Position feature determination method and device, equipment, storage medium and program product

Abstract

The invention discloses a position feature determination method and device, equipment, a storage medium and a program product. The position feature determination method comprises the steps that moving position information of a moving object is acquired; feature extraction is carried out on the moving position information, a base station position vector corresponding to each time step of the moving object is obtained, and the moving position information comprises a plurality of time steps; processing each base station position vector to obtain a hidden state vector corresponding to each time step; and determining a position feature vector according to each hidden state vector. According to the invention, the extraction efficiency of the position features is improved.

Description

[0001] Technology Neighborhood

[0002] The present application belongs to the field of data processing technology, and in particular, relates to a method, device, equipment, storage medium and program product for determining location features. Background Art

[0003] The movement trajectory of mobile objects such as terminals and vehicles can help the model predict the location of the mobile object at the next moment. The movement trajectory can be characterized by the location features of the mobile object.

[0004] The position feature includes the movement pattern and position movement law of the moving object, which can be used to train the model so that the model can more accurately predict the position of the moving object at the next moment.

[0005] In the exemplary technology, the position feature is extracted by using manual feature engineering. Manual extraction of position features has the problem of low efficiency in extracting position features. Summary of the invention

[0006] The embodiments of the present application provide a method, apparatus, device, storage medium and program product for determining position features to solve the problem of low efficiency in extracting position features.

[0007] In a first aspect, an embodiment of the present application provides a method for determining a location feature, comprising:

[0008] Obtaining the moving position information of the moving object;

[0009] Performing feature extraction on the mobile position information to obtain a base station position vector corresponding to the mobile object at each time step, the mobile position information including a plurality of the time steps;

[0010] Processing each of the base station position vectors to obtain a hidden state vector corresponding to each of the time steps;

[0011] A position feature vector is determined according to each of the hidden state vectors.

[0012] In one embodiment, the step of determining a position feature vector according to each of the hidden state vectors comprises:

[0013] Determine, according to the mobile position information, a grid corresponding to the mobile object at each time step;

[0014] Determine the spatial feature vector of the mobile object according to each of the grids;

[0015] A position feature vector is determined according to the spatial feature vector and each of the hidden feature vectors.

[0016] In another embodiment, the step of determining the spatial feature vector of the moving object according to each of the grids comprises:

[0017] Determine a neighboring grid corresponding to the current grid, and obtain occupancy information and movement trend information of the neighboring grid;

[0018] Clustering each of the neighborhood grids to obtain multiple clusters of neighborhood grids, and determining neighborhood grid clustering features corresponding to each cluster of neighborhood grids;

[0019] Determine the neighborhood grid occupancy characteristics according to the occupancy information, and determine the neighborhood grid movement trend characteristics according to the movement trend information;

[0020] Determine a neighborhood grid feature vector corresponding to the current grid according to neighborhood grid occupancy characteristics, neighborhood grid movement trend characteristics, and neighborhood grid clustering characteristics corresponding to the current grid;

[0021] A spatial feature vector is determined according to the neighborhood grid feature vector.

[0022] In another embodiment, the step of determining the position feature vector according to the neighborhood grid feature vector comprises:

[0023] Acquire a first category distribution feature of the interest points corresponding to the current grid, and determine a second category distribution feature of the interest points of a neighborhood grid corresponding to the current grid;

[0024] Determining a point of interest feature vector according to the first category distribution feature and the second category distribution feature;

[0025] A spatial feature vector is determined according to the interest point feature vector and the neighborhood grid feature vector.

[0026] In another embodiment, the step of determining the position feature vector according to the neighborhood grid feature vector comprises:

[0027] Acquire a first category attribute of the current grid, and determine a second category attribute of a neighborhood grid of the current grid;

[0028] Constructing a geographic semantic feature vector according to the first category attributes and the second category attributes;

[0029] A spatial feature vector is determined according to the geographic semantic feature vector and the neighborhood grid feature vector.

[0030] In another embodiment, the step of determining a position feature vector according to each of the hidden state vectors comprises:

[0031] The hidden state vector corresponding to the latest time step is determined as the position feature vector.

[0032] In another embodiment, the step of determining a position feature vector according to each of the hidden state vectors comprises:

[0033] An average value or a weighted sum of the hidden state vectors is determined as a position feature vector.

[0034] In another embodiment, the step of processing each of the base station position vectors to obtain a hidden state vector corresponding to each of the time steps includes:

[0035] Get the first hidden state vector corresponding to the previous time step;

[0036] A second hidden state vector corresponding to the current time step is determined according to the first hidden state vector and the base station position vector corresponding to the current time step.

[0037] In another embodiment, the step of determining the second hidden state vector corresponding to the current time step according to the first hidden state vector and the base station position vector corresponding to the current time step includes:

[0038] Determine an input gate parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and the first mapping relationship;

[0039] Determine a forget gate parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and the second mapping relationship;

[0040] Determine an output gate parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and a fourth mapping relationship;

[0041] Determine a candidate memory cell state parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and a fourth mapping relationship;

[0042] A second hidden state vector corresponding to the current time period is determined according to the input gate parameter, the forget gate parameter, the output gate parameter, and the candidate memory cell state parameter.

[0043] In another embodiment, the step of determining the second hidden state vector corresponding to the current time step according to the first hidden state vector and the base station position vector corresponding to the current time step includes:

[0044] Determine an update gate parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and the fifth mapping relationship;

[0045] Determine the reset gate parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and the sixth mapping relationship.

