CN116304987A - Mining and predicting method and system for space-time track of mobile user - Google Patents
Mining and predicting method and system for space-time track of mobile user Download PDFInfo
- Publication number
- CN116304987A CN116304987A CN202310331408.6A CN202310331408A CN116304987A CN 116304987 A CN116304987 A CN 116304987A CN 202310331408 A CN202310331408 A CN 202310331408A CN 116304987 A CN116304987 A CN 116304987A
- Authority
- CN
- China
- Prior art keywords
- track
- sequence
- time
- user
- term
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a method and a system for mining and predicting space-time trajectories of mobile users, wherein the scheme comprises the steps of preprocessing acquired historical space-time trajectory data of the current mobile users to obtain a time-ordered trajectory sequence and a space-time context attribute set; embedding the track sequence and the user characteristics to obtain track sequence characteristics and user characteristics respectively; inputting the track sequence with the embedded features into a hierarchical attention architecture taking a pre-trained self-attention neural network as a backbone network to obtain long-term track features and short-term track features; based on the embedded user characteristics, carrying out characteristic fusion on the long-term track characteristics and the short-term track characteristics to obtain fusion track characteristics; and processing by using a full connection layer based on the fusion track characteristics to obtain a track prediction result of the current mobile user.
Description
Technical Field
The invention belongs to the technical field of computer information communication and service calculation, and particularly relates to a method and a system for mining and predicting space-time trajectories of mobile users.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
Mobile users refer to users who can communicate while moving or while briefly staying with a car-mounted table, a handset or a portable communication device; in recent years, with the rapid development of software and hardware technologies, the number of mobile users has increased dramatically. The user can log in and access the mobile service platform through the mobile communication network at any time and any place to obtain various personalized services provided by the relevant platform. And personalized mobile services such as various map applications, travel, social networking sites and the like bring convenience to users and attract more customer resources for the users.
The personalized service is provided to the mobile user without the support of large amounts of user data. In the scenario of predicting the space-time trajectory of the mobile user, the mobile terminal device acquires the space-time trajectory data of the user in real time through various sensors such as a GPS position sensor and the like, and the relevant service provider can obtain and utilize the data uploaded by the user to improve the personalized service effect under the condition of meeting the corresponding privacy protocol standard, for example: the cell phone service website Foursquare, gowalla and the like encourages users to share the current geographical location information with others to provide a geographical location-based service.
However, the inventors have found that there are still a number of problems with existing prediction methods: firstly, the mobile user only interacts with certain specific places, so that the generated data is very sparse, and the accuracy of the track prediction of the mobile user is affected; secondly, the trajectory of the mobile user is influenced by time and geographical location factors, and in time, the trajectory of the mobile user exhibits different periodic characteristics, such as daily life, weekend leisure, holidays and the like, and in addition, the trajectory of the user also has different performances on both long-term and short-term time scales, for example, the user tends to visit an indoor office place in the long term, but for some reasons, the user frequently visits an outdoor entertainment place in the near term; geographically, mobile user trajectories often exhibit spatial clustering, i.e., locations visited by users are often geographically closely spaced. Based on the above problems, it is difficult for the existing prediction method to mine complex behavior patterns of the user constrained by space-time factors under limited mobile user trajectory data to predict the user trajectory.
Disclosure of Invention
In order to solve the problems, the invention provides a mining and predicting method and a system for a space-time track of a mobile user, wherein the scheme models the long-term and short-term space-time tracks of the mobile user by adopting a self-supervision learning range-based self-attention network model, and the scheme fully utilizes space-time context information of the track of the mobile user, so that a complex flow mode of the mobile user can be effectively modeled, and further, the problem of predicting the track of the user by utilizing a complex behavior mode of a limited data mining user constrained by space-time factors is solved, so that the current mobile service platform can accurately and timely predict the track of the mobile user, infer the demands of the user and provide corresponding high-quality personalized services.
