Model training, data processing methods, devices, equipment, storage media and programs

Abstract

The present invention discloses a model training, data processing method, device, equipment, storage medium and program, the method comprising: obtaining multiple track point data corresponding to complete track information, each track point data including a first position category corresponding to the track point, performing interference processing on the first position category in the first track point data to obtain interference track point data, processing the interference track point data and the second track point data by a preset model to be trained to obtain a predicted position category corresponding to the first track point data, and updating the model parameters of the preset model according to the first position category and the predicted position category corresponding to the first track point data. The preset model trained by the present invention has the ability to complete the position category of the track point, so that the trained preset model can be used to complete the incomplete track information to obtain the complete track information.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a model training, data processing method, device, equipment, storage medium and program. Background Art

[0002] In some application scenarios, such as traffic congestion simulation scenarios and disease transmission simulation scenarios, it is necessary to analyze and model user travel characteristics. Travel characteristics can be obtained by analyzing user trajectory information.

[0003] In the related art, the user's trajectory information can be recorded by a terminal device. Specifically, the terminal device can collect the user's location. In this way, the user's trajectory information can be obtained based on the multiple locations of the user collected by the terminal device within a preset time period. For example, the user's trajectory information can be: position A (10 o'clock) -> position B (11 o'clock) -> position C (12 o'clock) -> position D (13 o'clock) -> position E (14 o'clock).

[0004] However, in actual applications, the frequency with which the terminal device records the user's location depends on the user's usage scenario and habits, resulting in the possibility that the trajectory information collected by the terminal device is incomplete. For example, the terminal device only records location A (10 o'clock) -> location E (14 o'clock), and the locations of the intermediate time points (11 o'clock, 12 o'clock, and 13 o'clock) are missing. Therefore, how to complete the incomplete trajectory information is a technical problem that needs to be solved urgently. Summary of the invention

[0005] The main purpose of the present invention is to provide a model training, data processing method, device, equipment, storage medium and program to provide a way to complete incomplete trajectory information.

[0006] In a first aspect, the present invention provides a model training method, the method comprising:

[0007] Acquire multiple track point data corresponding to the complete track information, each track point data includes a first position category corresponding to the track point;

[0008] Performing interference processing on a first position category in first trajectory point data to obtain interference trajectory point data, wherein the first trajectory point data is part of the plurality of trajectory point data;

[0009] The interference trajectory point data and the second trajectory point data are processed by a preset model to be trained to obtain a predicted position category corresponding to the first trajectory point data, wherein the second trajectory point data is trajectory point data other than the first trajectory point data among the plurality of trajectory point data;

[0010] The model parameters of the preset model are updated according to the first position category and the predicted position category corresponding to the first trajectory point data.

[0011] In a possible implementation, the interference trajectory point data and the second trajectory point data are processed by a preset model to be trained to obtain a predicted position category corresponding to the first trajectory point data, including:

[0012] Processing the interference trajectory point data and the second trajectory point data by using the first sub-model in the preset model to obtain a feature vector corresponding to the interference trajectory point data;

[0013] Processing the feature vector by using a second sub-model in the preset model to obtain a predicted position category corresponding to the first trajectory point data;

[0014] Updating the model parameters of the preset model according to the first position category corresponding to the first trajectory point data and the predicted position category, including:

[0015] The model parameters of the first sub-model and the second sub-model are updated according to the first position category and the predicted position category corresponding to the first trajectory point data.

[0016] In a possible implementation, the interference trajectory point data and the second trajectory point data are processed by the first sub-model in the preset model to obtain a feature vector corresponding to the interference trajectory point data, including:

[0017] Determine a position category vector corresponding to the interference trajectory point data;

[0018] Determine a position category vector corresponding to the second trajectory point data;

[0019] The position category vector corresponding to the interference trajectory point data and the position category vector corresponding to the second trajectory point data are input into the first sub-model, so that the first sub-model outputs the feature vector corresponding to the interference trajectory point data.

[0020] In a possible implementation, interference processing is performed on the first position category in the first trajectory point data, including any one of the following:

[0021] Replacing the first location category in the first track point data with preset information;

[0022] Replacing the first position category in the first track point data with other position categories;

[0023] The first position category in the first trajectory point data remains unchanged.

[0024] In a possible implementation, obtaining multiple track point data corresponding to the complete track information includes:

[0025] Acquire first trajectory information corresponding to a preset time period, the first trajectory information including N trajectory point data, where N is an integer greater than 1;

[0026] Determine whether the first trajectory information is complete trajectory information according to the value of N and the distribution of trajectory moments corresponding to the N trajectory point data, and the trajectory moment is the moment when the trajectory point data is collected;

[0027] If yes, the N track point data are determined as the plurality of track point data;

[0028] If not, the N track point data are supplemented to obtain the plurality of track point data.

[0029] In a possible implementation, the N track point data are supplemented to obtain the plurality of track point data, including:

[0030] Obtaining the trajectory time corresponding to the N trajectory point data;

[0031] Determine, according to the trajectory moments corresponding to the N trajectory point data, a trajectory moment corresponding to at least one trajectory point data to be supplemented;

[0032] Generate the at least one track point data to be supplemented according to the track time corresponding to the N track point data and the at least one track point data to be supplemented;

[0033] The plurality of trajectory point data are obtained according to the N trajectory point data and the at least one trajectory point data to be supplemented.

[0034] In a possible implementation, generating the at least one track point data to be supplemented according to the track time corresponding to the N track point data and the at least one track point data to be supplemented includes:

[0035] For any track point data to be supplemented, determine the target track point data from the N track point data according to the track time corresponding to the track point data to be supplemented, and the time difference between the track time corresponding to the target track point data and the track time corresponding to the track point data to be supplemented is the smallest;

[0036] The target track point data is determined as the track point data to be supplemented.

[0037] In a possible implementation, updating the model parameters of the preset model according to the first position category and the predicted position category corresponding to the first trajectory point data includes:

[0038] Determining a loss function according to the first position category corresponding to the first trajectory point data and the predicted position category;

[0039] The model parameters of the preset model are updated through the loss function.

[0040] In a possible implementation, after updating the model parameters of the preset model according to the first location category and the predicted location category corresponding to the first trajectory point data, the method further includes:

[0041] Determine whether the updated preset model converges;

[0042] If yes, the updated preset model is used as the trained model;

[0043] If not, the preset model is repeatedly trained until the updated preset model converges.

