al and homeSatisfy dis(s al ,s home )≤600 meters, it is identified as the "return home" destination. If the alighting station is s al and work Satisfy dis(s al ,s work )≤600 meters, it is identified as a “commuting” purpose.

[0073] Thereafter, the first model and the second model are applied to the remaining unrecognized first and second category card swiping data, respectively. Figure 2 As shown, the following steps are included:

[0074] Step S3-1, determining whether the home site address and the workplace site address can be identified, if yes, determining it as the first type of travel data, and executing step S3-2; otherwise, determining it as the second type of travel data, and executing step S3-4;

[0075] Step S3-2, identifying the home site address and the work site address of each user based on the identification rules, and identifying the travel purpose, wherein the travel purpose includes commuting and going home, and executing step S3-3;

[0076] Step S3-3, using the trained first model to predict the travel purpose of the remaining first category travel data, the predicted travel purposes include leisure and medical treatment, and then ending;

[0077] Step S3-4, use the trained second model to predict the travel purpose of the second type of travel data, the predicted travel purposes include commuting, going home, leisure and medical treatment, and end.

[0078] The hierarchical identification and prediction in step S3-2 and step S3-3 reduce the prediction time of the first type of travel data and improve efficiency.

[0079] For the constructed travel purpose recognition model, the effectiveness of the travel model is evaluated using multi-dimensional evaluation indicators such as accuracy, precision, recall, and Jensen-Shannon divergence. The prediction results of the first model and the second model of this patent are compared with the true value on the test set, and the confusion matrix obtained is as follows: Figure 3 and Figure 4 As shown in the figure, the accuracy of the first model and the second model on their respective data sets is 86.7% and 78.3% respectively. Considering the overall prediction process of the model (including the results of recognition based on recognition rules), the accuracy can reach 92.7%. The model trained on the corresponding travel data of different categories has better accuracy in predicting travel purpose than the model trained on the same data.

[0080] The travel survey data and the prediction method of the present invention are used to predict the proportion of travel purposes, wherein the method of the present invention includes predictions based on bus card swiping data on working days and non-working days, such as Figure 5 As shown, it can be seen that the distributions of the two are basically the same, the proportion of commuting on weekdays is higher than that on non-working days, and the proportion of leisure and entertainment on weekdays is lower than that on non-working days, thus verifying the accuracy and rationality of the method of the present invention.

[0081] Example 3

[0082] 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.

[0083] Multiple components in the device are connected to the I / O interface, including: input units, such as keyboards, mice, etc.; output units, such as various types of displays, speakers, etc.; storage units, such as disks, optical disks, etc.; and communication units, such as network cards, modems, wireless communication transceivers, etc. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunication networks.

[0084] The processing unit performs the various methods and processes described above, such as methods S1 to S4. For example, in some embodiments, methods S1 to S4 may be implemented as a computer software program, which 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 on 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 S4 described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute methods S1 to S4 in any other appropriate manner (e.g., by means of firmware).

[0085] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, 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), complex programmable logic devices (CPLDs), and the like.

[0086] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes 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 the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0087] In the context of the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0088] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for predicting the travel purpose of public transportation users based on multi-source data, characterized in that: The method comprises the following steps: Step S1, establishing feature engineering based on joint analysis of multi-source data of urban residents, where the multi-source data includes travel survey data, bus card swiping data and point of interest data; Step S2, dividing the records of the travel survey data in which the travel mode is conventional public transportation into the first type of travel data and the second type of travel data according to whether both the home address and the workplace address are filled in the records, and using the machine learning model to construct the first model and the second model respectively; Step S3, identifying the home station address and the workplace station address of each user in the bus card swiping data, and classifying the data into the first type of card swiping data and the second type of card swiping data according to whether the home station address and the workplace station address of the user can be identified; Step S4, predicting the travel purpose using the first model and the second model respectively for the card swiping data classified in step S3.

2. The method for predicting the travel purpose of public transportation users based on multi-source data according to claim 1 is characterized in that: The process of establishing feature engineering includes: according to the travel starting point and travel end point of the travel survey data, and the boarding and alighting stations of the bus card swiping data, respectively calculating the proportion of different types of points of interest within the surrounding distance threshold range, as well as the cosine value of the departure time, the cosine value of the arrival time and the travel time as the characteristics of the user's travel.

3. The method for predicting the travel purpose of public transportation users based on multi-source data according to claim 1 is characterized in that: If the home address and work address of the travel survey data mentioned in a certain item are not missing, it will be classified as the first category of travel data, otherwise it will be classified as the second category of travel data.

4. The method for predicting the travel purpose of public transportation users based on multi-source data according to claim 1 is characterized in that: The process of constructing the first model includes: filtering out records of the first type of travel data whose travel purpose is commuting or going home, dividing the remaining data into a training set and a test set, and using a machine learning model to train and generate a first model; The construction process of the second model includes: selecting all second-category travel data, dividing them into a training set and a test set, and using a machine learning model to train and generate a second model.

5. The method for predicting the travel purpose of public transportation users based on multi-source data according to claim 4 is characterized in that: The machine learning model is an XGBoost model.

6. The method for predicting the travel purpose of public transportation users based on multi-source data according to claim 1 is characterized in that: The step S3 of identifying the home site address and workplace site address of each user in the bus card swiping data is specifically as follows: in the bus card swiping data for a period of time, the departure site with the largest number of first trips for the user and a proportion of the number of trips to all days exceeding a first threshold is identified as the user's home site address, and the departure site with the largest number of last trips for many days for the user and a proportion of the number of trips to all days exceeding a second threshold is identified as the user's workplace site address.

7. The method for predicting the travel purpose of public transportation users based on multi-source data according to claim 1 is characterized in that: The process of dividing the card swiping data into the first type and the second type includes: If both the home stop address and the workplace stop address can be identified in a bus travel record, it is classified as the first type of card swiping data, otherwise it is classified as the second type of card swiping data.

8. The method for predicting the travel purpose of public transportation users based on multi-source data according to claim 1 is characterized in that: The step S4 comprises: For the first type of card swiping data, if the distance between the alighting station and the home station address is ≤ the third threshold, it is identified as the purpose of "going home"; if the distance between the alighting station and the workplace station address is ≤ the fourth threshold, it is identified as the purpose of "commuting"; for the remaining first type of card swiping data with unidentified travel purposes, the first model is used to predict the travel purpose; For the second type of card swiping data, the second model is used to predict the travel purpose.

9. The method for predicting the travel purpose of public transportation users based on multi-source data according to claim 1, characterized in that: The method further includes evaluating the prediction results based on multi-dimensional evaluation indicators, including: For travel survey data, the prediction results are evaluated in terms of accuracy, precision, and recall; For bus card swiping data, the Jensen-Shannon divergence of the prediction results is calculated for evaluation.

10. 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 any one of claims 1 to 9 is implemented.

Images