[0046] Determine a candidate hidden state parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and a seventh mapping relationship;

[0047] A second hidden feature vector corresponding to the current time step is determined according to the update gate parameter, the reset gate parameter and the candidate hidden state parameter.

[0048] In a second aspect, an embodiment of the present application provides a device for determining a position feature, including:

[0049] An acquisition module, used for acquiring the moving position information of the moving object;

[0050] An extraction module, used for performing feature extraction on the mobile position information to obtain a base station position vector corresponding to the mobile object at each time step, wherein the mobile position information includes a plurality of the time steps;

[0051] A processing module, used to process each of the base station position vectors to obtain a hidden state vector corresponding to each of the time steps;

[0052] A determination module is used to determine a position feature vector according to each of the hidden state vectors.

[0053] In a third aspect, an embodiment of the present application provides a device for determining a location feature, including: a processor and a memory storing computer program instructions;

[0054] When the processor executes the computer program instructions, the method for determining position characteristics as described in one aspect is implemented.

[0055] In a fourth aspect, an embodiment of the present application provides a computer storage medium, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method for determining position features as in the first aspect is implemented.

[0056] In a fifth aspect, an embodiment of the present application provides a computer program product. When instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes the location feature determination method as in the first aspect.

[0057] The location feature determination method, apparatus, device, storage medium and program product provided in the embodiments of the present application obtain the mobile location information of the mobile object, perform feature extraction on the mobile location information to obtain the base station location vector corresponding to the mobile object at each time step, and process each base station location vector to obtain the hidden state vector corresponding to each time step, and determine the location feature vector based on each hidden feature vector. In the present application, multiple base station location vectors in a time series are obtained by performing feature extraction on the mobile location information, so that the location feature vector is automatically extracted by the hidden state vector obtained by processing each base station location vector, without the need for manual extraction of location features, thereby improving the efficiency of location feature extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0059] Figure 1 It is a schematic diagram of a scene involved in realizing the method for determining location features of the present application;

[0060] Figure 2 It is a flowchart of the first embodiment of the method for determining position features of the present application;

[0061] Figure 3 is a flow chart of a second embodiment of the method for determining position features of the present application;

[0062] Figure 4 is a flow chart of a third embodiment of the method for determining position features of the present application;

[0063] Figure 5 is a flow chart of a fourth embodiment of the method for determining position features of the present application;

[0064] Figure 6 is a flowchart of a fifth embodiment of the method for determining position features of the present application;

[0065] Figure 7 is a flow chart of a sixth embodiment of the method for determining position features of the present application;

[0066] Figure 8 is a schematic diagram of functional modules of a position feature determination device provided in an embodiment of the present application;

[0067] Fig. 9 It is a structural diagram of a location feature determination device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0068] The features and exemplary embodiments of various aspects of the present application will be described in detail below. In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is only to provide a better understanding of the present application by illustrating the examples of the present application.

[0069] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "include..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0070] The movement trajectory of mobile objects such as terminals and vehicles can help the model predict the location of the mobile object at the next moment. The movement trajectory can be characterized by the location features of the mobile object.

[0071] The position feature includes the movement pattern and position movement law of the moving object, which can be used to train the model so that the model can more accurately predict the position of the moving object at the next moment.

[0072] The inventors of the present application discovered that position features are extracted by artificial feature engineering, and that artificial extraction of position features has the problem of low efficiency in extracting position features.

[0073] The inventor of the present application therefore conceived of obtaining multiple base station position vectors in a time series by performing feature extraction on the mobile position information, and then automatically extracting the position feature vector from the hidden state vector obtained by processing each base station position vector, eliminating the need for manual extraction of position features and improving the efficiency of position feature extraction.

[0074] The embodiments of the present application provide a method, apparatus, device, storage medium and program product for determining location features. The following first describes a clustering description of the framework to which the method for determining location features provided by the embodiments of the present application can be applied.

[0075] like Figure 1As shown, the position feature determination device 100 obtains the mobile position information of the mobile object from the database 200. The position feature determination device 100 performs feature extraction on the mobile position information to obtain the base station position vector corresponding to the mobile object at each time step. The position feature determination device 100 processes the base station position vector to obtain the hidden state vector corresponding to each time step, thereby determining the position feature vector through each hidden state vector. The position feature determination device 100 outputs the position feature vector. For example, the position feature determination device 100 transmits the position feature vector to the device 300 for training the model, so that the device 300 for training the model can train a model for predicting the position of other mobile objects based on the position feature vector.

[0076] The following combination Figure 1 , the location feature determination method provided in the embodiment of the present application is introduced.

[0077] Figure 2 FIG. 2 is a flow chart showing a first embodiment of the method for determining position features provided by the present application. Figure 2 As shown, the method for determining the position feature includes:

[0078] Step S201, obtaining the moving position information of the moving object.

[0079] In this embodiment, the execution subject is a location feature determination device. For the convenience of description, the device is used below to refer to the location feature determination device. The device can be a server or any terminal device with a location feature extraction function.

[0080] The device obtains the mobile position information of the mobile object. The mobile object can be a user terminal, a vehicle-mounted terminal, etc. in a mobile state that is connected to the base station signal. The mobile position information refers to the information composed of the mobile position of the mobile object at different time steps. The time step is a timestamp. The mobile position information includes the position at different time steps, that is, the mobile position information includes multiple time steps.