[0044] In a possible implementation, each track point data further includes at least one of the following: time information corresponding to the track point, a position corresponding to the track point, an area to which the position corresponding to the track point belongs, and coordinate information corresponding to the track point.

[0045] In a second aspect, the present invention provides a data processing method, comprising:

[0046] Acquire incomplete trajectory information, where the incomplete trajectory information includes M trajectory point data, each trajectory point data includes a first position category corresponding to the trajectory point, where M is an integer greater than 1 and less than K, and K is the number of trajectory point data included in the complete trajectory information;

[0047] The M trajectory point data are processed by a trained preset model to obtain K trajectory point data; wherein the preset model is trained by the above data processing method.

[0048] In a third aspect, the present invention provides a model training device, comprising:

[0049] An acquisition module, used to acquire a plurality of trajectory point data corresponding to the complete trajectory information, each trajectory point data including a first position category corresponding to the trajectory point;

[0050] a processing module, configured to perform interference processing on a first position category in first trajectory point data to obtain interference trajectory point data, wherein the first trajectory point data is part of the plurality of trajectory point data;

[0051] The processing module is further used to process the interference trajectory point data and the second trajectory point data through a preset model to be trained to obtain a predicted position category corresponding to the first trajectory point data, and the second trajectory point data is trajectory point data other than the first trajectory point data among the multiple trajectory point data;

[0052] An updating module is used to update the model parameters of the preset model according to the first position category and the predicted position category corresponding to the first trajectory point data.

[0053] In a fourth aspect, the present invention provides a data processing device, comprising:

[0054] an acquisition module, configured to acquire incomplete trajectory information, wherein the incomplete trajectory information includes M trajectory point data, each trajectory point data includes a first position category corresponding to the trajectory point, wherein M is an integer greater than 1 and less than K, and K is the number of trajectory point data included in the complete trajectory information;

[0055] A processing module is used to process the M trajectory point data through a trained preset model to obtain K trajectory point data; wherein the preset model is trained by the method described in any one of the first negative aspects.

[0056] In a fifth aspect, the present invention provides an electronic device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the steps of the model training method as described in any one of the first aspect are implemented, or the steps of the data processing method as described in the second aspect are implemented.

[0057] In a sixth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the model training method as described in any one of the first aspects are implemented, or the steps of the data processing method as described in the second aspect are implemented.

[0058] In the seventh aspect, the present invention provides a computer program product, comprising: a computer program, which, when executed by a processor, implements the steps of the model training method as described in any one of the first aspects, or implements the steps of the data processing method as described in the second aspect.

[0059] The model training, data processing method, device, equipment, storage medium and program provided by the present invention can obtain multiple trajectory point data corresponding to the complete trajectory information, each trajectory point data includes a first position category corresponding to the trajectory point, and the first position category in the first trajectory point data is interfered to obtain the interfered trajectory point data, the first trajectory point data is part of the trajectory point data in the multiple trajectory point data, the interfered trajectory point data and the second trajectory point data are processed by the preset model to be trained to obtain the predicted position category corresponding to the first trajectory point data, the second trajectory point data is the trajectory point data in the multiple trajectory point data except the first trajectory point data, and the model parameters of the preset model are updated according to the first position category and the predicted position category corresponding to the first trajectory point data. Through the above training process, the preset model obtained has the ability to complete the position category of the trajectory point, so that the trained preset model can be used to complete the incomplete trajectory information to obtain the complete trajectory information. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 A schematic diagram of a system architecture applicable to an embodiment of the present invention;

[0061] Figure 2A A schematic diagram of complete trajectory information provided by an embodiment of the present invention;

[0062] Figure 2B A schematic diagram of incomplete trajectory information provided by an embodiment of the present invention;

[0063] Figure 3 A flowchart of a model training method provided by an embodiment of the present invention;

[0064] Figure 4 A schematic diagram of a model processing process provided by an embodiment of the present invention;

[0065] Figure 5 A flowchart of another model training method provided by an embodiment of the present invention;

[0066] Figure 6 A schematic diagram of another model training process provided by an embodiment of the present invention;

[0067] Figure 7 A flowchart of a data processing method provided by an embodiment of the present invention;

[0068] Figure 8 A schematic diagram of the structure of a model training device provided by an embodiment of the present invention;

[0069] Fig. 9 A schematic diagram of the structure of a data processing device provided by an embodiment of the present invention;

[0070] Fig.10 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention.

[0071] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0072] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to enable the scope of the present invention to be fully communicated to those skilled in the art.

[0073] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0074] User trajectory information can be used as a data source for analyzing travel characteristics. However, in actual applications, the user trajectory information collected by terminal devices is often incomplete. If travel characteristics are extracted based on incomplete trajectory information, the accuracy of travel characteristics will be affected. Therefore, how to complete incomplete trajectory information is a technical problem that needs to be solved urgently.

[0075] In order to facilitate understanding of the technical solution of the present invention, firstly Figure 1 The system architecture involved in the embodiments of the present invention is introduced.

[0076] Figure 1 FIG. 1 is a schematic diagram of a system architecture applicable to an embodiment of the present invention. Figure 1 As shown, the data collection device is used to collect the user's trajectory information. Generally, the data collection device is a device that moves with the user. Exemplary data collection devices include but are not limited to: smart phones, laptops, tablet computers, smart wearable devices, vehicle-mounted devices, etc.

[0077] The data collection device has a global positioning system (GPS) function. The data collection device can collect data with GPS coordinates and time information in real time or at preset time intervals or according to the user's configuration information. By matching the GPS coordinates with the corresponding positions, the GPS coordinates can be converted into the corresponding stop positions. The stop positions can also be called points of interest or track points. The collected track points are arranged in chronological order to form the user's track information.

[0078] After the data collection device collects the user's trajectory information, it stores the trajectory information in the trajectory information database. The training device can learn and model multiple trajectory information in the trajectory information database to obtain a trajectory supplement model. The trajectory supplement model has the ability to complete incomplete trajectory information.

[0079] The trajectory supplement model can be deployed in the execution device. The incomplete trajectory information is input into the execution device, and the execution device uses the trajectory supplement model to complete the incomplete trajectory information to obtain the complete trajectory information. It should be understood that the execution device can be a terminal device, such as a smart phone, a tablet computer, a laptop computer, a vehicle-mounted device, etc. The execution device can also be a server or a cloud, etc.