According to a first aspect of an embodiment of the present invention, there is provided a method for mining and predicting a spatio-temporal trajectory of a mobile user, including:
preprocessing the acquired historical space-time track data of the current mobile user to obtain a track sequence ordered in time and a space-time context attribute set;
embedding the track sequence and the user characteristics to obtain track sequence characteristics and user characteristics respectively;
inputting the track sequence with the embedded features into a hierarchical attention architecture taking a pre-trained self-attention neural network as a backbone network to obtain long-term track features and short-term track features;
based on the embedded user characteristics, carrying out characteristic fusion on the long-term track characteristics and the short-term track characteristics to obtain fusion track characteristics;
and processing by using a full connection layer based on the fusion track characteristics to obtain a track prediction result of the current mobile user.
Further, the preprocessing comprises a plurality of track session sequences with equal length, wherein the track session sequences are obtained by dividing the historical space-time track data of the mobile user according to fixed lengths, and the track sequences are obtained by time sequencing.
Further, the set of spatio-temporal context attributes is constructed from geographic location coordinate information and time stamps in a historical spatio-temporal trajectory of the mobile user.
Further, the track sequence after feature embedding comprises sequence position coding information, wherein the position coding is a multidimensional vector after feature embedding is carried out on each time step sequence position.
Further, the self-focusing neural network comprises a plurality of encoders, and each encoder comprises a multi-head self-focusing network and a feedforward neural network which are connected in a residual manner and are used for extracting information of different characteristic subspaces.
Further, the hierarchical attention architecture specifically performs the following operations: firstly, encoding an input track sequence by a pre-trained self-attention network to obtain a track characteristic sequence, wherein the track session sequence characteristic closest to the current time is used as a short-term track characteristic; after the remaining track session sequence features are averaged and pooled, modeling the time dependency relationship between the track session sequences through a self-attention mechanism module, and processing the output of the self-attention mechanism module and the short-term track features through a cross-attention module to obtain long-term track features.
Further, the feature fusion specifically includes: based on the short-term track characteristics, the long-term track characteristics and the user characteristics, the personalized weights of the user on the short-term track characteristics and the long-term track characteristics are obtained by utilizing a multi-layer perceptron, and the final track characteristics are obtained through weighted summation.
According to a second aspect of an embodiment of the present invention, there is provided a system for mining and predicting a spatio-temporal trajectory of a mobile user, including:
a data processing module configured to: preprocessing the acquired historical space-time track data of the current mobile user to obtain a track sequence ordered in time and a space-time context attribute set;
a feature embedding module configured to: embedding the track sequence and the user characteristics to obtain track sequence characteristics and user characteristics respectively;
a feature extraction module configured to: inputting the track sequence with the embedded features into a hierarchical attention architecture taking a pre-trained self-attention neural network as a backbone network to obtain long-term track features and short-term track features;
a feature fusion module configured to: based on the embedded user characteristics, carrying out characteristic fusion on the long-term track characteristics and the short-term track characteristics to obtain fusion track characteristics;
a spatiotemporal trajectory prediction module configured to: and processing by using a full connection layer based on the fusion track characteristics to obtain a track prediction result of the current mobile user.
According to a third aspect of the embodiment of the present invention, there is provided an electronic device, including a memory, a processor and a computer program running on the memory, where the processor implements the method for mining and predicting a space-time trajectory of a mobile user when executing the program.
According to a fourth aspect of embodiments of the present invention, there is provided a non-transitory computer readable storage medium having stored thereon a computer program which when executed by a processor implements the method of mining and predicting spatiotemporal trajectories of mobile users.
The one or more of the above technical solutions have the following beneficial effects:
(1) The invention provides a method and a system for mining and predicting a space-time track of a mobile user, wherein the scheme is characterized in that feature extraction is carried out on space-time track data of the mobile user, a space-time context attribute set is constructed, a self-supervision paradigm-based process for pre-training a self-attention network model is adopted, space-time context information can be integrated into the sequence features of the track of the user, the characterization effect is effectively enhanced, and the sparse problem of the track data of the user is relieved.
(2) The hierarchical attention architecture adopted in the scheme can effectively mine the movement modes and the periodicity of different time scales in the user track sequence; meanwhile, the adopted feature fusion method can effectively capture personalized preferences of users on track features of different time scales, and effective prediction information is extracted.