[0081] Step S202, extracting features from the mobile position information to obtain a base station position vector corresponding to the mobile object at each time step, wherein the mobile position information includes multiple time steps.

[0082] After obtaining the mobile location information, feature extraction is performed on the mobile location information to obtain the base station location vector corresponding to the mobile object at each time step. Exemplarily, the device obtains the ID of the base station or the base station location coordinates in the record corresponding to each timestamp, performs unique hot encoding on the ID of the base station, and converts the base station location coordinates into a real-valued vector. Each base station location vector is represented as X=(x 1 ,x 2 ,...,x T ), where x tRepresents the base station position vector at the tth time step.

[0083] Furthermore, after obtaining the mobile location information, the mobile location information is cleaned and preprocessed, that is, noise and outliers are removed, as follows:

[0084] Remove duplicate records: Delete duplicate data in mobile location information;

[0085] Eliminate invalid records: remove records with missing key fields or abnormal positions;

[0086] Timestamp normalization: convert timestamps in different formats or time zones into a unified timestamp representation;

[0087] Position coordinate denoising: Smooth the base station position coordinates to remove high-frequency noise.

[0088] Step S203, processing each base station position vector to obtain a hidden state vector corresponding to each time step.

[0089] After obtaining each base station position vector, each base station position vector is processed to obtain a hidden state vector corresponding to each time step.

[0090] Exemplarily, each base station position vector X=(x 1 ,x 2 ,...,x T ) is used as input and input into the recurrent neural network for feature extraction to obtain the hidden state vector corresponding to each time step. The hidden feature vector contains the temporal dependency and context information of the trajectory sequence up to the current time step.

[0091] Step S204: determining a position feature vector according to each hidden state vector.

[0092] After obtaining each hidden state vector, a position feature vector is determined based on each hidden state vector.

[0093] In one example, the hidden state vector corresponding to the latest time step is used as the position feature vector.

[0094] In another example, the average value or weighted sum of the hidden state vectors is determined as the position feature vector.

[0095] In this embodiment, the position feature vector is a base station zipper trajectory feature vector, and the position feature vector includes the movement mode and position transfer rule of the mobile object at the base station level.

[0096] In this embodiment, the mobile position information of the mobile object is obtained, and the mobile position information is feature extracted to obtain the base station position vector corresponding to the mobile object at each time step, and each base station position vector is processed to obtain the hidden state vector corresponding to each time step, and the position feature vector is determined based on each hidden feature vector. In this embodiment, multiple base station position vectors in a time series are obtained by feature extraction of the mobile position information, so that the position feature vector is automatically extracted by the hidden state vector obtained by processing each base station position vector, without the need for manual extraction of position features, thereby improving the extraction efficiency of position features.

[0097] Reference Figure 3 , Figure 3 The flowchart of the second embodiment of the method for determining position features of the present application is shown. Based on the first embodiment, step S204 includes:

[0098] Step S301, determining the grid corresponding to the moving object at each time step according to the moving position information.

[0099] In this embodiment, the constructed position feature vector is used for training the model, so the position feature vector needs to be extracted more accurately.

[0100] In this regard, the device determines the grid corresponding to the mobile object at each time step based on the mobile position information. Specifically, the area on the device map is divided into a plurality of grids, each grid is defined as a grid, and the mobile object moves between each grid.

[0101] The device can map the MR (Measurement Report) fingerprint positioning sequence (mobile position information) into a grid sequence, where the grid sequence G = (g 1 ,g 2 ,...,g T ), where g T Indicates the grid ID where the moving object is located at the tth time step.

[0102] Step S302: determining the spatial feature vector of the moving object according to each grid.

[0103] The grid ID is a unique identifier of the grid, and the unique identifier is, for example, the row and column numbers of the grid or the hash value of the row and column numbers.

[0104] Map each grid ID to an embedded vector of fixed length, represented as Where W e is the grid embedding matrix, g t is the unique hot vector representation of the grid ID at the tth time step, and the grid embedding vector The feature representation of the grid where the user is located at the t-th time step. After obtaining each grid feature vector, the device can determine the spatial feature vector according to each grid feature vector. Exemplarily, each grid feature vector is fused to obtain a fused feature. Since the fused feature is obtained by fusion of the grid feature vectors, the grid feature vector has a spatial property, and thus the fused feature can be used as a spatial feature vector.

[0105] Step S303: determining a position feature vector according to the spatial feature vector and each hidden feature vector.

[0106] After obtaining the spatial feature vector, the device constructs a temporal feature vector based on each hidden feature vector, so that the spatial feature vector and the temporal feature vector are fused to obtain the position feature vector. For example, the temporal feature vector can be obtained by fusion of each hidden feature vector.

[0107] Exemplarily, each hidden state vector is represented as h seq =[h 1 ,h 2 ,...,h T ], the device performs a pooling operation on each hidden state vector to obtain a time series feature vector of fixed length. The time series feature vector is represented by hpooled = Pooling (hseq). Common pooling operations include maximum pooling, average pooling, etc. pooled The position feature vector can be obtained by concatenating the spatial feature vector, which is the MR fingerprint positioning feature vector.

[0108] In this embodiment, the position feature vector can be used for the subsequent construction of a large position model and position sensing applications, such as those used in offline scenarios, such as commuting route planning. The above-mentioned use of the hidden state vector of the latest time step and the average or weighted sum of each hidden state vector as the position feature vector is applied to real-time scenarios, such as the tracking scenario of a moving object.