[0080] It should be noted that in actual applications, the trajectory information stored in the database does not necessarily come from the data acquisition device, but can also be received from other devices. It should also be noted that the training device does not necessarily build the model based entirely on the trajectory information in the database, but can also obtain trajectory information from the cloud or other places to build the model. Figure 1 The system architecture shown should not be regarded as limiting the embodiments of the present invention.

[0081] In the technical solution of the present invention, all the acquisition, storage and application of user trajectory information involved comply with the provisions of relevant laws and regulations, take necessary confidentiality measures, and do not violate public order and good morals.

[0082] The trajectory information in the embodiment of the present invention adopts a relatively sparse data format. A trajectory includes multiple trajectory points arranged in chronological order. There is a certain time interval between adjacent trajectory points. For example, the time interval can be 1 hour, or 30 minutes, or 2 hours, etc.

[0083] Accordingly, each track information includes multiple track point data. Each track point data is used to describe the relevant information of the user's stop location at a certain moment. Exemplarily, each track point data may include time information, location, and relevant description of the location. Among them, the relevant description of a location may include: the location category corresponding to the location (for example: shopping mall, residence, industrial area, restaurant, etc.), the administrative area to which the location belongs (for example, city, district, etc.), and the coordinate information corresponding to the location (for example, longitude, latitude, etc.).

[0084] In practical applications, the density of trajectory points in user travel trajectories varies greatly, which may make the collected trajectory information incomplete.

[0085] In this embodiment, complete trajectory information means that each preset time in a preset period corresponds to trajectory point data. Incomplete trajectory information means that part of the preset time in a preset period corresponds to trajectory point data, while the other part of the preset time does not correspond to trajectory point data. For example, the preset period can be 24 hours, 12 hours, etc. The interval between adjacent preset moments can be 1 hour, 30 minutes, etc. Figure 2A and Figure 2B Examples are given for complete trajectory information and incomplete trajectory information.

[0086] Figure 2A A schematic diagram of complete trajectory information provided by an embodiment of the present invention. Assume that the preset time period is 12 hours, for example, from 6:00 to 18:00, and the preset time is every hour. Figure 2A The complete trajectory information includes the trajectory point data corresponding to 6 o'clock, 7 o'clock, 8 o'clock, 9 o'clock, 10 o'clock, 11 o'clock, 12 o'clock, 13 o'clock, 14 o'clock, 15 o'clock, 16 o'clock, 17 o'clock, and 18 o'clock. Each trajectory point data includes: the time information corresponding to the trajectory point, the location corresponding to the trajectory point, the area to which the location corresponding to the trajectory point belongs, and the coordinate information corresponding to the trajectory point.

[0087] Figure 2B A schematic diagram of incomplete trajectory information provided by an embodiment of the present invention. Figure 2B The incomplete trajectory information includes the trajectory point data corresponding to 6 o'clock, 8 o'clock, 10 o'clock, 15 o'clock, 16 o'clock, 17 o'clock, and 18 o'clock, but lacks the trajectory point data corresponding to 7 o'clock, 9 o'clock, 11 o'clock, 12 o'clock, 13 o'clock, and 14 o'clock. Each trajectory point data includes: the time information corresponding to the trajectory point, the location corresponding to the trajectory point, the area to which the location corresponding to the trajectory point belongs, and the coordinate information corresponding to the trajectory point.

[0088] based on Figure 1The system architecture shown in the figure, the embodiment of the present invention provides a model training method and a data processing method. Among them, the model training method can be executed by a training device, and the data processing method can be executed by an execution device. In some application scenarios, the training device and the execution device can be the same device. In other application scenarios, the training device and the execution device can be different devices.

[0089] In the technical solution of the present invention, a trajectory supplement model can be obtained by learning and training multiple trajectory point data corresponding to the complete trajectory information, so that the trajectory supplement model has the ability to complete the incomplete trajectory information. Furthermore, the trajectory supplement model can be used to complete the incomplete trajectory information to obtain the complete trajectory information.

[0090] It should be noted that in the embodiment of the present invention, since the positions of the trajectory points are relatively sparse, and in practical applications, the position categories corresponding to the positions of the trajectory points are often of concern, the embodiment of the present invention only describes the completion of the position categories of the trajectory points as an example. Accordingly, the trajectory supplement model trained by the embodiment of the present invention has the ability to complete the position categories of the trajectory points. It should be understood that when it is necessary to complete other information of the trajectory points, a similar method can be used, and the embodiment of the present invention will not be described in detail.

[0091] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0092] Figure 3 1 is a flow chart of a model training method provided by an embodiment of the present invention. The method of this embodiment can be executed by a training device. After the training process of this embodiment, a trajectory supplement model can be obtained. Figure 3 As shown, the method of this embodiment may include:

[0093] S301: Acquire multiple track point data corresponding to the complete track information, each track point data includes a first position category corresponding to the track point.

[0094] In this embodiment, the multiple track point data corresponding to the complete track information can be used to train the preset model (i.e., the track supplement model). Figure 2A The data of each track point in .

[0095] In the embodiment of the present invention, it is necessary to enable the trained preset model to have the ability to complete the location category. Therefore, each trajectory point data involved in the training includes the first location category corresponding to the trajectory point. The first location category can also be called the real location category or the actual location category, which is used to distinguish it from the subsequent predicted location category.

[0096] Optionally, each track point data may further include at least one of the following: time information corresponding to the track point, location corresponding to the track point, region to which the location corresponding to the track point belongs, coordinate information corresponding to the track point, etc. It should be understood that when the track point data involved in the training contain more track point related information, the preset model obtained by training has a stronger ability to complete the location category.

[0097] S302: performing interference processing on a first position category in first trajectory point data to obtain interference trajectory point data, where the first trajectory point data is part of the plurality of trajectory point data.

[0098] Specifically, some of the track point data in the plurality of track point data may be determined as the first track point data. The number of the first track point data may be one or more. Further, interference processing is performed on the first position category in each first track point data to obtain interference track point data.

[0099] The interference processing refers to adding disturbance to the first position category in the first trajectory point data. Optionally, the interference processing of the first position category in the first trajectory point data may include any one of the following:

[0100] (1) The first position category in the first track point data is replaced with preset information.