(3) The scheme can effectively solve the problem of predicting the user track by utilizing the complex behavior mode of the limited data mining user constrained by space-time factors, so that the mobile service platform can accurately and timely predict the track of the mobile user, infer the requirement of the user and provide corresponding high-quality personalized service.
Additional aspects of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention.
FIG. 1 is a flow chart of a method for mining and predicting a space-time trajectory of a mobile user according to an embodiment of the present invention;
FIG. 2 is a flow chart of training a self-care network model based on a self-supervision paradigm according to one embodiment of the present invention.
Detailed Description
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present invention.
Embodiments of the invention and features of the embodiments may be combined with each other without conflict.
Term interpretation:
four-fork key (quadkey): in a map tile system, four-way keys are numeric references to a geographic region at a particular map zoom level, each four-way key uniquely identifying a tile at the current zoom level. Given the GPS coordinates of a certain location and the map zoom level, the four-way key may be calculated by microsoft, which has to be applied to the map tile system.
Mobile service platform: refers to a comprehensive platform that provides various services and applications based on mobile internet technology, including but not limited to social networking, map navigation, online entertainment, and a variety of other applications. The mobile service platform is mainly based on mobile terminal equipment, such as smart phones, tablet computers and the like, and provides connection and interaction between users and services.
Embodiment one:
an object of the present embodiment is to provide a method for mining and predicting a space-time trajectory of a mobile user.
As shown in fig. 1, to solve the problems existing in the prior art, the present embodiment provides a method for mining and predicting a space-time trajectory of a mobile user, including:
step 1: preprocessing the acquired historical space-time track data of the current mobile user to obtain a track sequence ordered in time and a space-time context attribute set;
step 2: embedding the track sequence and the user characteristics to obtain track sequence characteristics and user characteristics respectively;
step 3: inputting the track sequence with the embedded features into a hierarchical attention architecture taking a pre-trained self-attention neural network as a backbone network to obtain long-term track features and short-term track features;
step 4: based on the embedded user characteristics, carrying out characteristic fusion on the long-term track characteristics and the short-term track characteristics to obtain fusion track characteristics;
step 5: and processing by using a full connection layer based on the fusion track characteristics to obtain a track prediction result of the current mobile user.
The above technical solution of the present embodiment can predict a location that a user wants to access at time t+1 according to information such as a track sequence and a space-time context of the user before time t.
In the step 1, the obtained historical space-time track data of the current mobile user is preprocessed to obtain a track sequence ordered by time and a space-time context attribute set;
in an implementation, the mobile user historical space-time trajectory data is sign-in information of the user on the existing mobile service platform, for example: the historical spatiotemporal trajectory of mobile user u may be represented as T (u) = { c 1 ,c 2 ,c 3 ,…c t And (c), where c i E T (u) is a triplet (l) i ,t i ,g i ) Representing the user at a time stamp t i Visit site l i ，g i ＝<longitude,latitude>For site l i Longitude and latitude coordinate information of (a).
In one or more embodiments, in order to better model the flow pattern and periodicity of the user track, the historical space-time track data of the mobile user is divided into a plurality of track session sequences with equal length by a preset length k (k is a positive integer, and can be specifically set according to actual requirements).
In a specific embodiment, the space-time context attribute set is constructed by a time stamp and GPS coordinates of a user check-in position, specifically: discretizing the user sign-in timestamp into 48 different time attributes respectively representing 24 hours of weekdays and weekends; the GPS coordinates are coded into four-way keys (quadkey) by a map tile system and the spatial attributes of all sites are obtained by sliding window segmentation. For example, a particular place l may be encoded as a four-way key "03201011" at a map tile system resolution level of 8, and may be partitioned into "0320, 3201, 2010, 0101, 1011" at a sliding window size of 4, where each string of length 4 is considered a spatial attribute of the current place. All temporal attributes and all spatial attributes constructed by places constitute a spatio-temporal context attribute set a= { a 1 ,a 2 ,…a |A| }. For example, in the special case of a dataset with only the above-mentioned location l (its quadtree is "03201011" at resolution level 8), the constructed set of attributes A is timeThe union of the attribute sets {1,2, … } and the spatial attribute sets { "0320", "3201", "2010", "0101", "1011" }.