[0109] In this embodiment, the position feature vector includes multi-dimensional information such as the time dependence and spatial relationship of the position of the mobile object, which can comprehensively characterize the position behavior pattern of the mobile object. The position feature vector can be used for the subsequent construction of a large position model and position sensing applications.

[0110] Reference Figure 4 , Figure 4 The schematic diagram of the process of the third embodiment of the method for determining position features provided by the present application is shown. Figure 3 In the illustrated embodiment, step S302 includes:

[0111] Step S401, determining a neighboring grid corresponding to the current grid, and obtaining occupancy information and movement trend information of the neighboring grid.

[0112] In this embodiment, each grid has adjacent grids, and the adjacent grids are defined as neighborhood grids.

[0113] The neighborhood range is defined as the surrounding N×N grids with the current grid of the moving object as the center, where N is an odd number. The neighborhood range can be expressed as where g t Indicates the grid where the moving object is located at the tth time step, g i,j Relative to g t The (i,j)th neighborhood grid of .

[0114] The mobile position information includes the occupancy status and movement trend of each grid. The occupancy status refers to whether the current grid is occupied by a mobile object, and the movement trend refers to the trend of the mobile object moving from the current grid to the neighboring grid. The device obtains the occupancy information and movement trend information of the neighboring grid from the mobile position information.

[0115] Step S402 : clustering each neighborhood grid to obtain multiple clusters of neighborhood grids, and determining the neighborhood grid clustering features corresponding to each cluster of neighborhood grids.

[0116] After determining the neighborhood grids, the device will cluster each neighborhood grid. Specifically, grids with similar functions or similar behavior patterns are clustered into one category. For example, the grid location can be used for parking, and the behavior pattern is similar, for example, the mobile object stays in the grid.

[0117] The device clusters each neighborhood grid to obtain multiple clusters of neighborhood grids, and determines the neighborhood grid clustering features corresponding to each cluster of neighborhood grids.

[0118] The neighborhood grid clustering feature can be expressed as c i,j =cluster(g i,j ), where cluster(·) represents the clustering function, which maps the neighborhood grid clustering features to the corresponding category labels.

[0119] Step S403: determining the neighborhood grid occupancy characteristics according to the occupancy information, and determining the neighborhood grid movement trend characteristics according to the movement trend information.

[0120] After the device obtains the occupancy information, it determines the neighborhood grid occupancy characteristics based on the occupancy information. Specifically, for each neighborhood grid g i,j ∈N(g t ), calculate the neighborhood grid occupancy characteristics o i,j , indicating whether the grid is occupied by the user at the current time step.

[0121] The neighborhood grid occupancy feature can be expressed as o i,j =I(g i,j =g t ), where I(·) is an indicative function, which takes the value of 1 when the condition is met, otherwise it takes the value of 0. For example, when the neighborhood grid is occupied, it takes the value of 1, and when it is not occupied, it takes the value of 0.

[0122] The device can also determine the neighborhood grid movement trend characteristics based on the movement trend information. Specifically, for each neighborhood grid g i,j ∈N(g t ), calculate its moving trend feature m i,j , indicating the trend of the moving object moving from the current grid to the neighborhood grid.

[0123] The movement trend characteristics of the neighborhood grid can be calculated by counting the frequency of the moving object moving from the current grid to the neighborhood grid in the historical trajectory, that is, Where count(g t → i,j ) represents the moving object from grid g t Move to grid g i,j The number of times.

[0124] Step S404, determining a neighborhood grid feature vector corresponding to the current grid according to the neighborhood grid occupancy feature, the neighborhood grid movement trend feature and the neighborhood grid clustering feature corresponding to the current grid.

[0125] After obtaining the neighborhood grid occupancy characteristics, the neighborhood grid movement trend characteristics and the neighborhood grid clustering characteristics, the device determines the neighborhood grid feature vector based on the neighborhood grid occupancy characteristics, the neighborhood grid movement trend characteristics and the neighborhood grid clustering characteristics.

[0126] Specifically, the neighborhood grid occupancy feature, the neighborhood grid movement trend feature, and the neighborhood grid clustering feature are combined into a feature vector, which is represented by fi,j=[oi,j,m i,j ,c i,j ] The current grid has multiple neighboring grids. The feature vectors of each neighboring grid corresponding to the current grid are concatenated to obtain the neighborhood grid feature vector

[0127] Step S405, determining a spatial feature vector according to the neighborhood grid feature vector.

[0128] After the neighborhood grid feature vectors are obtained, the spatial feature vectors are determined based on the neighborhood grid feature vectors. Exemplarily, the spatial feature vectors can be obtained by concatenating the neighborhood grid feature vectors corresponding to each grid.

[0129] In this embodiment, the device obtains multiple features of the domain grid based on the domain grid of the current grid, thereby obtaining spatial features with precise spatial relationships based on each feature.

[0130] Reference Figure 5 , Figure 5 The flowchart of the fourth embodiment of the method for determining position features provided by the present application is shown. Figure 4 In the illustrated embodiment, step S405 includes:

[0131] Step S501, obtaining the first category distribution characteristics of the interest points corresponding to the current grid, and determining the second category distribution characteristics of the interest points of the neighboring grid corresponding to the current grid.

[0132] In this embodiment, for each grid, it is necessary to consider the category distribution characteristics of POI (Point of Interest) within its area. The point of interest contains information such as name, category, coordinates, and classification. The point of interest is marked in the map in advance. After the map is divided into grids, each grid (grid) has the corresponding category and other information of the point of interest. In this regard, the device obtains the first category distribution characteristics of the point of interest corresponding to the current grid, and determines the second distribution characteristics of the point of interest of the neighborhood grid corresponding to the current grid. The device can directly obtain the category distribution characteristics from the grid information corresponding to the grid.