[0101] Exemplarily, the preset information may be a special symbol. The special symbol may be a predefined vector or a randomly initialized vector. For example, when performing interference processing, the first position category in the first trajectory point data may be replaced with the symbol [MASK].

[0102] (2) Replacing the first position category in the first trajectory point data with other position categories.

[0103] For example, assuming that the location categories include: residence, shopping mall, restaurant, industrial area, and scenic spot, the first location category in the first trajectory point data is shopping mall, and when performing interference processing, the first location category in the first trajectory point data can be replaced with any one of residence, restaurant, industrial area, and scenic spot.

[0104] (3) The first position category in the first trajectory point data remains unchanged.

[0105] In a possible implementation, the above three interference processing methods can be used in combination. For example, in 80% of cases, method (1) can be used for interference processing, in 10% of cases, method (2) can be used for interference processing, and in the remaining 10% of cases, method (3) can be used for interference processing. By adopting different interference processing methods, the ability of the preset model to complete the location category can be improved.

[0106] S303: Processing the interference trajectory point data and the second trajectory point data through a preset model to be trained to obtain a predicted position category corresponding to the first trajectory point data, wherein the second trajectory point data is trajectory point data among the multiple trajectory point data except the first trajectory point data.

[0107] In this embodiment, since the first position category in the first trajectory point data has been scrambled, the preset model can predict the position category of the first trajectory point data based on the interference trajectory point data and the second trajectory point data to obtain the predicted position category corresponding to the first trajectory point data.

[0108] The preset model may be any network model based on machine learning. The preset model may be a single model or may be composed of multiple sub-models. This embodiment does not limit the structure of the preset model.

[0109] Figure 4 A schematic diagram of a model processing process provided by an embodiment of the present invention. Figure 4 See Figure 4 , assuming that the complete trajectory information includes k trajectory point data, the first position category in the first trajectory point data is E 1 , the first position category in the second trajectory point data is E 2 , ..., the first position category in the kth trajectory point data is E k . The jth trajectory point data is determined as the first trajectory point data, that is, the first position category (E j ) in the jth trajectory point data is scrambled.

[0110] Continue to see Figure 4 , the scrambled j-th trajectory point data (i.e., the interference trajectory point data) and other unscrambled trajectory point data (i.e., the second trajectory point data) are input into the preset model, so that the preset model predicts the position category of the j-th trajectory point data according to the scrambled j-th trajectory point data and other unscrambled trajectory point data, and obtains the predicted position category (i.e., E′ j ).

[0111] S304: updating the model parameters of the preset model according to the first position category and the predicted position category corresponding to the first trajectory point data.

[0112] Exemplarily, a loss function may be determined based on the first position category and the predicted position category corresponding to the first trajectory point data; and the model parameters of the preset model may be updated using the loss function.

[0113] Optionally, a cross entropy may be calculated based on the first position category and the predicted position category, and a loss function may be determined based on the cross entropy. Further, based on the loss function, a stochastic gradient descent method may be used to update the model parameters of the preset model.

[0114] It should be understood that in this embodiment, S301 to S304 are described by taking one iteration process as an example. In practical applications, the training of the preset model may include multiple iteration processes.

[0115] Exemplarily, after each update of the model parameters of the preset model, it is determined whether the updated preset model has converged. If so, the training ends and the updated preset model is used as the trained model. If not, the preset model is repeatedly trained (i.e., S301 to S304 are repeatedly executed) until the updated preset model converges.

[0116] The model training method provided in this embodiment includes: obtaining multiple trajectory point data corresponding to complete trajectory information, each trajectory point data includes a first position category corresponding to the trajectory point, performing interference processing on the first position category in the first trajectory point data to obtain interference trajectory point data, the first trajectory point data is part of the trajectory point data in the multiple trajectory point data, processing the interference trajectory point data and the second trajectory point data through the preset model to be trained to obtain the predicted position category corresponding to the first trajectory point data, the second trajectory point data is the trajectory point data in the multiple trajectory point data except the first trajectory point data, and updating the model parameters of the preset model according to the first position category and the predicted position category corresponding to the first trajectory point data. Through the above training process, the preset model obtained has the ability to complete the position category of the trajectory point, so that the trained preset model can be used to complete the incomplete trajectory information to obtain complete trajectory information.

[0117] Based on the above embodiment, the model training method provided by the present invention is described in more detail below in conjunction with a more specific embodiment. Figure 3 and Figure 4 The structure and processing process of the preset model are refined on the basis of the illustrated embodiment.

[0118] Figure 5 A flow chart of another model training method provided by an embodiment of the present invention. Figure 5 As shown, the method of this embodiment includes:

[0119] S501: Acquire multiple track point data corresponding to the complete track information, each track point data includes a first position category corresponding to the track point.

[0120] S502: performing interference processing on a first position category in first trajectory point data to obtain interference trajectory point data, where the first trajectory point data is part of the plurality of trajectory point data.

[0121] It should be understood that the specific implementation of S501 and S502 is different from Figure 3 S301 and S302 are similar and are not described here.

[0122] S503: Processing the interference trajectory point data and the second trajectory point data through the first sub-model in the preset model to obtain a feature vector corresponding to the interference trajectory point data, wherein the second trajectory point data is the trajectory point data among the multiple trajectory point data except the first trajectory point data.

[0123] S504: Processing the feature vector by using a second sub-model in a preset model to obtain a predicted position category corresponding to the first trajectory point data.

[0124] In this embodiment, the preset model may include a first sub-model and a second sub-model. The first sub-model is configured to obtain a feature vector corresponding to the interference trajectory point data according to the interference trajectory point data and the second trajectory point data. The second sub-model is configured to predict the predicted position category corresponding to the first trajectory point data according to the feature vector corresponding to the interference trajectory point data.

[0125] Optionally, the first sub-model may be a Bert model.

[0126] Optionally, the second sub-model may be a classification model. For example, the second sub-model may be a classification model based on Bayesian, a classification model based on a decision tree, a classification model based on a neural network, etc.

[0127] In one possible implementation, the position category vector corresponding to the interference trajectory point data can be determined, and the position category vector corresponding to the second trajectory point data can be determined. Then, the position category vector corresponding to the interference trajectory point data and the position category vector corresponding to the second trajectory point data are input into the first sub-model so that the first sub-model outputs the feature vector corresponding to the interference trajectory point data.