In the step 2, feature embedding is carried out on the track sequence and the user to respectively obtain track sequence features and user features;
in an implementation, the track sequence comprises a plurality of track session sequences of equal length.
The feature embedding process is implemented by an embedding layer, and is used for mapping the high-latitude one-hot code to the hidden vector (feature embedding vector) in the low-latitude space, and the hidden vector contains feature information extracted from the original data. The embedded layer comprises four feature embedded matrixes W L 、W A 、W U 、W P Feature embedding for location, spatiotemporal context attributes, user, and sequence position, respectively, in a location embedding matrix W L For example, each element therein represents a site-specific embedded vector. For a certain track session sequence in a given user track sequence, the sequence position i can be expressed as after feature embedding
Wherein->
Is a d-latitude embedded vector representing the location and sequence position, respectively, where d is a super parameter that can be determined experimentally. For user u, the embedded representation of the features, i.e. the user features e u ∈W U 。
In the step 3, the track sequence with embedded features is input into a hierarchical attention architecture taking a pre-trained self-attention neural network as a backbone network, and long-term track features and short-term track features are obtained;
in an implementation, the track sequence comprises a plurality of track session sequences of equal length.
As shown in fig. 1, the hierarchical attention architecture mainly performs the following operations:
first of all to pretrain the input user track sequenceThe self-care network codes, and the specific steps are as follows: track sequence S for a given user u u ＝{s 1 ,s 2 ,…s n Track session sequence s i ∈S u Can be expressed as after being coded
Wherein the nearest track session sequence feature +.>
For short-term trajectory characteristics H short The method comprises the steps of carrying out a first treatment on the surface of the In addition, the remaining n-1 track session sequence features are averaged and pooled and input to a self-attention mechanism module (i.e., the attribute structure of the tranformer Encoder) to model the time dependence between track session sequences to get->
Finally, by H short As a query, H' long Input of a Cross Attention module (i.e. Cross Attention structure of a transform decoder) as key (value) and value (value) results in long-term trajectory characteristics for each time step
In a specific embodiment, the pre-training process of the self-attention neural network model is specifically:
the track sequence with the embedded features is subjected to random masking according to preset fixed probability (specifically, the track sequence can be set according to actual requirements), then is input into a self-attention neural network to obtain time dependency characteristics, and the network is pre-trained by predicting the space-time context attribute of each time step.
Wherein the stochastic masking process is similar to the classical model Bert in natural language processing, randomly masking certain positions in the input sequence with a fixed probability, which are embedded as the same learnable vector.
The self-focusing neural network comprises N encoders, which are similar to Transformer Encoder and only differ in parameter setting, and each encoder comprises a multi-head self-focusing network and a feedforward neural network which are connected in a residual manner so as to effectively extract information of different characteristic subspaces.
Wherein each track session sequence in the track sequence is obtained after the self-attention neural network coding
k is a fixed length of the sequence of track sessions. The encoded result is transformed as follows:
wherein W is a learnable linear transformation matrix, sigma is a sigmoid activation function,
a transpose of the matrix is embedded for the spatiotemporal context attribute. The result H after the transformation is processed N′ Input to the binary cross entropy loss function (Binary Cross Entropy Loss) for pre-training, which process can be regarded as a multi-label classification task, output for the ith time step +.>
The label is a space-time context attribute set A corresponding to the current time step i E A. The pretraining process predicts and optimizes the real space-time context attribute of each time step by setting proper number of rounds (epochs) and Batch Size (Batch Size), and fuses space-time information into the model to enhance the characterization performance of the model.