[0133] Step S502: determining a feature vector of an interest point according to the first category distribution feature and the second category distribution feature.

[0134] After obtaining the first category distribution feature and the second category distribution feature, the interest point feature vector can be determined.

[0135] Exemplarily, the device fuses the POI category distribution feature vector dg in the grid and the neighborhood POI category distribution feature vector dN(g) to obtain a comprehensive POI category distribution feature vector. The fusion method uses an attention mechanism, dfused=αdg+(1-α)dN(g), where α is the fusion weight.

[0136] The device normalizes the comprehensive POI category distribution feature vector to obtain the point of interest feature vector. The point of interest feature vector meets certain numerical range and distribution requirements. The normalization is, for example: where |·| 2 represents the L2 norm.

[0137] Step S503: determining a spatial feature vector according to the feature vector of the point of interest and the feature vector of the neighborhood grid.

[0138] After obtaining the interest point feature vector, a spatial feature vector can be constructed based on the interest point feature vector and the neighborhood grid feature vector. Exemplarily, the interest point feature vector and the neighborhood grid feature vector corresponding to each grid are concatenated to obtain the spatial feature vector.

[0139] In this embodiment, the interest point feature vector is determined based on the category distribution characteristics of the interest points of the current grid and the category distribution characteristics of the neighborhood grid, and the spatial features with precise spatial relationship are obtained based on the interest point feature vector and the neighborhood grid feature vector.

[0140] Reference Figure 6 , Figure 6 A schematic diagram showing a flow chart of a fifth embodiment of a method for determining position features provided by the present application is shown. Figure 4 In the illustrated embodiment, step S405 includes:

[0141] Step S601, obtaining the first category attribute of the current grid, and determining the second category attribute of the neighborhood grid of the current grid.

[0142] In this embodiment, the spatial feature vector needs to consider the geographic semantic feature. The device obtains the first category attribute of the current grid and the second category attribute of the neighboring grid corresponding to the current grid.

[0143] Specifically, define a geographic semantic category set S = s 1 ,s 2 ,...,s M , where M represents the total number of geographic semantic categories. Geographic semantic categories can include commercial areas, residential areas, industrial areas, parks, schools, etc. For example, if the location corresponding to the current grid on the map is in the commercial area, then the first category attribute corresponds to the semantic category corresponding to the commercial area; if the location corresponding to the neighborhood grid on the map is in the industrial area, then the second category attribute corresponds to the semantic category corresponding to the industrial area.

[0144] For example, for each grid g, a mapping function f is used to map it to a corresponding geographic semantic category to obtain a first category attribute and a second category attribute. The mapping function can be expressed as f(g)=s i , where s i ∈S, represents the geographic semantic category to which the grid g belongs. The mapping function f can be constructed based on the grid’s POI composition, demographics, land use and other characteristics through rules, clustering, classification and other methods.

[0145] Step S602: constructing a geographic semantic feature vector according to the first category attributes and the second category attributes.

[0146] For each grid g, construct an M-dimensional geographic semantic feature vector s g, represents the geographic semantic category to which the raster belongs. The geographic semantic feature vector is represented by one-hot encoding, s g = [0,...,0,1,0,...,0], where the i-th element is 1, indicating that the grid g belongs to the geographic semantic category s i , other elements are 0. The element is the first category attribute.

[0147] For each grid g, consider the geographic semantic features in its neighborhood. Define the neighborhood of grid g as N(g), and the geographic semantic feature vector of the grid in the neighborhood as s g′ ,g′∈N(g). Aggregate the geographic semantic feature vectors of all grids in the neighborhood to obtain the neighborhood geographic semantic feature vector where |N(g)| represents the number of grids in the neighborhood.

[0148] The grid's geographic semantic feature vector sg and the neighborhood geographic semantic feature vector sN(g) are fused to obtain comprehensive geographic semantic features. The fusion method uses the attention mechanism, sfused=βsg+(1-β)sN(g), where β is the fusion weight.

[0149] The fused geographic semantic features are embedded and the high-dimensional one-hot encoding vector is mapped to a low-dimensional dense vector space to obtain the final geographic semantic feature vector. Embedding can be achieved by training a feedforward neural network or using a pre-trained word embedding model, such as e s =Wesfused, where W e is the embedding matrix, e s is the embedded geographic semantic feature vector.

[0150] Step S603, determining a spatial feature vector according to the geographic semantic feature vector and the neighborhood grid feature vector.

[0151] After obtaining the geographic semantic feature vector, the spatial feature vector can be obtained by concatenating the geographic semantic feature vector with the neighborhood raster feature vector.

[0152] In addition, the semantic feature vector, the neighborhood grid feature vector and the interest point feature vector may be concatenated to obtain a spatial feature vector.

[0153] In this embodiment, the geographic semantic feature vector is determined based on the category attributes of the grid and the category attributes of the domain grid, and then the spatial features with accurate spatial relationships are obtained based on the semantic feature vector and the neighborhood grid feature vector.

[0154] Reference Figure 7 , Figure 7 A schematic diagram showing a flow chart of a sixth embodiment of a method for determining position features provided by the present application is shown. Figures 2 to 6In any of the illustrated embodiments, step S203 includes:

[0155] Step S701, obtaining the first hidden state vector corresponding to the previous time step.