[0128] Furthermore, the feature vector corresponding to the interference trajectory point data is input into the second sub-model, so that the second sub-model outputs the predicted position category corresponding to the first trajectory point data.

[0129] S505: Update model parameters of the first sub-model and the second sub-model according to the first position category and the predicted position category corresponding to the first trajectory point data.

[0130] In this embodiment, the first sub-model and the second sub-model can be trained as a whole. That is, in each iteration process, the model parameters of the first sub-model and the second sub-model are synchronously updated.

[0131] Combine the following Figure 6 , the model training method of this embodiment is illustrated by example.

[0132] Figure 6 A schematic diagram of another model training process provided by an embodiment of the present invention. In this embodiment, an example is given in which the first sub-model is a Bert model and the second sub-model is a classification model. Figure 6 Take an iterative process as an example. P 1 represents the first trajectory point data, P 2 represents the second trajectory point data, ..., P k represents the kth trajectory point data. Assume that each trajectory point data includes: the area to which the location corresponding to the trajectory point belongs, the location category corresponding to the trajectory point, and the time information corresponding to the trajectory point. The time information includes date and time.

[0133] First, for each trajectory point data, the information in the trajectory point data is vectorized so as to serve as the input of the BERT model.

[0134] For example, since the location categories in the trajectory point data are usually represented by words, for example, the location categories are "residential", "commercial", "industrial area", etc. Therefore, for the location categories in the trajectory point data, the word-to-vector (word2vec) method can be used for vectorization to obtain the location category vector. Figure 6 In , E 1 represents the position category vector corresponding to the first trajectory point data, E 2 represents the position category vector corresponding to the second trajectory point data, ..., E k represents the position category vector corresponding to the kth trajectory point data.

[0135] The region to which the location in the trajectory point data belongs is usually represented by a word, for example, the region to which the location belongs is "Shenzhen", "Beijing", etc. Therefore, for the region to which the location in the trajectory point data belongs, the word-to-vector (word2vec) method can also be used for vectorization to obtain a region vector. Figure 6 In , C 1 represents the region vector corresponding to the first trajectory point data, C 2 represents the region vector corresponding to the second trajectory point data, ..., C k represents the region vector corresponding to the kth trajectory point data.

[0136] For the dates in the trajectory point data, we can distinguish between working days and non-working days. For example, working days are 0 and non-working days are 1, thus obtaining a date vector. Figure 6 In , D 1 represents the date vector corresponding to the first trajectory point data, D 2 represents the date vector corresponding to the second trajectory point data, ..., D k represents the date vector corresponding to the kth trajectory point data.

[0137] For the moments in the data point data (such as 6 o'clock, 7 o'clock, etc.), a trainable random initialization vector can be used, or the trigonometric function position encoding in the attention mechanism can be used to obtain the time vector. Figure 6 In , T 1 represents the time vector corresponding to the first trajectory point data, T 2 represents the time vector corresponding to the second trajectory point data, ..., T k represents the time vector corresponding to the kth trajectory point data.

[0138] After the above processing of the trajectory point data, four types of vectors are obtained, namely, location category vector, region vector, date vector and time vector. Figure 6 , the above four types of vectors can be used as input of the Bert model.

[0139] In a possible implementation, the above four types of vectors may be added, and the added vectors may be input into the Bert model.

[0140] In another possible implementation, the above four types of vectors may be concatenated, and the concatenated vectors may be input into the Bert model.

[0141] It should be noted that if the trajectory information also includes other more information, such as the location of the trajectory point, the user's residence, the workplace, the distance between the trajectory point and the residence, etc., the above method can be used to construct a vector and use it as the input of the Bert model.

[0142] Continue to see Figure 6 Before the trajectory point data is input into the Bert model, one or more first trajectory point data may be randomly determined, and interference processing may be performed on the position category vector corresponding to the first trajectory point data to obtain interference trajectory point data. Figure 6 In the example, the interference processing of the position category vector corresponding to the j-th trajectory point data is taken as an example.

[0143] Continue to see Figure 6 , the j-th trajectory point data after interference and other uninterrupted trajectory point data are input into the Bert model, so that the Bert model obtains the feature vector corresponding to the j-th trajectory point data after interference according to the j-th trajectory point data after interference and other uninterrupted trajectory point data (that is, Figure 6 The eigenvector O j in ). Figure 6As shown, the Bert model also outputs feature vectors corresponding to other uninterrupted trajectory point data (i.e., Figure 6 In this embodiment, the Bert model only needs to convert the feature vector corresponding to the j-th trajectory point data after interference (ie, Figure 6 The feature vector O j in is input into the classification model, so that the classification model predicts the predicted position category corresponding to the j-th trajectory point data according to the feature vector.

[0144] Furthermore, according to the real location category (i.e., the first location category) and the predicted location category corresponding to the jth trajectory point data, the cross entropy is obtained and the cross entropy is used as the loss function. The model parameters of the Bert model and the classification model are updated using the stochastic gradient descent method.

[0145] Repeat the above iterative process until the Bert model and classification model converge and the training ends.

[0146] In this embodiment, since the Bert model considers the contextual relationship between different trajectory point data when extracting feature vectors, the feature vector corresponding to the interference trajectory point data output by the Bert model is more accurate, thereby ensuring the accuracy of the predicted location category. In addition, in this embodiment, when training the preset model, various information such as the location category, the region to which the location belongs, the date, and the time in the trajectory point data are used to further improve the accuracy of the predicted location category.

[0147] Above Figures 3 to 6 The illustrated embodiment focuses on the model training process. In the above model training process, multiple track point data corresponding to the complete track information are required. Based on any of the above embodiments, the following describes how to obtain multiple track point data corresponding to the complete track information in combination with a specific embodiment. This embodiment can be used as a possible implementation of S301 and S501.

[0148] In this embodiment, the user's historical travel data can be obtained, and the historical travel data includes the user's track point data. The user's historical travel data is segmented based on the preset time period, and the track information corresponding to each preset time period is obtained. For example, the division point can be 0:00 a.m. every day, or 4:00 a.m. every day, etc.

[0149] For each preset time period after segmentation, obtain the first trajectory information corresponding to the preset time period. Assume that the first trajectory information includes N trajectory point data, where N is an integer greater than 1. According to the value of N and the distribution of the trajectory moments corresponding to the N trajectory point data, it can be determined whether the first trajectory information is complete trajectory information. The trajectory moment is the moment when the trajectory point data is collected.

[0150] For example, assuming that the preset time period is 24 hours, if the value of N is greater than or equal to 24, and each hour corresponds to trajectory point data (for example, each hour corresponds to trajectory point data), then the first trajectory information is determined to be complete trajectory information.

[0151] If the value of N is less than 24, it means that there is no trajectory point data in some hours, and therefore, the first trajectory information is determined to be incomplete trajectory information.

[0152] If the value of N is greater than or equal to 24, but the track time distribution of the N track point data is uneven, for example, some hours correspond to multiple track point data, while some hours correspond to no track point data. Therefore, in this case, the first track information is determined to be incomplete track information.

[0153] Further, when it is determined that the first trajectory information is complete trajectory information, the N trajectory point data in the first trajectory information can be determined as multiple trajectory point data corresponding to the complete trajectory information. When it is determined that the first trajectory information is incomplete trajectory information, the N trajectory point data in the first trajectory information can be completed to obtain multiple trajectory point data corresponding to the complete trajectory information.

[0154] In a possible implementation, the following completion method can be used:

[0155] Obtain the trajectory moments corresponding to the N trajectory point data; determine the trajectory moment corresponding to at least one trajectory point data to be supplemented according to the trajectory moments corresponding to the N trajectory point data; generate the at least one trajectory point data to be supplemented according to the trajectory moments corresponding to the N trajectory point data and the at least one trajectory point data to be supplemented; obtain the multiple trajectory point data according to the N trajectory point data and the at least one trajectory point data to be supplemented.

[0156] Optionally, for each trajectory point data to be supplemented, target trajectory point data is determined among the N trajectory point data according to the trajectory moment corresponding to the trajectory point data to be supplemented, and the time difference between the trajectory moment corresponding to the target trajectory point data and the trajectory moment corresponding to the trajectory point data to be supplemented is the smallest; the target trajectory point data is determined as the trajectory point data to be supplemented.

[0157] For example, assuming that among the N track point data, except for the track point data corresponding to 12 o'clock, the track point data corresponding to other whole points exist, then it can be determined that the track time corresponding to the track point data to be supplemented is 12 o'clock. In this case, the track point data corresponding to 11 o'clock can be filled as the track point data corresponding to 12 o'clock, or the track point data corresponding to 13 o'clock can be filled as the track point data corresponding to 12 o'clock. Thus, the filled track point data are used as multiple track point data corresponding to the complete track information.

[0158] It should be noted that the method for completing the trajectory point data provided in this embodiment is applicable to the case where the number of missing trajectory point data is small, that is, it is applicable to completing relatively complete trajectory information. For example, only one or two trajectory point data are missing in the trajectory information, or only the trajectory point data at night is missing in the trajectory information. In this case, the above-mentioned neighbor filling method can be used to complete the trajectory point data, so that the completed complete trajectory information can be used to train the preset model.

[0159] For trajectory information with a large number of missing trajectory point data, the trained preset model can be used to complete it. Figure 7 The illustrated embodiment describes a process of completing trajectory information using a preset model.

[0160] Figure 7 1 is a flow chart of a data processing method provided by an embodiment of the present invention. The method of this embodiment can be executed by an execution device. Figure 7 As shown, the method of this embodiment includes:

[0161] S701: Acquire incomplete trajectory information, where the incomplete trajectory information includes M trajectory point data, each trajectory point data includes a first position category corresponding to the trajectory point, where M is an integer greater than 1 and less than K, and K is the number of trajectory point data included in the complete trajectory information.

[0162] S702: Processing the M trajectory point data using a trained preset model to obtain K trajectory point data.

[0163] Among them, the preset model can be trained using the model training method provided by any of the above embodiments.

[0164] Optionally, each track point data may further include at least one of the following: time information corresponding to the track point, a position corresponding to the track point, an area to which the position corresponding to the track point belongs, and coordinate information corresponding to the track point.

[0165] by Figure 6The preset model shown is used as an example for illustration. The M trajectory point data are input into the Bert model, and the Bert model can output the feature vector corresponding to the missing data point data. The Bert model inputs the feature vector into the classification model, and the classification model predicts the predicted location category corresponding to the missing data point data. In this way, the predicted location category output by the classification model can be filled into the incomplete trajectory information, so that each trajectory point in the trajectory information corresponds to a location category, thereby obtaining complete trajectory information.

[0166] In this embodiment, the incomplete trajectory information is processed by using a trained classification model to obtain the complete trajectory information, thereby completing the incomplete trajectory information.

[0167] Figure 8 Schematic diagram of the structure of a model training device provided by an embodiment of the present invention. The device of this embodiment can be in the form of software and / or hardware. Figure 8 As shown, the model training device 800 provided in this embodiment may include: an acquisition module 801, a processing module 802 and an update module 803.

[0168] The acquisition module 801 is used to acquire multiple track point data corresponding to the complete track information, each track point data includes a first position category corresponding to the track point;

[0169] A processing module 802 is used to perform interference processing on a first position category in first trajectory point data to obtain interference trajectory point data, where the first trajectory point data is part of the plurality of trajectory point data;

[0170] The processing module 802 is further used to process the interference trajectory point data and the second trajectory point data through a preset model to be trained to obtain a predicted position category corresponding to the first trajectory point data, and the second trajectory point data is trajectory point data other than the first trajectory point data among the multiple trajectory point data;

[0171] The updating module 803 is used to update the model parameters of the preset model according to the first position category and the predicted position category corresponding to the first trajectory point data.

[0172] In a possible implementation, the processing module 802 is specifically configured to:

[0173] Processing the interference trajectory point data and the second trajectory point data by using the first sub-model in the preset model to obtain a feature vector corresponding to the interference trajectory point data;

[0174] Processing the feature vector by using a second sub-model in the preset model to obtain a predicted position category corresponding to the first trajectory point data;

[0175] The updating module 803 is specifically used to update the model parameters of the first sub-model and the second sub-model according to the first position category and the predicted position category corresponding to the first trajectory point data.

[0176] In a possible implementation, the processing module 802 is specifically configured to:

[0177] Determine a position category vector corresponding to the interference trajectory point data;

[0178] Determine a position category vector corresponding to the second trajectory point data;

[0179] The position category vector corresponding to the interference trajectory point data and the position category vector corresponding to the second trajectory point data are input into the first sub-model, so that the first sub-model outputs the feature vector corresponding to the interference trajectory point data.

[0180] In a possible implementation, the processing module 802 is specifically configured to perform any one of the following:

[0181] Replacing the first location category in the first track point data with preset information;

[0182] Replacing the first position category in the first track point data with other position categories;

[0183] The first position category in the first trajectory point data remains unchanged.

[0184] In a possible implementation, the acquisition module 801 is specifically used for:

[0185] Acquire first trajectory information corresponding to a preset time period, the first trajectory information including N trajectory point data, where N is an integer greater than 1;

[0186] Determine whether the first trajectory information is complete trajectory information according to the value of N and the distribution of trajectory moments corresponding to the N trajectory point data, and the trajectory moment is the moment when the trajectory point data is collected;

[0187] If yes, the N track point data are determined as the plurality of track point data;

[0188] If not, the N track point data are supplemented to obtain the plurality of track point data.

[0189] In a possible implementation, the acquisition module 801 is specifically used for:

[0190] Obtaining the trajectory time corresponding to the N trajectory point data;

[0191] Determine, according to the trajectory moments corresponding to the N trajectory point data, a trajectory moment corresponding to at least one trajectory point data to be supplemented;

[0192] Generate the at least one track point data to be supplemented according to the track time corresponding to the N track point data and the at least one track point data to be supplemented;

[0193] The plurality of trajectory point data are obtained according to the N trajectory point data and the at least one trajectory point data to be supplemented.

[0194] In a possible implementation, the acquisition module 801 is specifically used for:

[0195] For any track point data to be supplemented, determine the target track point data from the N track point data according to the track time corresponding to the track point data to be supplemented, and the time difference between the track time corresponding to the target track point data and the track time corresponding to the track point data to be supplemented is the smallest;

[0196] The target track point data is determined as the track point data to be supplemented.

[0197] In a possible implementation, the updating module 803 is specifically used for:

[0198] Determining a loss function according to the first position category corresponding to the first trajectory point data and the predicted position category;

[0199] The model parameters of the preset model are updated through the loss function.

[0200] In a possible implementation, the processing module 802 is further configured to:

[0201] Determine whether the updated preset model converges;

[0202] If yes, the updated preset model is used as the trained model;

[0203] If not, the preset model is repeatedly trained until the updated preset model converges.

[0204] In a possible implementation, each track point data further includes at least one of the following: time information corresponding to the track point, a position corresponding to the track point, an area to which the position corresponding to the track point belongs, and coordinate information corresponding to the track point.

[0205] The model training device provided in this embodiment can be used to execute the model training method provided in any of the above method embodiments. Its implementation principles and technical effects are similar and will not be elaborated here.

[0206] Fig. 9Schematic diagram of the structure of a data processing device provided by an embodiment of the present invention. The device of this embodiment can be in the form of software and / or hardware. Fig. 9 As shown, the data processing device 900 provided in this embodiment may include: an acquisition module 901 and a processing module 902 .

[0207] The acquisition module 901 is used to acquire incomplete trajectory information, where the incomplete trajectory information includes M trajectory point data, each trajectory point data includes a first position category corresponding to the trajectory point, where M is an integer greater than 1 and less than K, and K is the number of trajectory point data included in the complete trajectory information;

[0208] The processing module 902 is used to process the M trajectory point data through a trained preset model to obtain K trajectory point data; wherein the preset model is trained by the model training method in any of the above method embodiments.

[0209] The data processing device provided in this embodiment can be used to execute the data processing method in any of the above method embodiments. Its implementation principle and technical effects are similar and will not be described in detail here.

[0210] Fig.10 The electronic device of this embodiment can be used as a training device or an execution device. Fig.10 As shown, the electronic device 1000 provided in this embodiment includes: a memory 1001, a processor 1002, and a computer program stored in the memory 1001 and executable on the processor 1002. When the computer program is executed by the processor 1002, the steps of the model training method described in any of the foregoing embodiments are implemented, or the steps of the data processing method described in any of the foregoing embodiments are implemented.

[0211] Optionally, the memory 1001 may be independent or integrated with the processor 1002 .

[0212] The implementation principle and technical effects of the device provided in this embodiment can be found in the aforementioned embodiments and will not be described in detail here.

[0213] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the model training method described in any of the foregoing embodiments are implemented, or the steps of the data processing method described in any of the foregoing embodiments are implemented.

[0214] An embodiment of the present invention also provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the steps of the model training method described in any of the aforementioned embodiments, or implements the steps of the data processing method described in any of the aforementioned embodiments.

[0215] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation, for example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0216] The above-mentioned integrated module implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform some steps of the method described in each embodiment of the present invention.

[0217] It should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the invention may be directly implemented as being executed by a hardware processor, or may be implemented by a combination of hardware and software modules in the processor.

[0218] The memory may include a high-speed RAM memory, and may also include a non-volatile storage NVM, such as at least one disk memory, and may also be a USB flash drive, a mobile hard disk, a read-only memory, a magnetic disk or an optical disk, etc.

[0219] The above storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The storage medium can be any available medium that can be accessed by a general or special purpose computer.

[0220] An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the storage medium can also exist as discrete components in an electronic device or a main control device.

[0221] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0222] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0223] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0224] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A model training method, characterized in that: include: Acquire multiple track point data corresponding to the complete track information, wherein each track point data further includes at least one of the following: time information corresponding to the track point, a position corresponding to the track point, an area to which the position corresponding to the track point belongs, and coordinate information corresponding to the track point, and each track point data includes a first position category corresponding to the track point; Performing interference processing on a first position category in first trajectory point data to obtain interference trajectory point data, wherein the first trajectory point data is part of the plurality of trajectory point data; The interference trajectory point data and the second trajectory point data are processed by a preset model to be trained to obtain a predicted position category corresponding to the first trajectory point data, wherein the second trajectory point data is trajectory point data other than the first trajectory point data among the plurality of trajectory point data; The interfering trajectory point data and the second trajectory point data are processed by the preset model to be trained to obtain the predicted position category corresponding to the first trajectory point data, including: Processing the interference trajectory point data and the second trajectory point data by using the first sub-model in the preset model to obtain a feature vector corresponding to the interference trajectory point data; Processing the feature vector by using a second sub-model in the preset model to obtain a predicted position category corresponding to the first trajectory point data; updating the model parameters of the preset model according to the first position category corresponding to the first trajectory point data and the predicted position category; Updating the model parameters of the preset model according to the first position category corresponding to the first trajectory point data and the predicted position category, including: The model parameters of the first sub-model and the second sub-model are updated according to the first position category and the predicted position category corresponding to the first trajectory point data.

2. The method according to claim 1, characterized in that Processing the interference trajectory point data and the second trajectory point data by using the first sub-model in the preset model to obtain a feature vector corresponding to the interference trajectory point data includes: Determine a position category vector corresponding to the interference trajectory point data; Determine a position category vector corresponding to the second trajectory point data; The position category vector corresponding to the interference trajectory point data and the position category vector corresponding to the second trajectory point data are input into the first sub-model, so that the first sub-model outputs the feature vector corresponding to the interference trajectory point data.

3. The method according to claim 1, characterized in that Interference processing is performed on the first position category in the first trajectory point data, including any one of the following: Replacing the first location category in the first track point data with preset information; Replacing the first position category in the first track point data with other position categories; The first position category in the first trajectory point data remains unchanged.

4. The method according to any one of claims 1 to 3, characterized in that: Get multiple track point data corresponding to the complete track information, including: Acquire first trajectory information corresponding to a preset time period, the first trajectory information including N trajectory point data, where N is an integer greater than 1; Determine whether the first trajectory information is complete trajectory information according to the value of N and the distribution of trajectory moments corresponding to the N trajectory point data, and the trajectory moment is the moment when the trajectory point data is collected; If yes, the N track point data are determined as the plurality of track point data; If not, the N track point data are supplemented to obtain the plurality of track point data.

5. The method according to claim 4, characterized in that The N track point data are supplemented to obtain the plurality of track point data, including: Obtaining the trajectory time corresponding to the N trajectory point data; Determine, according to the trajectory moments corresponding to the N trajectory point data, a trajectory moment corresponding to at least one trajectory point data to be supplemented; Generate the at least one track point data to be supplemented according to the track time corresponding to the N track point data and the at least one track point data to be supplemented; The plurality of trajectory point data are obtained according to the N trajectory point data and the at least one trajectory point data to be supplemented.

6. The method according to claim 5, characterized in that Generating the at least one track point data to be supplemented according to the track time corresponding to the N track point data and the at least one track point data to be supplemented, comprising: For any track point data to be supplemented, determine the target track point data from the N track point data according to the track time corresponding to the track point data to be supplemented, and the time difference between the track time corresponding to the target track point data and the track time corresponding to the track point data to be supplemented is the smallest; The target track point data is determined as the track point data to be supplemented.

7. The method according to any one of claims 1 to 3, characterized in that: Updating the model parameters of the preset model according to the first position category corresponding to the first trajectory point data and the predicted position category, including: Determining a loss function according to the first position category corresponding to the first trajectory point data and the predicted position category; The model parameters of the preset model are updated through the loss function.

8. The method according to any one of claims 1 to 3, characterized in that: After updating the model parameters of the preset model according to the first position category and the predicted position category corresponding to the first trajectory point data, the method further includes: Determine whether the updated preset model converges; If yes, the updated preset model is used as the trained model; If not, the preset model is repeatedly trained until the updated preset model converges.

9. A data processing method, characterized in that: include: Acquire incomplete trajectory information, where the incomplete trajectory information includes M trajectory point data, each trajectory point data includes a first position category corresponding to the trajectory point, where M is an integer greater than 1 and less than K, and K is the number of trajectory point data included in the complete trajectory information; The M trajectory point data are processed by a trained preset model to obtain K trajectory point data; wherein the preset model is trained by the method described in any one of claims 1 to 8.

10. A model training device, characterized in that: include: an acquisition module, used to acquire multiple track point data corresponding to the complete track information, wherein each track point data further includes at least one of the following: time information corresponding to the track point, a position corresponding to the track point, an area to which the position corresponding to the track point belongs, and coordinate information corresponding to the track point, and each track point data includes a first position category corresponding to the track point; a processing module, configured to perform interference processing on a first position category in first trajectory point data to obtain interference trajectory point data, wherein the first trajectory point data is part of the plurality of trajectory point data; The processing module is further used to process the interference trajectory point data and the second trajectory point data through the preset model to be trained to obtain the predicted position category corresponding to the first trajectory point data, and the second trajectory point data is the trajectory point data other than the first trajectory point data among the multiple trajectory point data, wherein the processing module is further used to process the interference trajectory point data and the second trajectory point data through the first sub-model in the preset model to obtain the feature vector corresponding to the interference trajectory point data; process the feature vector through the second sub-model in the preset model to obtain the predicted position category corresponding to the first trajectory point data; update the model parameters of the preset model according to the first position category and the predicted position category corresponding to the first trajectory point data, including: updating the model parameters of the first sub-model and the second sub-model according to the first position category and the predicted position category corresponding to the first trajectory point data; An updating module is used to update the model parameters of the preset model according to the first position category and the predicted position category corresponding to the first trajectory point data.

11. A data processing device, characterized in that: include: an acquisition module, configured to acquire incomplete trajectory information, wherein the incomplete trajectory information includes M trajectory point data, each trajectory point data includes a first position category corresponding to the trajectory point, wherein M is an integer greater than 1 and less than K, and K is the number of trajectory point data included in the complete trajectory information; A processing module, used to process the M trajectory point data through a trained preset model to obtain K trajectory point data; wherein the preset model is trained by the method described in any one of claims 1 to 8.

12. An electronic device, characterized in that: The electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the steps of the model training method as described in any one of claims 1 to 8 are implemented, or the steps of the data processing method as described in claim 9 are implemented.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the model training method according to any one of claims 1 to 8 are implemented, or the steps of the data processing method according to claim 9 are implemented.

14. A computer program product, characterized in that The computer program product comprises a computer program, which, when executed by a processor, implements the steps of the model training method according to any one of claims 1 to 8, or implements the steps of the data processing method according to claim 9.

Images