In the step 4, based on the embedded user characteristics, characteristic fusion is carried out on the long-term track characteristics and the short-term track characteristics, and fusion track characteristics are obtained;
in the specific implementation, the long-term track feature, the short-term track feature and the embedded user feature are input into a fusion device to obtain a final track feature, and the track prediction result of the current mobile user is output after the final track feature is processed by a full-connection layer.
Wherein the input of the fusion device is the short-term track characteristic H of the user u short Long-term trajectory characteristics H long User characteristics e u ∈W U The output is the final trajectory feature, the process of which can be expressed as follows:
wherein concat col Concat row Representing connection vectors from the column and row directions, respectively. MLP is a common multi-layer perceptron structure, consisting of a linear layer and an active layer (ReLU).
The method is characterized in that the fusion track characteristics of the user at the time step t in the sequence are obtained by weighting and summing the long-term track characteristics and the short-term track characteristics of the user at the time step t, so that the dynamic dependence of the user on different characteristics can be captured, and the prediction effect is improved. />
The method comprises the step of predicting information of k time steps of a current track sequence of a user, namely final fusion track characteristics.
In the step 5, based on the fusion track characteristics, the full connection layer is utilized for processing, and a track prediction result of the current mobile user is obtained.
In particular implementations, as shown in FIG. 1, the final fused track feature H final The score of each candidate prediction place is obtained after the full connection layer processing
l is the number of candidate prediction sites.
In one or more embodiments, different loss functions can be used in the training process of the model, so that the training of the neural network is more controllable, and the robustness of the model is improved;
according to the scheme, the space-time context information can be integrated into the user track sequence characteristics by carrying out characteristic extraction on the user space-time track data on the mobile service platform and constructing a space-time context attribute set and adopting a process based on self-supervision paradigm pre-training, so that the characterization effect is effectively enhanced, and the sparseness problem of the user track data is relieved; the hierarchical network structure can effectively mine the movement modes and the periodicity of different time scales in the user track sequence; the fusion device can effectively capture personalized preferences of users on track features of different time scales and extract effective prediction information. Different loss functions are used in the model training process, so that the training of the neural network is more controllable, and the robustness of the model is improved.
Example two
It is an object of this embodiment to provide a system for mining and predicting a spatio-temporal trajectory of a mobile user.
A system for mining and predicting a space-time trajectory of a mobile user, comprising:
a data processing module configured to: preprocessing the acquired historical space-time track data of the current mobile user to obtain a track sequence ordered in time and a space-time context attribute set;
a feature embedding module configured to: embedding the track sequence and the user characteristics to obtain track sequence characteristics and user characteristics respectively;
a feature extraction module configured to: inputting the track sequence with the embedded features into a hierarchical attention architecture taking a pre-trained self-attention neural network as a backbone network to obtain long-term track features and short-term track features;
a feature fusion module configured to: based on the embedded user characteristics, carrying out characteristic fusion on the long-term track characteristics and the short-term track characteristics to obtain fusion track characteristics;
a spatiotemporal trajectory prediction module configured to: and processing by using a full connection layer based on the fusion track characteristics to obtain a track prediction result of the current mobile user.
Further, the system in this embodiment corresponds to the method in the first embodiment, and the technical details thereof have been described in the first embodiment, so that the details are not repeated here.
In further embodiments, there is also provided:
an electronic device comprising a memory and a processor and computer instructions stored on the memory and running on the processor, which when executed by the processor, perform the method of embodiment one. For brevity, the description is omitted here.
It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general purpose processors, digital signal processors DSP, application specific integrated circuits ASIC, off-the-shelf programmable gate array FPGA or other programmable logic device, discrete gate or transistor logic devices, discrete hardware components, or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory may include read only memory and random access memory and provide instructions and data to the processor, and a portion of the memory may also include non-volatile random access memory. For example, the memory may also store information of the device type.
A computer readable storage medium storing computer instructions which, when executed by a processor, perform the method of embodiment one.
The method in the first embodiment may be directly implemented as a hardware processor executing or implemented by a combination of hardware and software modules in the processor. The software modules may be located in a random access memory, flash memory, read only memory, programmable read only memory, or electrically erasable programmable memory, registers, etc. as well known in the art. The storage medium is located in a memory, and the processor reads the information in the memory and, in combination with its hardware, performs the steps of the above method. To avoid repetition, a detailed description is not provided herein.