[0156] Step S702: Determine a second hidden state vector corresponding to the current time step according to the first hidden state vector and the base station position vector corresponding to the current time step.

[0157] In this embodiment, the device obtains the first hidden state vector corresponding to the previous time step, and thereby determines the second hidden state vector corresponding to the current time step based on the first hidden state vector and the base station position vector corresponding to the current time step. The previous time step and the current time step are adjacent timestamps, and the previous time step is earlier than the current time step.

[0158] In one example, the device determines input gate parameters based on the first hidden state vector, the base station position vector corresponding to the current time step, and the first mapping relationship.

[0159] Input gate parameter i t =σ(Wi·[ht-1,x t ]+b i ), xt is the base station position vector of the current time step and ht-1 is the hidden state of the previous time step, σ, Wi and bi are constants, and the above formula constitutes the first mapping relationship.

[0160] The device determines a forget gate parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and the second mapping relationship. t =σ(Wf·[ht-1,x t ]+b f ), σ, Wf and bf are constants, and the above formula constitutes the second mapping relationship.

[0161] The device determines the output gate parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and the third mapping relationship. Output gate parameter o t =σ(W O ·[ht-1, x t ]+b o ), σ, Wo and bo are constants, and the above formula constitutes the third mapping relationship.

[0162] The device determines the candidate memory cell state parameters based on the first hidden state vector, the base station position vector corresponding to the current time step and the fourth mapping relationship.

[0163] Output gate parameter C t =tanh(W c [h t-1 , xt ]+b c ), W c 、b c is a constant, and the above formula constitutes the fourth mapping relationship.

[0164] The device determines the second hidden state vector corresponding to the current time step based on the input gate parameter, the forget gate parameter, the output gate parameter, and the candidate memory cell state parameter. Specifically, the memory cell state parameter C is determined based on the input gate parameter, the forget gate parameter, and the candidate memory cell state parameter, C=f t *C t-1 -1+i t *C t The device determines the second hidden state vector by output gate parameters and memory cell state parameters. The second hidden state vector h t =o t tanh(C).

[0165] In another example, the device determines the update gate parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and the fifth mapping relationship. Exemplarily, the update gate parameter z t =σ(Wz·[ht-1,x t ]+b z ), where σ, W z 、b z is a constant, and the formed formula is the fifth mapping relationship.

[0166] The device determines the reset gate parameters based on the first hidden state vector, the base station position vector corresponding to the current time step, and the sixth mapping relationship.

[0167] For example, resetting the gate parameter r t =σ(Wr·[ht-1,x t ]+b r ), where σ, W r 、b r is a constant, and the formed formula is the sixth mapping relationship.

[0168] The device determines the candidate hidden state parameters based on the first hidden state vector, the base station position vector corresponding to the current time step and the seventh mapping relationship.

[0169] For example, the candidate hidden state parameter h t =tanh(W h ·[rt*ht-1,x t ]+b h ), b h is a constant, and the formed formula is the seventh mapping relationship.

[0170] The device determines a second hidden feature vector corresponding to the current time step based on the update gate parameters, the reset gate parameters and the candidate hidden state parameters.

[0171] For example, the second hidden state parameter h t =(1-zt)*ht-1+z t *h t .

[0172] In this embodiment, the device can accurately determine the second hidden state vector corresponding to the current time step based on the first hidden state parameter of the previous time step and the base station position vector corresponding to the current time step.

[0173] Based on the above embodiments, the present application has the following advantages:

[0174] 1. According to the spatiotemporal characteristics of mobile location information, a gridded spatiotemporal division (raster) and data preprocessing method is proposed to transform the original discrete trajectory data into regular time series data, reducing data redundancy and noise and improving the efficiency and accuracy of location feature extraction.

[0175] 2. The gated recurrent unit (the above-mentioned input gate, output gate and other gated recurrent units) adaptively learns the temporal features and mines the temporal dependency and regional association patterns of the trajectory of the moving object. Compared with the traditional static feature extraction method, it has stronger feature expression and generalization capabilities.

[0176] 3. An end-to-end spatiotemporal feature extraction framework is proposed, which realizes the automatic extraction and optimization from raw location data to high-level feature representation. The framework is universal and scalable, and can flexibly embed different data processing, feature learning and model optimization modules to adapt to different data formats and application scenarios.

[0177] The method proposed in this application is expected to break through the bottleneck of traditional location big data analysis, fully tap the value of spatiotemporal trajectory data, and empower operators' intelligent operations and refined management. At the same time, this method can also be applied to location feature extraction in the fields of transportation, e-commerce, public safety, etc.

[0178] like Figure 8 As shown, the embodiment of the present application further provides a position feature determination device 800, and the position feature determination device 800 includes:

[0179] An acquisition module 810 is used to acquire the moving position information of the moving object;

[0180] An extraction module 820 is used to extract features from the mobile position information to obtain a base station position vector corresponding to the mobile object at each time step, and the mobile position information includes multiple time steps;

[0181] A processing module 830 is used to process each base station position vector to obtain a hidden state vector corresponding to each time step;

[0182] The determination module 840 is used to determine the position feature vector according to each hidden state vector.

[0183] In one embodiment, the location feature determination device 800 is further configured to:

[0184] According to the mobile position information, determine the grid corresponding to the mobile object at each time step;

[0185] Determine the spatial feature vector of the moving object according to each grid;

[0186] The position feature vector is determined based on the spatial feature vector and each hidden feature vector.

