Position prediction method, device, equipment and computer-readable storage medium

Abstract

The present application discloses a location prediction method, apparatus, device and computer-readable storage medium, including: obtaining metadata of an object to be predicted within a preset time period, and determining location data information of the object to be predicted based on the metadata, wherein the location data information includes at least the current location data information of the object to be predicted at the current time node; obtaining the first historical output information of a prediction model for the prediction object at the previous time node; the prediction model is at least used to predict the location information of the object to be predicted; determining a prediction time node based on the current time node and a preset time interval; inputting the current location data information, the first historical output information and the prediction time node into the prediction model to obtain the target location information of the object to be predicted at the prediction time node. When performing location prediction, adding the first historical output information can quickly obtain target location information with higher accuracy, thereby improving the efficiency of location prediction.

Description

Technical Field

[0001] The present application relates to the field of data mining, and is related to, but not limited to, a location prediction method, apparatus, device, and computer-readable storage medium. Background Art

[0002] In recent years, the continuous development and widespread adoption of mobile devices and communication technologies have led to an explosive growth in the spatiotemporal data generated by these devices. Simultaneously, various navigation and social networking applications have emerged, resulting in a significant increase in spatiotemporal data containing location information about users and mobile devices. Therefore, predicting the next location of a user or mobile device based on this location information is crucial for providing more convenient services, such as recommendation services, travel planning, and device tracking.

[0003] Taking into account the problems of signal, mobile device, server, etc., the lack of location information occurs frequently, resulting in the lack of completeness of location information. Therefore, when performing location prediction based on incomplete location information, related technologies such as deep recurrent neural networks, Markov models, and household group social relationship models will make location prediction more complex and time-consuming, and also make the location prediction results less accurate, thereby reducing the efficiency of location prediction and affecting the subsequent next decision based on the predicted location. Summary of the Invention

[0004] In view of this, embodiments of the present application provide a location prediction method, apparatus, device, and computer-readable storage medium.

[0005] The technical solution of the embodiment of the present application is implemented as follows:

[0006] The present invention provides a method for predicting a position, including:

[0007] Obtaining metadata of the object to be predicted within a preset time period, and determining location data information of the object to be predicted based on the metadata, wherein the location data information at least includes current location data information of the object to be predicted at a current time node;

[0008] Obtaining first historical output information of a prediction model for the prediction object at a previous time node; the prediction model is at least used to predict location information of the object to be predicted;

[0009] Determining a predicted time node based on the current time node and a preset time interval;

[0010] The current position data information, the first historical output information and the prediction time node are input into the prediction model to obtain the target position information of the object to be predicted at the prediction time node.

[0011] An embodiment of the present application provides a position prediction device, the device comprising:

[0012] A first acquisition module is configured to acquire metadata of an object to be predicted within a preset time period, and determine location data information of the object to be predicted based on the metadata, wherein the location data information at least includes current location data information of the object to be predicted at a current time node;

[0013] A second acquisition module is configured to acquire first historical output information of the prediction model for the prediction object at a previous time node; the prediction model is at least used to predict the position information of the object to be predicted;

[0014] A determination module, configured to determine a predicted time node based on the current time node and a preset time interval;

[0015] The input module is used to input the current position data information, the first historical output information and the prediction time node into the prediction model to obtain the target position information of the object to be predicted at the prediction time node.

[0016] An embodiment of the present application provides a location prediction device, the location prediction device comprising:

[0017] processor; and

[0018] a memory for storing a computer program executable on the processor;

[0019] Wherein, the computer program implements the above-mentioned position prediction method when executed by the processor.

[0020] An embodiment of the present application provides a computer-readable storage medium, wherein the computer storage medium stores computer-executable instructions, and the computer-executable instructions are configured to execute the above-mentioned location prediction method.

[0021] Embodiments of the present application provide a location prediction method, apparatus, device, and computer-readable storage medium, which obtain metadata of an object to be predicted within a preset time length, and determine location data information of the object to be predicted based on the metadata, wherein the location data information at least includes current location data information of the object to be predicted at the current time node; then, obtain first historical output information of a prediction model for the prediction object at a previous time node, wherein the prediction model is at least used to predict the location information of the object to be predicted, and the first historical output information includes relevant location data information of the object to be predicted at the previous time node; then, determine the prediction time node based on the current time node and the preset time interval; finally, input the current location data information, the first historical output information, and the prediction time node into the prediction model, thereby obtaining the target location information of the object to be predicted at the prediction time node. In this way, by obtaining the current position data information of the object to be predicted and the first historical output information of the previous time node, the input information of the prediction model is enriched. Then, the current position data information, the first historical output information and the prediction time node are input into the prediction model, so that the existing information can be fully utilized, so that the target position information of the object to be predicted at the prediction time node can be quickly obtained, and the accuracy of the target position information is high, achieving the purpose of quickly obtaining target position information with high accuracy, and ultimately improving the efficiency of position prediction, which is conducive to making the next decision based on the predicted position and can provide more convenient services. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In the drawings, which are not necessarily drawn to scale, like reference numerals may describe similar components throughout the different views.The drawings illustrate generally, by way of example and not limitation, various embodiments discussed herein.

[0023] Figure 1 A schematic diagram of an implementation flow of a location prediction method provided in an embodiment of the present application;

[0024] Figure 2 A schematic diagram of another implementation flow of the location prediction method provided in an embodiment of the present application;

[0025] Figure 3 A schematic diagram of an implementation flow of a method for determining similar objects provided in an embodiment of the present application;

[0026] Figure 4 A schematic diagram of an implementation flow of a method for determining similarity values ​​between metadata provided in an embodiment of the present application;

[0027] Figure 5 A schematic diagram of another implementation flow of the location prediction method provided in an embodiment of the present application;

[0028] Figure 6 A schematic diagram of metadata transmission provided in an embodiment of the present application;

[0029] Figure 7 A schematic diagram of an implementation flow of the data mining and clustering processing method provided in an embodiment of the present application;

[0030] Figure 8 A schematic diagram of an implementation flow of a method for determining user similarity provided in an embodiment of the present application;

[0031] Figure 9 A schematic diagram of the structure of the prediction module provided in the embodiment of the present application;

[0032] Figure 10 A schematic diagram of the working principle of the prediction module provided in the embodiment of the present application;