It will be appreciated by those skilled in the art that the modules or steps of the invention described above may be implemented by general-purpose computer means, alternatively they may be implemented by program code executable by computing means, whereby they may be stored in storage means for execution by computing means, or they may be made into individual integrated circuit modules separately, or a plurality of modules or steps in them may be made into a single integrated circuit module. The present invention is not limited to any specific combination of hardware and software.
While the foregoing description of the embodiments of the present invention has been presented in conjunction with the drawings, it should be understood that it is not intended to limit the scope of the invention, but rather, it is intended to cover all modifications or variations within the scope of the invention as defined by the claims of the present invention.
Claims (10)
1. A method for mining and predicting a space-time trajectory of a mobile user, comprising:
preprocessing the acquired historical space-time track data of the current mobile user to obtain a track sequence ordered in time and a space-time context attribute set;
embedding the track sequence and the user characteristics to obtain track sequence characteristics and user characteristics respectively;
inputting the track sequence with the embedded features into a hierarchical attention architecture taking a pre-trained self-attention neural network as a backbone network to obtain long-term track features and short-term track features;
based on the embedded user characteristics, carrying out characteristic fusion on the long-term track characteristics and the short-term track characteristics to obtain fusion track characteristics;
and processing by using a full connection layer based on the fusion track characteristics to obtain a track prediction result of the current mobile user.
2. A method of mining and predicting spatio-temporal trajectories of mobile users as claimed in claim 1, wherein said sequence of feature-embedded trajectories includes sequence position coding information, wherein the position coding is a multidimensional vector feature-embedded for each time step sequence position.
3. A method of mining and predicting a spatio-temporal trajectory of a mobile user according to claim 1, characterized in that said self-attention neural network comprises a plurality of encoders, each comprising a multi-headed self-attention network and a feed-forward neural network connected in a residual manner for extracting information of different characteristic subspaces.
4. The method for mining and predicting spatiotemporal trajectories of mobile users according to claim 1, wherein said hierarchical attention architecture performs the following operations: firstly, encoding an input track sequence by a pre-trained self-attention network to obtain a track characteristic sequence, wherein the track session sequence characteristic closest to the current time is used as a short-term track characteristic; after the remaining track session sequence features are averaged and pooled, modeling the time dependency relationship between the track session sequences through a self-attention mechanism module, and processing the output of the self-attention mechanism module and the short-term track features through a cross-attention module to obtain long-term track features.
5. The method for mining and predicting space-time trajectories of mobile users according to claim 1, wherein said feature fusion is specifically: based on the short-term track characteristics, the long-term track characteristics and the user characteristics, the personalized weights of the user on the short-term track characteristics and the long-term track characteristics are obtained by utilizing a multi-layer perceptron, and the final track characteristics are obtained through weighted summation.
6. The method for mining and predicting a spatiotemporal trajectory of a mobile user according to claim 1, wherein said preprocessing comprises dividing historical spatiotemporal trajectory data of the mobile user by a fixed length to obtain a plurality of equal-length trajectory session sequences, and time-ordering to obtain a trajectory sequence.
7. The method of mining and predicting a spatiotemporal trajectory of a mobile user of claim 1, wherein the set of spatiotemporal context attributes is constructed from geographic location coordinate information and time stamps in a historical spatiotemporal trajectory of the mobile user.
8. A system for mining and predicting a space-time trajectory of a mobile user, comprising:
a data processing module configured to: preprocessing the acquired historical space-time track data of the current mobile user to obtain a track sequence ordered in time and a space-time context attribute set;
a feature embedding module configured to: embedding the track sequence and the user characteristics to obtain track sequence characteristics and user characteristics respectively;
a feature extraction module configured to: inputting the track sequence with the embedded features into a hierarchical attention architecture taking a pre-trained self-attention neural network as a backbone network to obtain long-term track features and short-term track features;
a feature fusion module configured to: based on the embedded user characteristics, carrying out characteristic fusion on the long-term track characteristics and the short-term track characteristics to obtain fusion track characteristics;
a spatiotemporal trajectory prediction module configured to: and processing by using a full connection layer based on the fusion track characteristics to obtain a track prediction result of the current mobile user.