[0187] In one embodiment, the location feature determination device 800 is further configured to:

[0188] Determine the neighboring grid corresponding to the current grid, and obtain the occupancy information and movement trend information of the neighboring grid;

[0189] Clustering each neighborhood grid to obtain multiple clusters of neighborhood grids, and determining the neighborhood grid clustering features corresponding to each cluster of neighborhood grids;

[0190] Determine the neighborhood grid occupancy characteristics based on the occupancy information, and determine the neighborhood grid movement trend characteristics based on the movement trend information;

[0191] Determine the neighborhood grid feature vector corresponding to the current grid according to the neighborhood grid occupancy characteristics, neighborhood grid movement trend characteristics and neighborhood grid clustering characteristics corresponding to the current grid;

[0192] The spatial feature vector is determined based on the neighborhood grid feature vector.

[0193] In one embodiment, the location feature determination device 800 is further configured to:

[0194] Obtaining a first category distribution feature of the interest point corresponding to the current grid, and determining a second category distribution feature of the interest point of the neighboring grid corresponding to the current grid;

[0195] Determine a feature vector of the point of interest according to the first category distribution feature and the second category distribution feature;

[0196] The spatial feature vector is determined according to the feature vector of the interest point and the feature vector of the neighborhood grid.

[0197] In one embodiment, the location feature determination device 800 is further configured to:

[0198] Get the first category attribute of the current grid, and determine the second category attribute of the neighboring grid of the current grid;

[0199] According to the first category attributes and the second category attributes, a geographic semantic feature vector is constructed;

[0200] The spatial feature vector is determined according to the geographic semantic feature vector and the neighborhood grid feature vector.

[0201] In one embodiment, the location feature determination device 800 is further configured to:

[0202] The hidden state vector corresponding to the latest time step is determined as the position feature vector.

[0203] In one embodiment, the location feature determination device 800 is further configured to:

[0204] Determine the average or weighted sum of the hidden state vectors as the position feature vector.

[0205] In one embodiment, the location feature determination device 800 is further configured to:

[0206] Get the first hidden state vector corresponding to the previous time step;

[0207] A second hidden state vector corresponding to the current time step is determined according to the first hidden state vector and the base station position vector corresponding to the current time step.

[0208] In one embodiment, the location feature determination device 800 is further configured to:

[0209] Determine an input gate parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and the first mapping relationship;

[0210] Determine the forget gate parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and the second mapping relationship

[0211] Determine an output gate parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and a fourth mapping relationship;

[0212] Determine a candidate memory cell state parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and a fourth mapping relationship;

[0213] According to the input gate parameters, the forget gate parameters, the output gate parameters and the candidate memory cell state parameters, a second hidden state vector corresponding to the current time period is determined.

[0214] In one embodiment, the location feature determination device 800 is further configured to:

[0215] Determine an update gate parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and the fifth mapping relationship;

[0216] Determine a reset gate parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and a sixth mapping relationship;

[0217] Determine a candidate hidden state parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and the seventh mapping relationship;

[0218] A second hidden feature vector corresponding to the current time step is determined according to the update gate parameters, the reset gate parameters and the candidate hidden state parameters.

[0219] Fig. 9 A schematic diagram of the hardware structure of a location feature determination device provided in an embodiment of the present application is shown.

[0220] The location feature determination device may include a processor 901 and a memory 902 storing computer program instructions.

[0221] Specifically, the processor 901 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0222] The memory 902 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 902 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, the memory 902 may include a removable or non-removable (or fixed) medium. In appropriate cases, the memory 902 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 902 is a non-volatile solid-state memory.

[0223] In certain embodiments, the memory 902 may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk storage media device, an optical storage media device, a flash memory device, an electrical, optical or other physical / tangible memory storage device. Thus, in general, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to an aspect of the present disclosure.

[0224] The processor 901 implements any one of the location feature determination methods in the above embodiments by reading and executing computer program instructions stored in the memory 902 .

[0225] In one example, the location feature determination device may further include a communication interface 903 and a bus 910. Fig. 9 As shown, the processor 901, the memory 902, and the communication interface 903 are connected via a bus 910 and communicate with each other.

[0226] The communication interface 903 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiments of the present application.

[0227] Bus 910 includes hardware, software or both, and the parts of online data flow billing equipment are coupled to each other. For example, but not limitation, bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front-end bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 910 may include one or more buses. Although the present application embodiment describes and shows a specific bus, the present application considers any suitable bus or interconnection.

[0228] In addition, in combination with the terminal device movement and location feature determination method in the above embodiments, the present application embodiment can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when the computer program instructions are executed by the processor, any one of the location feature determination methods in the above embodiments is implemented.

[0229] An embodiment of the present application also provides a computer program product, including a computer program, which, when executed, implements any one of the location feature determination methods in the above embodiments.

[0230] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and a technician in this field can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.

[0231] The functional blocks shown in the above block diagram can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present application are programs or code segments that are used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link by a data signal carried in a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, etc. The code segment can be downloaded via a computer network such as the Internet, an intranet, etc.

[0232] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.

[0233] Aspects of the present disclosure are described above with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable terminal equipment motion device to produce a machine so that these instructions executed by the processor of the computer or other programmable terminal equipment motion device enable the implementation of the function / action specified in one or more boxes of the flowchart and / or block diagram. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field programmable logic circuit. It can also be understood that each box in the block diagram and / or flowchart and the combination of boxes in the block diagram and / or flowchart can also be implemented by dedicated hardware that performs a specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0234] The above is only a specific implementation method of the present application. The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, the specific working process of the above-described system, module and unit can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here. It should be understood that the protection scope of the present application is not limited to this. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in this application, and these modifications or replacements should be included in the protection scope of the present application.

Claims

1. A method for determining position features, characterized in that: include: Obtaining the moving position information of the moving object; Performing feature extraction on the mobile position information to obtain a base station position vector corresponding to the mobile object at each time step, the mobile position information including a plurality of the time steps; Processing each of the base station position vectors to obtain a hidden state vector corresponding to each of the time steps; A position feature vector is determined according to each of the hidden state vectors.

2. The method for determining position features according to claim 1, characterized in that: The step of determining the position feature vector according to each of the hidden state vectors comprises: Determine, according to the mobile position information, a grid corresponding to the mobile object at each time step; Determine the spatial feature vector of the mobile object according to each of the grids; A position feature vector is determined according to the spatial feature vector and each of the hidden feature vectors.

3. The method for determining position features according to claim 2, characterized in that: The step of determining the spatial feature vector of the moving object according to each of the grids comprises: Determine a neighboring grid corresponding to the current grid, and obtain occupancy information and movement trend information of the neighboring grid; Clustering each of the neighborhood grids to obtain multiple clusters of neighborhood grids, and determining neighborhood grid clustering features corresponding to each cluster of neighborhood grids; Determine the neighborhood grid occupancy characteristics according to the occupancy information, and determine the neighborhood grid movement trend characteristics according to the movement trend information; Determine a neighborhood grid feature vector corresponding to the current grid according to neighborhood grid occupancy characteristics, neighborhood grid movement trend characteristics, and neighborhood grid clustering characteristics corresponding to the current grid; A spatial feature vector is determined according to the neighborhood grid feature vector.

4. The method for determining position features according to claim 3, characterized in that: The step of determining the position feature vector according to the neighborhood grid feature vector comprises: Acquire a first category distribution feature of the interest points corresponding to the current grid, and determine a second category distribution feature of the interest points of a neighborhood grid corresponding to the current grid; Determining a point of interest feature vector according to the first category distribution feature and the second category distribution feature; A spatial feature vector is determined according to the interest point feature vector and the neighborhood grid feature vector.

5. The method for determining position features according to claim 3, characterized in that: The step of determining the position feature vector according to the neighborhood grid feature vector comprises: Acquire a first category attribute of the current grid, and determine a second category attribute of a neighborhood grid of the current grid; Constructing a geographic semantic feature vector according to the first category attributes and the second category attributes; A spatial feature vector is determined according to the geographic semantic feature vector and the neighborhood grid feature vector.

6. The method for determining position features according to claim 1, characterized in that: The step of determining the position feature vector according to each of the hidden state vectors comprises: The hidden state vector corresponding to the latest time step is determined as the position feature vector.

7. The method for determining position features according to claim 1, characterized in that: The step of determining the position feature vector according to each of the hidden state vectors comprises: An average value or a weighted sum of the hidden state vectors is determined as a position feature vector.

8. The method for determining position features according to any one of claims 1 to 7, characterized in that: The step of processing each of the base station position vectors to obtain a hidden state vector corresponding to each time step comprises: Get the first hidden state vector corresponding to the previous time step; A second hidden state vector corresponding to the current time step is determined according to the first hidden state vector and the base station position vector corresponding to the current time step.

9. The method for determining position features according to claim 8, characterized in that: The step of determining the second hidden state vector corresponding to the current time step according to the first hidden state vector and the base station position vector corresponding to the current time step comprises: Determine an input gate parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and the first mapping relationship; Determine a forget gate parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and the second mapping relationship; Determine an output gate parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and a fourth mapping relationship; Determine a candidate memory cell state parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and a fourth mapping relationship; A second hidden state vector corresponding to the current time period is determined according to the input gate parameter, the forget gate parameter, the output gate parameter, and the candidate memory cell state parameter.

10. The method for determining position features according to claim 8, characterized in that: The step of determining the second hidden state vector corresponding to the current time step according to the first hidden state vector and the base station position vector corresponding to the current time step comprises: Determine an update gate parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and the fifth mapping relationship; Determine a reset gate parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and a sixth mapping relationship; Determine a candidate hidden state parameter based on the first hidden state vector, the base station position vector corresponding to the current time step, and a seventh mapping relationship; A second hidden feature vector corresponding to the current time step is determined according to the update gate parameter, the reset gate parameter and the candidate hidden state parameter.

11. A device for determining position characteristics, characterized in that: include: An acquisition module, used for acquiring the moving position information of the moving object; An extraction module, used for performing feature extraction on the mobile position information to obtain a base station position vector corresponding to the mobile object at each time step, wherein the mobile position information includes a plurality of the time steps; A processing module, used to process each of the base station position vectors to obtain a hidden state vector corresponding to each of the time steps; A determination module is used to determine a position feature vector according to each of the hidden state vectors.

12. A device for determining position characteristics, characterized in that: include: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, the method for determining position characteristics as described in any one of claims 1-10 is implemented.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed by a processor, the method for determining position features as described in any one of claims 1 to 10 is implemented.

14. A computer program product, characterized in that When the instructions in the computer program product are executed by a processor of an electronic device, the electronic device is enabled to execute the location feature determination method as described in any one of claims 1-10.

Images