[0033] Figure 11 A schematic diagram of an implementation flow of the data preprocessing method provided in an embodiment of the present application;

[0034] Figure 12 A schematic diagram of the structure of a position prediction device provided in an embodiment of the present application;

[0035] Figure 13 A schematic diagram of the structure of a location prediction device according to an embodiment of the present application. DETAILED DESCRIPTION

[0036] In order to make the purpose, technical solutions and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limiting this application. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0037] In the following description, reference is made to “some embodiments”, which describes a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0038] In the following description, the terms "first\second\third" involved are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It can be understood that "first\second\third" can be interchanged with a specific order or sequence where permitted, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0039] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0040] To address the problems of the related art, embodiments of the present application provide a method for predicting a position. This method can be implemented using a computer program. When executed, the computer program completes each step of the method for predicting a position. In some embodiments, the computer program can be executed by a processor in a position prediction device. Figure 1 A schematic diagram of an implementation flow of the location prediction method provided in an embodiment of the present application is shown as follows: Figure 1 As shown, the method includes:

[0041] Step S101 : obtaining metadata of an object to be predicted within a preset time period, and determining location data information of the object to be predicted based on the metadata.

[0042] Here, the location prediction device can obtain the metadata of the object to be predicted within a preset time period from the server periodically or in real time. Of course, the location prediction device can also obtain the metadata of the object to be predicted within a preset time period from the server based on the acquisition instruction. The server can be a primary server, a backup server, a physical server or a virtual server. The embodiments of this application are only illustrative and not limiting. Metadata, also known as intermediary data or relay data, is data that describes data, mainly information that describes the attributes of the data, and is used to support functions such as indicating storage location, historical data, resource search, and file records. Metadata is information about the organization of data, data domains, and their relationships.

[0043] The location data information at least includes the current location data information of the object to be predicted at the current time node. The object to be predicted can be the location prediction device itself, or a user using the location prediction device. The preset duration can be the last 12 months, 8 months, 6 months, etc. The preset duration can be a default value or a custom value. The current time node is used to characterize the current time. The current time node can be a time node accurate to the second, for example: a year b month c day d hour e minute f second; it can also be a time node accurate to the hour, for example: a year b month c day d hour; of course, it can also be a time node accurate to the day. The embodiments of the present application are merely illustrative and are not limited to this.

[0044] In an embodiment of the present application, metadata can be preprocessed to obtain location information, semantic information, and time information, and the location information, semantic information, and time information are determined as location data information. Among them, the location information includes information such as a location identifier (Identity document, ID), location description information, etc., the semantic information includes information such as a semantic ID, semantic description information, and time information includes information such as a time point and a time interval. Here, when performing preprocessing, different information can be distinguished by the identifiers of different information or the content of the information itself. Based on this, the location information, semantic information, and time information can be obtained from the metadata, that is, the location data information can be obtained from the metadata.

[0045] Step S102: obtaining first historical output information of the prediction model for the prediction object at a previous time node.

[0046] Here, the prediction model is at least used to predict the position information of the object to be predicted. The prediction model includes two gated recurrent units (GRUs), wherein the previous time node refers to the time node of the prediction model's last prediction. The first historical output information includes at least position information, and may also include semantic information, time information, etc., thereby being able to reflect the position-related information of the object to be predicted at the historical prediction time point. In addition, based on the GRU's cyclic principle, the first historical output information includes the position semantic-related information of the previous node.

[0047] Step S103: determining a predicted time node based on the current time node and a preset time interval.

[0048] Here, the preset time interval may be 3 hours, 6 hours, 12 hours, etc. The preset time interval may be a default value or a custom setting value.

[0049] In an embodiment of the present application, a preset time interval can be accumulated on the basis of the current time node to obtain an accumulated result, which is then used as the predicted time node.

[0050] Step S104: input the current position data information, the first historical output information and the prediction time node into the prediction model to obtain the target position information of the object to be predicted at the prediction time node.

[0051] In an embodiment of the present application, the prediction model includes three inputs, which are the current position data information in step S101, the first historical output information in step S102, and the prediction time node in step S103. Based on this, the current position data information and the prediction time node are input into a gated loop unit, and the first historical output information is input into another gated loop unit, and then training and prediction are performed based on the neural network to obtain the prediction result, that is, the prediction model can output the target position information of the object to be predicted at the prediction time node.

[0052] Among them, when the prediction model makes predictions based on these three inputs, it can also first determine the first weight factor corresponding to the current position data information, the second weight factor corresponding to the first historical output information, and the third weight factor corresponding to the prediction time node, and the sum of these three weight factors is 1; then based on these three weight factors, the current position data information, the first historical output information and the prediction time node, the target position information of the object to be predicted at the prediction time node is predicted. When determining the three weight factors, the weight factors can be determined by a preset matrix, the current position data information, the first historical output information and the prediction time node. The preset matrix can adjust the weight factors of metadata in different sparsity situations. Therefore, when making predictions, the situation where the metadata is sparse data can be fully considered, thereby improving the accuracy of position prediction in the case of sparse data.

[0053] For example, when the metadata is not sparse data, when making a prediction, the current position data information and the first historical output information both have a high reference significance. Then, the first weight factor and the second weight factor can be set to similar values ​​or values ​​with a small difference. At this time, the first weight factor corresponding to the current position data information is set to 0.4, the second weight factor corresponding to the first historical output information is set to 0.3, and the third weight factor corresponding to the prediction time node is set to 0.3; when the metadata is sparse data, when making a prediction, it indicates that the input data was incomplete when the prediction was made at the previous time node, so the accuracy of the obtained first historical output information is not very high. Based on this, the reference significance of the current position data information is higher than the reference significance of the first historical output information. Then, the first weight factor is greater than the second weight factor, and the difference between the first weight factor and the second weight factor is large. At this time, the first weight factor corresponding to the current position data information is set to 0.5, the second weight factor corresponding to the first historical output information is set to 0.2, and the third weight factor corresponding to the prediction time node is set to 0.3.

[0054] An embodiment of the present application provides a location prediction method, which obtains metadata of an object to be predicted within a preset time length, and determines location data information of the object to be predicted based on the metadata, wherein the location data information at least includes current location data information of the object to be predicted at the current time node; then, obtains first historical output information of a prediction model for the prediction object at a previous time node, wherein the prediction model is at least used to predict the location information of the object to be predicted, and the first historical output information includes relevant location data information of the object to be predicted at the previous time node; then, determines a prediction time node based on the current time node and a preset time interval; finally, inputs the current location data information, the first historical output information, and the prediction time node into the prediction model, thereby obtaining the target location information of the object to be predicted at the prediction time node. In this way, by obtaining the current position data information of the object to be predicted and the first historical output information of the previous time node, the input information of the prediction model is enriched. Then, the current position data information, the first historical output information and the prediction time node are input into the prediction model, wherein the current position data information and the prediction time node are input into a gated loop unit, and the first historical output information is input into another gated loop unit. Then, the prediction result is obtained through training and prediction, and the existing information is fully utilized, so that the target position information of the object to be predicted at the prediction time node can be quickly obtained, and the accuracy of the target position information is high, so as to achieve the purpose of quickly obtaining target position information with high accuracy, and ultimately improve the efficiency of position prediction, which is conducive to making the next decision based on the predicted position and can provide more convenient services.

[0055] Based on the above embodiments, the present application further provides a location prediction method, such as Figure 2 As shown, before executing step S102, the method may further execute the following steps S201 to S208:

[0056] Step S201: determining whether the metadata meets a data sparsity condition based on the multiple time intervals.

[0057] Here, the location data information further includes multiple time intervals, which refer to the intervals between different metadata acquisition moments.

[0058] When implementing step S201 , it may be determined whether there is a target time interval greater than an interval threshold among the multiple time intervals.

[0059] Here, multiple time intervals can be directly obtained from the location data information. The interval threshold can be 1 week, 2 weeks, 1 month, etc., and can be a default value or a custom value. If a target time interval greater than the interval threshold exists among the multiple time intervals, it is determined that the metadata meets the sparse condition, and the process proceeds to step S203. If no target time interval greater than the interval threshold exists among the multiple time intervals, indicating that each time interval is less than or equal to the interval threshold, it is determined that the metadata does not meet the sparse condition, and the process proceeds to step S202.

[0060] Step S202: obtaining first historical output information of the prediction model for the prediction object at a previous time node.

[0061] Here, the implementation process of step S202 may refer to the implementation process of the above-mentioned step S102.

[0062] Step S203: Acquire a set of metadata other than the metadata.

[0063] Here, the other metadata set includes other metadata for objects other than the object to be predicted. Metadata may also include user information. Therefore, the location prediction device may also obtain other metadata for objects other than the object to be predicted from a server based on the user information. Since the server contains multiple other objects, the obtained metadata sets correspond to multiple other objects.

[0064] Step S204 : Based on the metadata and the other metadata sets, determine whether there is a similar object in the other objects that meets a similarity condition with the object to be predicted.

[0065] When implementing step S204, the position information and semantic information of each other metadata can be determined based on each other metadata in the other metadata set, and then mining and clustering processing can be performed to obtain the position semantic cluster set of each other metadata; then, the position information and semantic information are synthesized to obtain the position semantic information of the object to be predicted; then, the longest position semantic common sequence is determined based on the position semantic cluster set and the position semantic information, and the similarities between the metadata and each other metadata are determined based on the longest position semantic common sequence; further, the highest similarity value is determined. When the highest similarity value is greater than or equal to the similarity threshold, it indicates that there is a similar object that meets the similarity condition with the object to be predicted, and the process proceeds to step S205; when the highest similarity value is less than the similarity threshold, it indicates that there is no similar object that meets the similarity condition with the object to be predicted, and the process proceeds to step S207.

[0066] Step S205 : obtaining second historical output information of the prediction model for the similar object at a previous time node.

[0067] Here, the prediction model is also used to predict the location information of the similar objects. If the other metadata of the similar objects is not sparse data, that is, the other metadata corresponding to the similar objects is complete data, then based on the complete other metadata of the similar objects and the prediction model, a second historical output information with a high prediction accuracy at the previous time node can be obtained.

[0068] Step S206: input the current position data information, the second historical output information and the prediction time node into the prediction model to obtain the target position information of the object to be predicted at the prediction time node.

[0069] The implementation steps of step S206 may refer to the implementation steps of step S104 , wherein step S206 differs from step S104 in that the first historical output information in step S104 is replaced by the second historical output information corresponding to the similar object as input information.

[0070] Step S207: extract the current position information and the current semantic information from the current position data information.

[0071] Here, similar to the location data information, the current location data information includes the current location information and the current semantic information, wherein the current location information includes the current location ID, location description information and other information, and the semantic information includes the semantic ID, semantic description information and other information. The current location information and current semantic information can be extracted based on the identification of different information.

[0072] Step S208 : inputting the current position information, current semantic information, first historical output information and prediction time node into the prediction model to obtain target position information of the object to be predicted at the prediction time node.

[0073] Here, the current position information and the current semantic information can represent the position-related information of the object to be predicted at the current time node. At this time, there is no similar object to the object to be predicted. Then, the relevant information itself is directly input into the prediction model. After prediction by the prediction model, the target position information of the object to be predicted at the prediction time node can be obtained.

[0074] In an embodiment of the present application, through steps S201 to S208, it is first determined whether the metadata meets the sparse condition based on the time interval. If the metadata does not meet the sparse condition and the metadata is not sparse data, the first historical output information of the prediction model for the object to be predicted at the previous time node is obtained. If the metadata meets the sparse condition and the metadata is sparse data, the similarity threshold is determined based on the longest position semantic common sequence. When the highest similarity threshold is greater than or equal to the similarity threshold, similar objects can be determined. Further, the second historical output information of the prediction model for similar objects at the previous time node is obtained, and then the current position data information, the second historical output parameter and the prediction time node of the object to be predicted are input into the prediction model, thereby obtaining target position information with high accuracy at the prediction time node of the object to be predicted; and when the highest similarity threshold is less than the similarity threshold, the similar object cannot be determined. Further, the current position information and current semantic information are extracted from the current position data information, and the current position information, current semantic information, the first historical output information and the prediction time node are input into the prediction model, thereby also obtaining relatively accurate target position information of the object to be predicted at the prediction time node. In this way, the target position information of the object to be predicted at the prediction time node can be quickly obtained, and the accuracy of the target position information is high, achieving the purpose of quickly obtaining target position information with high accuracy, and ultimately improving the efficiency of position prediction.

[0075] In some embodiments, when implementing step S204, Figure 3 As shown, this can be achieved by following steps S2041 to S2047:

[0076] Step S2041 : determining the location information and semantic information of each other metadata based on each other metadata in the other metadata set.

[0077] Here, other metadata also includes location information, semantic information, and time information. Location information includes information such as a location ID and location description information. The location ID can be the latitude and longitude coordinates of the location, for example, (X, Y). The location description information can be attribute information such as the name and nature of the location, for example, the location description information can be Building A, Office Building, Company, Shopping Mall, etc. Semantic information includes information such as a semantic ID and semantic description information. The semantic ID can be a preset code, for example, 0123. The semantic description can be a description of the state at the corresponding location, for example, the semantic description information can be "working," "training," "entertainment," "business trip," etc. Therefore, the location information and semantic information of each other metadata can be determined from each other metadata based on different IDs.

[0078] Step S2042: Mine and cluster the location information and semantic information of each of the other metadata to obtain the location-semantic clustering set of each of the other metadata.

[0079] Here, location pattern mining can be performed on each other user through user information, location information, and time information to obtain a location pattern set. Semantic pattern mining can also be performed on each other user through user information, semantic information, and time information to obtain a semantic pattern set. Among them, the pattern mining algorithm can be the Prefix-Projected Pattern Growth (PrefixSpan) algorithm. For the pattern mining algorithm, the embodiments of the present application are only for illustrative purposes and are not limited. Then, the location pattern set and the semantic pattern set are aggregated to obtain a location-semantic clustering set.

[0080] Step S2043: Determine the similarity values between the metadata and each of the other metadata in the set of the other metadata.

[0081] When implementing Step S2043, as Figure 4 shown, it can be implemented through the following Steps S431 to S433:

[0082] Step S431: Synthesize the location information and the semantic information to obtain location-semantic information.

[0083] Here, the location-semantic information includes a set of location-semantic sequences of the object to be predicted. For example, the location information is <Building A, Square B, Community C>, and the corresponding semantic information is <Company, Restaurant, Home>. Then, the corresponding set of location-semantic sequences after synthesis is <Company - Building A, Restaurant - Square B, Home - Community C>. Further, the set of location-semantic sequences includes 3 sequences.

[0084] Step S432: Based on the location-semantic information and the location-semantic clustering set, determine the longest location-semantic common sequence between the metadata and each of the other metadata.

[0085] Here, the location-semantic clustering set includes a set of location-semantic sequences of other objects. When implementing Step S432, the location-semantic information and the location-semantic clustering sets of each of the other metadata are compared one by one. The set of identical location-semantic sequences with the same sequences as those included in the location-semantic information is determined from the location-semantic clustering set, and the one with the largest number of sequences included in the set of identical location-semantic sequences is determined as the longest location-semantic common sequence. Among them, when there is no set of identical location-semantic sequences, it is considered that the longest location-semantic common sequence is an empty set.

[0086] Step S433 : determining, based on the longest position semantic common sequence, each similarity value between the metadata and each other metadata in the set of other metadata.

[0087] Here, taking a piece of other metadata as an example, when determining the similarity between the metadata and the other metadata, the number of location semantics contained in the longest common location semantic sequence can be used as the numerator. Next, the first number of location semantics contained in the location semantic sequence set is obtained, along with the second number of location semantics contained in the location semantic sequence set of the other metadata. Based on this, the sum of the first and second numbers is calculated, and this sum is used as the denominator. Finally, the final ratio is used to determine the similarity value. For example, if the longest location semantic sequence is <restaurant-B Plaza, home-C Community>, the location semantic sequence set is <company-A Building, restaurant-B Plaza, home-C Community>, and the location semantic sequence set of the other metadata is <library-D Building, company-A Building, restaurant-B Plaza, home-C Community>, then the numerator is 2, the first number is 3, the second number is 4, and the denominator is 7. Therefore, the similarity value is 2 / 7. This translates to a similarity value of approximately 0.286.

[0088] Step S2044: determine the highest similarity value from each similarity value.

[0089] Step S2045: Determine whether the highest similarity value is greater than or equal to a similarity threshold.

[0090] Here, the similarity threshold can be 0.6, 0.8, etc. When the highest similarity threshold is greater than or equal to the similarity threshold, step S2046 is entered, that is, it indicates that there are similar objects to the object to be predicted; when the highest similarity threshold is less than the similarity threshold, step S2047 is entered, that is, it indicates that there are no similar objects to the object to be predicted.

[0091] Step S2046: Determine whether there is a similar object that meets similarity conditions with the object to be predicted.

[0092] At this time, if the highest similarity value is greater than or equal to the similarity threshold, it indicates that there are similar objects that meet similarity conditions with the object to be predicted, wherein the similar objects are other objects corresponding to the highest similarity threshold.

[0093] Step S2047: Determine whether there is no similar object that meets the similarity condition with the object to be predicted.

[0094] Through steps S2041 to S2047, first, based on each other metadata in the other metadata set, the position information and semantic information of each other metadata are determined, and then mining and clustering processing is performed to obtain the position semantic cluster set of each other metadata; then, the position information and semantic information are synthesized to obtain the position semantic information of the object to be predicted; then, based on the position semantic cluster set and the position semantic information, the longest position semantic common sequence is determined, and based on the longest position semantic common sequence, the various similarities between the metadata and each other metadata are determined; further, the highest similarity value is determined, and when the highest similarity value is greater than or equal to the similarity threshold, it indicates that there is a similar object that meets the similarity condition with the object to be predicted; when the highest similarity value is less than the similarity threshold, it indicates that there is no similar object that meets the similarity condition with the object to be predicted, thereby judging whether there is a similar object, and when a similar object exists, the similar object is determined.

[0095] Based on the above embodiments, the embodiments of the present application further provide a location prediction method. The embodiments of the present application are to solve the problem of data sparsity in user location prediction and the prediction process. On the one hand, through mining and clustering, the data is accurate, complete, and consistent. On the other hand, the model is improved. By connecting the improved gated recurrent unit and GRU in series, a larger network and model can be constructed in the later stage. Because it has two gates, the computational efficiency is also higher than that of other gates, which improves the accuracy of the next position of the user position and the accuracy of the first K positions.

[0096] In the embodiments of this application, user location data can be obtained through mobile communication base stations and satellite signals, or through check-in data on a certain platform. This application takes into account the situation where individual user location data is missing for a certain time period and addresses this issue through two parts: data preprocessing and the design of a prediction model. The overall module of the device for predicting the location of individual users is as follows, mainly divided into the following three modules: location data preprocessing, location data mining and clustering, and location prediction.

[0097] like Figure 5 As shown, the location prediction method provided in the embodiment of the present application can be implemented through the following steps S501 to S509:

[0098] Step S501: Obtain individual user metadata.

[0099] Here, the metadata of individual users can be obtained from the server. The transmission diagram of user metadata is as follows: Figure 6 As shown. Figure 6In the case of a mobile phone, a positioning chip is installed in the terminal. This allows the terminal to locate its own position by accessing base stations and also by satellite. Once the terminal is connected to the internet, this terminal / user location information is transmitted to a workstation via the network. The workstation aggregates the information and then uploads it to a server. Therefore, the laptop can obtain this terminal / user location information from the server for analysis and processing.

[0100] Step S502: pre-process individual user metadata to obtain processed data.

[0101] Here, preprocessing can include parsing and decomposition. After preprocessing, the processed data may include user information, geographic location information, semantic information, time information, distance information, etc. User information may include a user identity document (ID); geographic location information includes a location ID and location information; semantic information includes a semantic ID and semantic information; time information includes a time interval; and distance information includes a distance interval. During preprocessing, information other than user information, geographic location information, semantic information, time information, and distance information may also be filtered out.

[0102] Step S503: determine whether the processed data meets the sparse condition.

[0103] Here, the sparseness of individual user spatiotemporal data within a time interval is determined. Specifically, the sparse condition is that the individual user's metadata lacks location information within a week or two weeks, and the time interval is determined based on the sparsity of the metadata. If the processed data meets the sparse condition, the individual user data is sparse and is recorded as data B. The process proceeds to step S504 to obtain all user metadata. If the processed data does not meet the sparse condition, the individual user data is not sparse and is recorded as data A. The process proceeds to step S508, where the processed data is directly input into the prediction module.

[0104] Step S504: Obtain all user metadata.

[0105] Here, all user metadata except the individual user metadata in step S501 are acquired from the server.

[0106] Step S505: Perform data mining and clustering processing on all user metadata.

[0107] When implementing step S505, all user metadata are mined and clustered using the PrefixSpan algorithm data mining method, association rules and longest common subsequence clustering method to obtain a processed data set.

[0108] In some embodiments, the location tag is the basic data, and usually the semantic sequence and the location sequence are in a one-to-many relationship. For example, semantic sequences such as "class", "lunch break", and "self-study" can correspond to the location sequence "D school". Therefore, semantic pattern mining can be performed on the basis of metadata pattern mining. According to the unique user identification ID, the location and / or semantic pattern mining is performed based on the PrefixSpan algorithm. The data information required for location pattern mining includes user ID, location ID, and time information. The data information required for semantic mining pattern includes user ID, time information, semantic ID, and semantic information. Each user's location and / or semantic pattern may have more than one pattern over a long period of time. For example, a user has two patterns <home, company> and <home, school, company> within a certain period of time. However, a user has only one pattern within a fixed time period.

[0109] In some other embodiments, such as Figure 7 As shown, step S505 can be implemented through the following steps S5051 to S50510:

[0110] Step S5051: pre-process all user metadata to obtain a processed data set.

[0111] Here, the dataset includes semantic label sequences and position label sequences.

[0112] Step S5052: Obtain a position tag sequence in the data set.

[0113] Step S5053: Perform position pattern mining on the position tag sequence.

[0114] Step S5054: Obtain a location pattern set.

[0115] Step S5055: determine whether semantic quantity is needed.

[0116] If semantic quantity is required, the process proceeds to step S5057 , that is, continues to obtain semantic sequences; if semantic sequences are not required, the process proceeds to step S5056 , and directly outputs the user location cluster set.

[0117] Step S5056: output the user location cluster set.

[0118] Step S5057: Obtain a semantic tag sequence based on the user location cluster set and the data set.

[0119] Step S5058: perform semantic pattern mining on the semantic tag sequence.

[0120] Step S5059: Obtain a semantic pattern set.

[0121] Step S50510: output all user locations and / or semantic cluster sets.

[0122] Step S506: determine whether there are similar users similar to the individual user among all users.

[0123] Here, the location and / or semantic clustering set of each user among all users is recorded as data C. Individual users similar to data B are searched from data C, and the data of these similar individual users are used to predict the location. Similarity is determined by the user similarity metric value derived from association rules and the longest common subsequence. If the maximum similarity metric value in the set is less than the set threshold, that is, the similarity metric value does not meet the conditions and there are no similar users similar to the individual user, then the data is discarded and the process proceeds to step S508, where the individual user data is directly input into the prediction module; if the maximum similarity metric value in the set is greater than or equal to the set threshold, that is, the similarity metric value meets the conditions, then the data D of the similar user corresponding to the maximum similarity metric value is selected. If there are similar users similar to the individual user, the process proceeds to step S507, where the data of the similar users is input into the prediction module. The set threshold can be set to 0.68, which can also be a default value or a custom value.

[0124] In the embodiments of this application, Figure 8 As shown, obtaining the similarity between two users can be achieved by following steps S5061 to S5065:

[0125] Step S5061: Acquire the location and / or semantic pattern set of the first user.

[0126] Here, the location and / or semantic pattern set of the first user may be denoted as user-A.

[0127] Step S5062: Acquire the location and / or semantic pattern set of the second user.

[0128] Here, the location and / or semantic pattern set of the first user may be denoted as user-B.

[0129] Step S5063: Determine the longest common subsequence between the first user and the second user.

[0130] Here, the subsequence in user-A and the subsequence in user-B are analyzed to determine the common subsequence in the two pattern sets, which is recorded as LCS(user-A, user-B).

[0131] Step S5064: Calculate the pattern similarity between the first user and the second user.

[0132] When implementing step S5064, it can be calculated using the following formula (1):

[0133]

[0134] in,

[0135] Step S5065: Calculate the similarity between the two users.

[0136] When implementing step S5065, it can be calculated using the following formula (2):

[0137]

[0138] Where user-A={p A1 ,p A2 ,...,p An} and user-B={p B1 ,p B2 ,...,p Bm} are the pattern sets of the first user and the second user respectively. S(p Ai )、S(p Bj ) is the support of the association rules, that is, they are p Ai , p Bj The number of users divided by the total number of users.

[0139] Step S507: inputting the data of similar users into the prediction module.

[0140] Here, as Figure 9 As shown in FIG, the prediction module is composed of an improved gated recurrent unit, a gated recurrent unit unit, an excitation function layer and a training module, wherein the excitation function can be SoftMax.

[0141] Among them, when the prediction is realized by the improved gated recurrent unit, when the data of an individual user is judged to be sparse and has a great influence on the subsequent model training, the data can be processed through steps S501 to S505, and then there are two prediction methods to choose from: the first prediction method, there is another user whose similarity with the individual user is greater than the set maximum threshold, in this case, the next possible location can be predicted by the user with high similarity. The second prediction method, if other users with suitable similarity cannot be obtained, then the user's own location and / or semantic pattern set is used to predict the location that the user is more inclined to in the future. If the user data is sparse but has little influence on the model training, that is, there is no data in the user metadata within a short time interval, or there are short time intervals between a lot of data, then the next location can be predicted by selecting a suitable model. The general premise of conventional models in this field is that the data is a specified sequence, such as dividing the data into equal time intervals. However, for the case where there is no data in a short time period, it is not conducive to feature training. The traditional gated recurrent neural network system is based on position / semantic state sequence information as the input gate, but when there is no data within a certain period of time, the historical sequence state information is forgotten by the model. Based on this, Figure 10 As shown, the embodiment of the present application proposes to use the time state information as the first input, add an interface with the time state as the input to the basic gated recurrent unit, and use a sigmoid gate to control the proportion of the two in the prediction. When the sequence state information disappears, the time state information still plays a role in the model. Figure 10 :

[0142] h n-1 The first input is the hidden state passed down from the previous time point. This hidden state contains relevant information of the previous node, such as time point, position and / or semantic information.

[0143] The second input is the location and / or semantic information of the current time node.

[0144] The third input is the user's individual code, which inputs the predicted next time point and the time interval between it and the current time point.

[0145] refer to Figure 10 , f n The calculation is as shown in formula (3), i n The calculation is as follows:

[0146]

[0147]

[0148] Among them, both formulas (3) and (4) control the location, semantics, and time state information of individual user history and current input.

[0149] refer to Figure 10 , The calculation is as shown in formula (5):

[0150]

[0151] Formula (3) represents the update process of the location and semantic state of an individual user over a long period of time.

[0152] refer to Figure 10 , The calculation is as shown in formula (6), h n The calculation is as shown in formula (7):

[0153]

[0154]

[0155] Among them, formula (5) obtains different neurons according to the weights of the control position, semantic information hidden state and time information hidden state. u is the individual user code, p n-1 is the last position vector, s n-1 is the previous semantic vector. t n is the time when the next location / semantic information of an individual user is predicted, t d is the time interval between the predicted time and the last time. Gate function The weights of the sequence hidden state and the time hidden state are restricted to between 0 and 1. f , A i , A h , A T , is the linear transformation matrix, b f , b i , b h , b T , is the linear transformation offset. σ is the activation function sigmoid, h n-1 is the hidden state vector of the previous time period.

[0156] exist Figure 10 In the output, h n It will be used as the first input for the next prediction, and the cycle will be implemented in this way, so that the first input of each prediction includes the relevant information of all previous nodes, and can ensure accurate prediction when the data is sparse.

[0157] Step S508: inputting the individual user's data into the prediction module.

[0158] like Figure 11 As shown, step S508 can be implemented through steps S5081 to S5084:

[0159] Step S5081: One-hot encoding is performed on individual user data.

[0160] Here, given a user's historical location and / or semantic sequence, their next location / semantic information is predicted. Because the location / semantic data for individual users is discrete, to facilitate calculations between discrete values, one-hot encoding is used to convert the discrete data into a sparse matrix.

[0161] Step S5082: perform matrix decomposition and dimensionality reduction.

[0162] Step S5083: linear conversion.

[0163] Here, the five input features of individual users are linearly transformed and combined into two input features: sequence features and time characteristics Among them, the five input feature quantities are: data position and / or semantic vector, last position vector, last semantic vector, last time vector, and time interval vector between last time and predicted time.

[0164] Step S5084: input the improved gated recurrent unit.

[0165] The two input features obtained above are used as the input of the improved gated recurrent unit. The training module uses the stochastic gradient descent algorithm to reduce the cross entropy loss.

[0166] Step S509: output the prediction result.

[0167] Here, the similarity threshold between an individual user and other users can be set to 0.68, because the next location of an individual user is a certain geographical location coordinate, but there are many optional geographical locations, and multiple prediction results will be given during the actual prediction. Therefore, in this field, the accuracy of the first K predicted locations is usually used as the performance of location prediction. As shown in Table 1, taking the "New York" location as an example, when K is 10, the accuracy is 0.7142, which is greater than the similarity threshold. Therefore, in the embodiment of the present application, K is taken as 10, that is, the prediction accuracy of the first 10 is compared. It is also based on the developer application programming interface (API) check-in data provided by Foursquare. Xavier is used to initialize the weights of the entire network, and the stochastic gradient descent with learning rate decay is used to reduce the loss function, and the hidden layer dimension that minimizes the cross entropy loss is selected to be 228 layers.

[0168] Table 1 Example of “New York” location prediction accuracy

[0169] Dataset Top 1 accuracy Top 5 accuracy Top 10 accuracy New York 0.3504 0.6321 0.7142

[0170] In this way, through steps S501 to S509, individual user metadata is obtained, and the individual user metadata is preprocessed to obtain processed data information; then it is determined whether the processed data is sparse data. If the processed data is sparse data, all user metadata is further obtained, and data mining and clustering processing is performed on all user metadata to obtain all user positions and / or semantic cluster sets; based on this, it is determined by similarity whether there are similar users similar to the individual user among all users. If there are similar users, the similar users are input into the prediction module including the improved gated recurrent unit; if there are no similar users or the processed data of the individual user is not scarce data, the individual user data is directly input into the prediction module including the improved gated recurrent unit. In this way, based on the position prediction under sparse data, the prediction time can be accelerated, and the accuracy of the position prediction can be improved, thereby improving the prediction efficiency.

[0171] Based on the foregoing embodiments, an embodiment of the present application provides a position prediction device, and the modules included in the device, as well as the units included in each module, can be implemented by a processor in a computer device; of course, they can also be implemented by a specific logic circuit; in the implementation process, the processor can be a central processing unit (CPU), a microprocessor (MPU), a digital signal processor (DSP) or a field programmable gate array (FPGA), etc.

[0172] The present application further provides a position prediction device. Figure 12 A schematic diagram of the structure of the position prediction device provided in the embodiment of the present application is shown in FIG. Figure 12 As shown, the position prediction device 1200 includes:

[0173] A first acquisition module 1201 is configured to acquire metadata of an object to be predicted within a preset time period, and determine location data information of the object to be predicted based on the metadata, wherein the location data information includes at least current location data information of the object to be predicted at a current time node;

[0174] The second acquisition module 1202 is configured to acquire first historical output information of the prediction model for the prediction object at a previous time node; the prediction model is at least used to predict the location information of the object to be predicted;

[0175] A determination module 1203 is configured to determine a predicted time node based on the current time node and a preset time interval;

[0176] The input module 1204 is used to input the current position data information, the first historical output information and the prediction time node into the prediction model to obtain the target position information of the object to be predicted at the prediction time node.

[0177] In some embodiments, the determination module 1203 is further configured to determine whether the metadata satisfies a data sparsity condition based on the multiple time intervals; and the location prediction device 1200 further includes:

[0178] The third acquisition module is used to obtain the first historical output information of the prediction model for the prediction object at the previous time node when it is determined that the data sparsity condition is not met.

[0179] In some embodiments, the position prediction device 1200 further includes:

[0180] a fourth acquisition module, configured to acquire, when it is determined that the metadata satisfies a data sparsity condition, a set of other metadata other than the metadata, wherein the set of other metadata includes other metadata of other objects other than the object to be predicted;

[0181] The determination module 1203 is further configured to determine whether there is a similar object in the other objects that satisfies a similarity condition with the object to be predicted based on the metadata and the other metadata set;

[0182] A fifth acquisition module is configured to, when the similar object exists, acquire second historical output information of the prediction model for the similar object at a previous time node, wherein the prediction model is further configured to predict location information of the similar object;

[0183] The input module 1204 is further configured to input the current position data information, the second historical output information and the prediction time node into the prediction model to obtain the target position information of the object to be predicted at the prediction time node.

[0184] In some embodiments, the current location data information includes current location information and current semantic information, and the location prediction device 1200 further includes:

[0185] an extraction module, configured to extract the current position information and the current semantic information from the current position data information when the similar object does not exist;

[0186] The input module 1204 is further configured to input the current position information, current semantic information, first historical output information and prediction time node into the prediction model to obtain target position information of the object to be predicted at the prediction time node.

[0187] In some embodiments, based on the metadata and the other metadata sets, the determining module 1203 includes:

[0188] a first determining unit, configured to determine respective similarity values ​​between the metadata and respective other metadata in the set of other metadata;

[0189] a second determining unit, configured to determine a highest similarity value from among the similarity values;

[0190] a third determination unit, configured to determine whether the highest similarity value is greater than or equal to a similarity threshold, wherein when the highest similarity value is greater than or equal to the similarity threshold, it is determined that there is a similar object that meets the similarity condition with the object to be predicted; and when the highest similarity value is less than the similarity threshold, it is determined that there is no similar object that meets the similarity condition with the object to be predicted.

[0191] In some embodiments, the determination module 1203 is further configured to determine the location information and semantic information of each other metadata based on each other metadata in the other metadata set; and the location prediction device 1200 further includes:

[0192] The mining and clustering module is used to mine and cluster the location information and semantic information of each other metadata to obtain a location semantic cluster set of each other metadata.

[0193] In some embodiments, the location data information further includes location information and semantic information, and the determining module 1203 further includes:

[0194] a synthesis unit, configured to synthesize the position information and the semantic information to obtain position semantic information;

[0195] a fourth determining unit, configured to determine, based on the location semantic information and the location semantic cluster set, a longest location semantic common sequence between the metadata and each other metadata;

[0196] The fifth determining unit is configured to determine, based on the longest position semantic common sequence, each similarity value between the metadata and each other metadata in the set of other metadata.

[0197] In some embodiments, the determining module 1203 further includes:

[0198] A sixth determination unit is used to determine whether there is a target time interval greater than an interval threshold among the multiple time intervals; wherein, when the target time threshold exists, it is determined that the metadata meets the sparse condition; when the target time interval does not exist, it is determined that the metadata does not meet the sparse condition.

[0199] In some embodiments, the first acquisition module 1201 includes:

[0200] A preprocessing unit, configured to preprocess the metadata to obtain location information, semantic information, and time information;

[0201] A seventh determining unit is configured to determine the position information, semantic information, and time information as the position data information.

[0202] In some embodiments, the input module 1204 includes:

[0203] an eighth determining unit, configured to determine weight factors based on a preset matrix, the current location data information, the first historical output information, and the prediction time node, wherein the weight factors include a first weight factor corresponding to the current location data information, a second weight factor corresponding to the first historical output information, and a third weight factor corresponding to the prediction time node, wherein the preset matrix is ​​used to adjust the weight factors of the metadata under different sparse conditions;

[0204] A prediction unit is used to predict the target position information of the object to be predicted at the prediction time node based on the first weight factor, the second weight factor, the third weight factor, the current position data information, the first historical output information and the prediction time node.

[0205] It should be noted that the description of the position prediction device in the embodiment of the present application is similar to the description of the method embodiment described above, and has similar beneficial effects as the method embodiment, so it will not be repeated here. For technical details not disclosed in the embodiment of the present device, please refer to the description of the method embodiment of the present application for understanding.

[0206] It should be noted that in the embodiment of the present application, if the above-mentioned location prediction method is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application, or the part that contributes to the relevant technology, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM), a magnetic disk or an optical disk. In this way, the embodiment of the present application is not limited to any specific combination of hardware and software.

[0207] Accordingly, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps in the position prediction method provided in the above embodiment are implemented.

[0208] The embodiment of the present application provides a location prediction device, Figure 13 A schematic diagram of the structure of the position prediction device provided in the embodiment of the present application is shown as follows: Figure 13As shown, the location prediction device 1300 includes: a processor 1301, at least one communication bus 1302, a user interface 1303, at least one external communication interface 1304, and a memory 1305. The communication bus 1302 is configured to enable communication between these components. The user interface 1303 may include a display screen, and the external communication interface 1304 may include a standard wired interface and a wireless interface. The processor 1301 is configured to execute the location prediction method program stored in the memory to implement the steps of the location prediction method provided in the above embodiment.

[0209] The description of the above embodiments of the location prediction device and storage medium is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment. For technical details not disclosed in the embodiments of the location prediction device and storage medium of this application, please refer to the description of the method embodiment of this application for understanding.

[0210] It should be understood that "one embodiment" or "an embodiment" mentioned throughout the specification means that the specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application. The above-mentioned serial numbers of the embodiments of the present application are for description only and do not represent the advantages and disadvantages of the embodiments.

[0211] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0212] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0213] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed across multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the embodiment of the present application.

[0214] In addition, all functional units in the embodiments of the present application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the above-mentioned integrated units can be implemented in the form of hardware or in the form of hardware plus software functional units.

[0215] Those skilled in the art will understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above-mentioned method embodiments; and the aforementioned storage medium includes: mobile storage devices, ROM, disks or optical disks, and other media that can store program codes.

[0216] Alternatively, if the above-mentioned integrated unit of the present application is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application, or the part that contributes to the relevant technology, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling an AC to execute all or part of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROMs, magnetic disks or optical disks.

[0217] The above is merely an embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A location prediction method, characterized in that: The method comprises: Obtaining metadata of the object to be predicted within a preset time period, and determining location data information of the object to be predicted based on the metadata, wherein the location data information at least includes current location data information of the object to be predicted at a current time node; Obtaining first historical output information of the prediction model for the prediction object at a previous time point; the prediction model is at least used to predict the position information of the object to be predicted; the first historical output information includes location-related information of the object to be predicted at a historical prediction time point; Determining a predicted time node based on the current time node and a preset time interval; The current position data information, the first historical output information and the prediction time node are input into the prediction model to obtain the target position information of the object to be predicted at the prediction time node.

2. The method according to claim 1, wherein The location data information further includes a plurality of time intervals. Before obtaining the first historical output information of the prediction model for the prediction object at the previous time node, the method further includes: determining, based on the plurality of time intervals, whether the metadata satisfies a data sparsity condition; When it is determined that the data sparsity condition is not met, first historical output information of the prediction model for the prediction object at a previous time node is obtained.

3. The method according to claim 2, characterized in that The method further comprises: When it is determined that the metadata meets the data sparse condition, obtaining a set of other metadata other than the metadata, wherein the set of other metadata includes other metadata of other objects other than the object to be predicted; Based on the metadata and the other metadata sets, determining whether there is a similar object among the other objects that meets a similarity condition with the object to be predicted; When the similar object exists, obtaining second historical output information of the prediction model for the similar object at a previous time node, the prediction model is further used to predict location information of the similar object; The current position data information, the second historical output information and the prediction time node are input into the prediction model to obtain the target position information of the object to be predicted at the prediction time node.

4. The method according to claim 3, wherein The current position data information includes current position information and current semantic information, and the method further includes: When the similar object does not exist, extracting the current position information and the current semantic information from the current position data information; The current position information, current semantic information, first historical output information and prediction time node are input into the prediction model to obtain target position information of the object to be predicted at the prediction time node.

5. The method according to claim 3, characterized in that The determining, based on the metadata and the other metadata sets, whether there is a similar object in the other objects that satisfies a similarity condition with the object to be predicted includes: determining respective similarity values ​​between the metadata and respective other metadata in the set of other metadata; Determine the highest similarity value from among the similarity values; Determine whether the highest similarity value is greater than or equal to a similarity threshold, wherein, when the highest similarity value is greater than or equal to the similarity threshold, it is determined that there is a similar object that meets the similarity condition with the object to be predicted; when the highest similarity value is less than the similarity threshold, it is determined that there is no similar object that meets the similarity condition with the object to be predicted.

6. The method according to claim 5, characterized in that Before determining whether there is a similar object among the other objects that satisfies a similarity condition with the object to be predicted based on the metadata and the other metadata sets, the method further includes: Determining, based on each other metadata in the other metadata set, location information and semantic information of each other metadata; Mining and clustering are performed on the location information and semantic information of each other metadata to obtain a location semantic cluster set of each other metadata.

7. The method according to claim 6, characterized in that The location data information further includes location information and semantic information. The determining of each similarity value between the metadata and each other metadata in the set of other metadata includes: synthesizing the position information and the semantic information to obtain position semantic information; Determining, based on the location semantic information and the location semantic cluster set, a longest location semantic common sequence between the metadata and each other metadata; Determine, based on the longest position semantic common sequence, respective similarity values ​​between the metadata and respective other metadata in the set of other metadata.

8. A position prediction device, characterized in that: The device comprises: A first acquisition module is configured to acquire metadata of an object to be predicted within a preset time period, and determine location data information of the object to be predicted based on the metadata, wherein the location data information at least includes current location data information of the object to be predicted at a current time node; A second acquisition module is configured to acquire first historical output information of the prediction model for the prediction object at a previous time point; the prediction model is at least used to predict the position information of the object to be predicted; the first historical output information includes information related to the position of the object to be predicted at a historical prediction time point; A determination module, configured to determine a predicted time node based on the current time node and a preset time interval; The input module is used to input the current position data information, the first historical output information and the prediction time node into the prediction model to obtain the target position information of the object to be predicted at the prediction time node.

9. A location prediction device, characterized in that: The device comprises: processor; and a memory for storing a computer program executable on the processor; Wherein, when the computer program is executed by a processor, it implements the position prediction method described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, which are configured to execute the position prediction method described in any one of claims 1 to 7.

Images