9. An electronic device comprising a memory, a processor and a computer program stored for execution on the memory, wherein the processor implements a method of mining and predicting a spatio-temporal trajectory of a mobile user when executing the program of any one of claims 1-7.
10. A non-transitory computer readable storage medium having stored thereon a computer program, which when executed by a processor implements a method of mining and predicting a spatiotemporal trajectory of a mobile user as claimed in any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310331408.6A CN116304987A (en) | 2023-03-28 | 2023-03-28 | Mining and predicting method and system for space-time track of mobile user |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310331408.6A CN116304987A (en) | 2023-03-28 | 2023-03-28 | Mining and predicting method and system for space-time track of mobile user |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116304987A true CN116304987A (en) | 2023-06-23 |
Family
ID=86813122
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310331408.6A Pending CN116304987A (en) | 2023-03-28 | 2023-03-28 | Mining and predicting method and system for space-time track of mobile user |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116304987A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117216614A (en) * | 2023-09-22 | 2023-12-12 | 哈尔滨工业大学 | Track characterization mining method based on space-time information extraction |
-
2023
- 2023-03-28 CN CN202310331408.6A patent/CN116304987A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117216614A (en) * | 2023-09-22 | 2023-12-12 | 哈尔滨工业大学 | Track characterization mining method based on space-time information extraction |
| CN117216614B (en) * | 2023-09-22 | 2024-03-08 | 哈尔滨工业大学 | Track characterization mining method based on space-time information extraction |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Bidirectional GRU networks‐based next POI category prediction for healthcare | |
| CN111400620B (en) | User trajectory position prediction method based on space-time embedded Self-orientation | |
| CN108804646B (en) | Point of interest sign-in prediction method integrating deep learning and factorization machine | |
| CN113139140B (en) | Tourist attraction recommendation method based on space-time perception GRU and combined with user relationship preference | |
| Zou et al. | Air quality prediction based on a spatiotemporal attention mechanism | |
| CN112035743A (en) | Data recommendation method and device, computer equipment and storage medium | |
| CN113782096B (en) | Method and device for predicting unpaired probability of RNA (ribonucleic acid) base | |
| CN116304987A (en) | Mining and predicting method and system for space-time track of mobile user | |
| Wang et al. | A hybrid Markov and LSTM model for indoor location prediction | |
| CN111695019A (en) | Method and device for identifying associated account | |
| CN114579882A (en) | Address query method, method for obtaining geocoding prediction model and corresponding device | |
| CN111949877A (en) | Personalized interest point recommendation method and system | |
| Wen et al. | MSSRM: A multi-embedding based self-attention spatio-temporal recurrent model for human mobility prediction | |
| CN113269379B (en) | Method and device for determining attributes of resource objects, storage medium and computer equipment | |
| CN116090504A (en) | Training method and device for graphic neural network model, classifying method and computing equipment | |
| CN106600053B (en) | User attribute prediction system based on space-time trajectory and social network | |
| Wang et al. | A Spatial‐Temporal Self‐Attention Network (STSAN) for Location Prediction | |
| Sun et al. | Tcsa-net: a temporal-context-based self-attention network for next location prediction | |
| CN116257704A (en) | Point-of-interest recommendation method based on user space-time behaviors and social information | |
| CN116049887A (en) | Privacy track release method and device based on track prediction | |
| CN115662517A (en) | Sequence prediction method, device, medium, and electronic apparatus | |
| Yang et al. | Deep Learning‐Based Destination Prediction Scheme by Trajectory Prediction Framework | |
| CN115482019A (en) | Activity attention prediction method and device, electronic equipment and storage medium | |
| CN113487163A (en) | Method and device for service prediction based on geographical location information | |
| CN115455276A (en) | Method and device for recommending object, computer equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |