CN115169526B - Base station representation learning method, system and storage medium based on deep learning - Google Patents
Base station representation learning method, system and storage medium based on deep learning Download PDFInfo
- Publication number
- CN115169526B CN115169526B CN202210548996.4A CN202210548996A CN115169526B CN 115169526 B CN115169526 B CN 115169526B CN 202210548996 A CN202210548996 A CN 202210548996A CN 115169526 B CN115169526 B CN 115169526B
- Authority
- CN
- China
- Prior art keywords
- base station
- working days
- neural network
- vector set
- encoder
- 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.)
- Active
Links
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
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W64/00—Locating users or terminals or network equipment for network management purposes, e.g. mobility management
- H04W64/003—Locating users or terminals or network equipment for network management purposes, e.g. mobility management locating network equipment
-
- 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
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Signal Processing (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computer Networks & Wireless Communication (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The invention provides a base station representation learning method, a system and a storage medium based on deep learning, wherein the method comprises the following steps: acquiring base station information, wherein the base station information comprises unique identification code information of a base station, static characteristic data of the base station and relevant dynamic track data of the base station; determining an adjacency matrix between base stations based on the acquired longitude and latitude of the base stations and the base station related dynamic track data, and constructing a first base station relation diagram based on the adjacency matrix; constructing a first feature vector set based on the base station static feature data and the base station related dynamic track data; and constructing a self-encoder neural network model, inputting the first base station relation diagram and the first characteristic vector set into an encoder to obtain a base station expression vector set, and inputting the base station expression vector set into a decoder to obtain a reconstructed second base station relation diagram and second characteristic vector set. The method can efficiently realize the representation learning of the base station, thereby better applying the base station information in various data mining processes.
Description
Technical Field
The invention relates to the technical field of big data mining, in particular to a base station representation learning method, a system and a storage medium based on deep learning.
Background
The base station is an important mobile communication facility, and mobile communication users can obtain services through the base station, so that the requirements of call receiving, short message receiving and sending, mobile internet surfing and the like can be met. Therefore, the base station plays an important role in mobile communication of large data. For example, in the mobile communication track data, the base station information in a certain track point indicates the position range in which the user is located at this time.
With the development of deep learning technology, it has become mainstream to apply a deep learning model to mine mobile communication track data in recent years. However, in the existing method, the base station identification code is generally mapped to longitude and latitude information of the base station, then the longitude and latitude data is mapped to grid codes, and the mobile communication track data can be transmitted to a neural network model for training by utilizing a grid code embedding technology. Firstly, the preprocessing mode cannot realize end-to-end mobile communication data mining; secondly, because of different base station systems, environments and other factors, the area ranges represented by different base stations are also different, and the characteristics and grids of the base stations are not consistent; furthermore, compared with the grid, the base station also has the attribution information of an operation enterprise, and in the same geographic range, there may be base stations of multiple operators. And how to perform representation learning on the base station and convert the characteristics of the base station into a low-dimensional space vector so that the deep learning model can better utilize the base station information in various mining tasks (such as track prediction and abnormal base station detection), thus the method is a technical problem to be solved.
Disclosure of Invention
In view of the above, the present invention provides a method, system and storage medium for deep learning based base station representation learning, which solves one or more of the problems in the prior art.
According to one aspect of the invention, the invention discloses a base station representation learning method based on deep learning, which comprises the following steps:
acquiring base station information, wherein the base station information comprises unique identification code information of a base station, base station static characteristic data and base station related dynamic track data, the base station static characteristic data comprises position area codes, cell codes, longitudes, latitudes, addresses, affiliated operators, direction angles and POI information in a specific range, and the base station related dynamic track data comprises the accessed time of the base station, the stay time of users in the base station and the number of access users of the base station;
determining an adjacency matrix between base stations based on the acquired longitude and latitude of the base stations and the related dynamic track data of the base stations, and constructing a first base station relation diagram based on the adjacency matrix;
constructing a first feature vector set based on the base station static feature data and the base station related dynamic track data;
and constructing a self-encoder neural network model, inputting the first base station relation diagram and the first characteristic vector set into an encoder to obtain a base station expression vector set, and inputting the base station expression vector set into a decoder to obtain a reconstructed second base station relation diagram and second characteristic vector set.
In some embodiments of the present invention, the accessed time of the base station includes a time period with the most accessed times of the base station and a time period with the least accessed times of the base station;
the time for the user to stay in the base station comprises the average value and the median value of the time for the user to stay in the base station; and/or
The access user quantity of the base station comprises total number of access users of the base station on working days and non-working days, average number of access users of the base station on working days and non-working days, maximum number of access users of the base station on working days and non-working days, median number of access users of the base station on working days and non-working days and minimum number of access users of the base station on working days and non-working days.
In some embodiments of the invention, each element in the adjacency matrix is 0 or 1; when the distance between the first base station and the second base station is smaller than the preset distance, and the total number of times that the first base station and the second base station serve as adjacent track points is not smaller than the preset number of times, the corresponding element in the adjacent matrix is 1.
In some embodiments of the invention, the method further comprises:
constructing a loss function;
and updating weight parameters of the self-encoder neural network model.
In some embodiments of the invention, the loss function is: wherein V is i First characteristic direction for the ith base stationQuantity (S)>The second eigenvector of the ith base station, N is the total number of base stations, h i Is the expression vector of the ith base station, lambda is the supermodel parameter, M is the total number of neighbor base stations of the ith base station, h j Is a representation vector of a neighbor base station j of the i-th base station.
In some embodiments of the invention, the decoder includes self-attention modules Q, K, V each of which is σ (W k h k-1 ) The method comprises the steps of carrying out a first treatment on the surface of the Wherein σ is a nonlinear activation function, and a Relu function can be selected; w (W) k Is a trainable weight coefficient matrix with the size of N multiplied by L, wherein N is the number of base stations, and L is h k-1 Size W of (2) k The initial value is randomly initialized to be a real number between 0 and 1; h is a k-1 Is the output of the neural network of the k-1 layer.
In some embodiments of the invention, the encoder and the decoder each comprise three layers of a graph attention neural network layer.
In some embodiments of the invention, the method further comprises:
setting the iteration times and the learning rate of the self-encoder neural network model;
training the self-encoder neural network model.
According to another aspect of the present invention there is also disclosed a deep learning based base station representation learning system comprising a processor and a memory, said memory having stored therein computer instructions for executing the computer instructions stored in said memory, the system implementing the steps of the method according to any of the embodiments described above when said computer instructions are executed by the processor.
According to yet another aspect of the present invention, a computer-readable storage medium is also disclosed, on which a computer program is stored, which program, when being executed by a processor, carries out the steps of the method according to any of the embodiments above.
According to the base station representation learning method and system based on deep learning, the representation vector of the base station is obtained through self-encoder neural network model learning based on the first base station relation diagram and the first feature vector, namely, the base station features are projected to the low-dimensional space vector, so that the representation learning of the base station can be efficiently realized, and the base station information can be better applied in various data mining processes.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims thereof as well as the appended drawings.
It will be appreciated by those skilled in the art that the objects and advantages that can be achieved with the present invention are not limited to the above-described specific ones, and that the above and other objects that can be achieved with the present invention will be more clearly understood from the following detailed description.
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 application, illustrate and together with the description serve to explain the invention. The components in the figures are not necessarily to scale, emphasis instead being placed upon illustrating the principles of the invention. Corresponding parts in the drawings may be exaggerated, i.e. made larger relative to other parts in an exemplary device actually manufactured according to the present invention, for convenience in showing and describing some parts of the present invention. In the drawings:
fig. 1 is a flow chart of a deep learning-based base station representation learning method according to an embodiment of the invention.
Fig. 2 is a schematic structural diagram of a deep learning-based base station representation learning model according to an embodiment of the present invention.
Fig. 3 is a schematic diagram of a data updating process of an encoder of a base station representation learning model according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the embodiments of the present invention will be described in further detail with reference to the accompanying drawings. The exemplary embodiments of the present invention and their descriptions herein are for the purpose of explaining the present invention, but are not to be construed as limiting the invention.
It should be noted that, in order to avoid obscuring the present invention due to unnecessary details, only structures and/or processing steps closely related to the solution according to the present invention are shown in the drawings, while other details not greatly related to the present invention are omitted.
It should be emphasized that the term "comprises/comprising" when used herein is taken to specify the presence of stated features, elements, steps or components, but does not preclude the presence or addition of one or more other features, elements, steps or components.
In order to better perform representation learning on a base station, and convert the base station characteristics into low-dimensional space vectors, so that a deep learning model can better utilize base station information in various mining tasks.
Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the drawings, the same reference numerals represent the same or similar components, or the same or similar steps.
Fig. 1 is a schematic flow chart of a deep learning-based base station representation learning method according to an embodiment of the invention, as shown in fig. 1, the base station representation learning method at least includes steps S10 to S40.
Step S10: the method comprises the steps of obtaining base station information, wherein the base station information comprises unique identification code information of a base station, base station static characteristic data and base station related dynamic track data, the base station static characteristic data comprises position area codes, cell codes, longitudes, latitudes, addresses, affiliated operators, direction angles and POI information in a specific range, and the base station related dynamic track data comprises the accessed time of the base station, the stay time of users in the base station and the number of access users of the base station.
The base station information can be stored in advance in the base station information table or corresponding database, namelyInformation indicating the base station to be learned can be obtained, and the unique identification code information refers to the ID information of the base station. The first through N base stations can be referred to as BSs, respectively 1 ,BS 2 …BS n ，BS 1 A unique identification code representing the base station 1; correspondingly, the static characteristic data of base stations 1 to n can be respectively denoted as e 1 ,e 2 …e n The method comprises the steps of carrying out a first treatment on the surface of the The dynamic trace data associated with base stations 1 through n may be denoted as f 1 ,f 2 …f n 。
In this step, a location area code (LAC, location Area Code) is used to identify the different location areas, one location area may contain one or more cells, the location area code is contained in a location area identification code (LAI). A Cell Identifier (CI) is used as a unique identification code for a Cell, and is used in combination with a location area identification code (LAI) to identify the Cell covered by each base station in the network. The address refers to a specific location of the base station, such as the XX street XX number in XX city XX area XX street XX. The operators to which the base station belongs, such as mobile, unicom or telecommunications, in this embodiment the base station information table, each operator may be represented by a corresponding code. The direction angle of the base station is determined by the direction of the antenna in the base station, and the POI information in the specific range refers to specific information of the POI in a specific area range around the base station, wherein the POI in the embodiment can be specifically a store, a hospital, a school, a gas station and the like. It should be appreciated that the specific range may be set according to actual needs, for example, in an embodiment, the specific range may be defined as a circular area with a radius of k km around the base station.
The accessed time of the base station may include a time period in which the number of times the base station is accessed is the greatest, and a time period in which the number of times the base station is accessed is the least. At this time, the accessed data of the base station may be acquired first, for example, the number of times each base station is accessed in one day, the specific time of each access, etc.; further, the number of times of the access of each base station in each time interval (time period) in one day is counted; the time interval may be, for example, hours, and the number of times each base station is accessed in each hour of the day is counted. After counting the total number of times the base station is accessed in each hour in a day, it can be further determined that the base station is accessed in the time period with the most and least times in the day, and the time period with the most times in which the base station is accessed can be understood as the most frequent base station in the time period.
The time that the user stays at the base station may specifically include an average value of the time that the user stays at the base station and a median value. Before determining the average value and median value of the time of the user staying in the base station, the total time of each user staying in the base station and the total number of the users visited by the base station in a preset time period (such as one day) can be obtained, and then the average value and median value of the time of each user staying in the base station can be calculated based on the sum of the total time of each user staying in the base station and the total number of the users.
The number of access users of the base station comprises the total number of access users of the base station on working days and non-working days, the average number of access users of the base station on working days and non-working days, the maximum number of access users of the base station on working days and non-working days, the median number of access users of the base station on working days and non-working days and the minimum number of access users of the base station on working days and non-working days. In this embodiment, the number of access users of the base station includes workday and non-workday, because the difference between the office place covered by the base station and the entertainment place is considered to have a certain influence on the characteristics of the base station, for example, the number of access people on workday of the base station covering more office places on workday is larger than the number of access people on non-workday, and it is not easy to understand that the state of the base station covering more entertainment places is just opposite. Therefore, the semantic features of the base station can be better extracted by simultaneously acquiring the related information of the access users on working days and non-working days. For example, whether the user is on a working day or a non-working day, the number of users accessing the base station in each hour in a day can be obtained first, then the total number of users accessing the base station in a day is counted, finally the number of users accessing the base station in each hour in the day is compared, the time period with the maximum and minimum number of users accessing the base station in the day and the specific number of users in each time period can be determined, and the average number of the number of users accessing the base station in the day can be calculated based on the counted total number of users accessing the base station in the day.
Step S20: and determining an adjacency matrix between the base stations based on the acquired longitude and latitude of the base stations and the related dynamic track data of the base stations, and constructing a first base station relation diagram based on the adjacency matrix.
In this step, in order to further determine the adjacency matrix, the adjacency matrix of the base station may be denoted as M, and each element in the matrix M may be 0 or 1. Specifically, when the distance between the first base station and the second base station is smaller than the preset distance, and the total number of times of the first base station and the second base station serving as the adjacent track points is not smaller than the preset number of times, the corresponding element in the adjacent matrix is 1, and in addition, the elements in the adjacent matrix M are all marked as 0. Exemplary, when the distance between the third base station and the fourth base station is greater than the preset distance, and the total number of times that the third base station and the fourth base station serve as adjacent track points is not less than the preset number of times; or when the distance between the third base station and the fourth base station is larger than the preset distance and the times of the third base station and the fourth base station serving as adjacent track points is smaller than the preset times; the corresponding elements in the adjacency matrix M are all marked 0. It should be understood that the preset distance and the preset number of times can be defined according to actual needs; in one embodiment, the predetermined distance may be defined as 3 km, and the predetermined number of times may be set as 20 times.
Since each element in the adjacency matrix reflects the adjacent base station of the base station and the distance between the base station and the adjacent base station, the first base station relation graph is constructed with the base station as the vertex and the distance between the base station and the adjacent base station as the edge in this step.
Step S30: and constructing a first feature vector set based on the base station static feature data and the base station related dynamic track data.
In this step, the feature vector of each base station may be first determined, and then a first feature vector combination may be established based on the determined feature vector of each base station. The eigenvector composition of the base station may include: area code, cell code, longitude, latitude, address, affiliated operator, direction angle, POI information in a specific range, time zone with the most accessed times of a base station, time zone with the least accessed times of the base station, time average value and median value of stay of a user in the base station, total number of access users of the base station on working days and non-working days, average number of access users on working days and non-working days, maximum number of access users on working days and non-working days, median number of access users on working days and non-working days and minimum number of access users on working days and non-working days, and the like. The POI information in the specific range can be specifically the total number of POIs of each type in a preset range around the base station, such as the number of shopping places, the number of entertainment places, the number of dining places and the like; and in this embodiment the base station feature vector may include more or less content in addition to those listed above.
Step S40: and constructing a self-encoder neural network model, inputting the first base station relation diagram and the first characteristic vector set into an encoder to obtain a base station expression vector set, and inputting the base station expression vector set into a decoder to obtain a reconstructed second base station relation diagram and second characteristic vector set.
The method comprises the steps of constructing a self-encoder neural network model to perform representation learning on a base station, wherein the self-encoder neural network model comprises an encoder module and a decoder module, the input of the encoder module is a first base station relation diagram and a first characteristic vector set, and the output of the encoder is a vector set (the base station represents the vector set) obtained by representation learning; the input of the decoder is the base station representation vector set output by the encoder, and the reconstructed second base station relationship graph and second feature vector set output by the decoder.
In addition, in order to obtain a more ideal self-encoder neural network model, the self-encoder neural network model is further subjected to iterative training. Specifically, the base station representation learning method further includes the following steps: constructing a loss function; and updating weight parameters of the self-encoder neural network model. The loss function is used to measure the degree of inconsistency between the actual value and the predicted value learned from the encoder neural network model constructed, the smaller the loss function, the more robust the model. Specifically, when the self-encoder neural network model is built, relevant super parameters, such as the iteration times and the learning rate of the self-encoder neural network model, are also required to be set; after the number of iterations and the learning rate of the self-encoder neural network model are set, the self-encoder neural network model may also be trained based on the set number of iterations and the learning rate. The data update process of the encoder of the base station representation learning model is specifically shown in fig. 3.
In the invention, the training and optimizing targets of the self-encoder neural network model are as follows: the cost of the characteristic vector reconstruction of the base station is minimum, and the standard vector similarity between the base station and the neighbor base station is maximized. Then exemplary, the constructed loss function isWherein V is i For the first eigenvector of the ith base station, -/->The second eigenvector of the ith base station, N is the total number of base stations, h i Is the expression vector of the ith base station, lambda is the supermodel parameter, M is the total number of neighbor base stations of the ith base station, h j Is a representation vector of a neighbor base station j of the i-th base station.
In one embodiment of the present invention, the decoder includes a self-attention module, wherein Q (query), K (key), V (value) of the self-attention module are sigma (W) k h k-1 ) The method comprises the steps of carrying out a first treatment on the surface of the Wherein σ is a nonlinear activation function, and a Relu function can be selected; w (W) k Is a trainable weight coefficient matrix with the size of N multiplied by L, wherein N is the number of base stations, and L is h k-1 Size W of (2) k The initial value is randomly initialized to be a real number between 0 and 1; h is a k-1 Is the output of the neural network of the k-1 layer. In this embodiment, both the encoder and decoder may include three layers of the graph attention neural network layer, then k may take on the values 1,2,3. It will be appreciated that the inclusion of three layers of the graph attention neural network layer in both the encoder and decoder is merely a preferred example and may vary depending on the particular application scenario.
Fig. 2 is a schematic structural diagram of a base station representation learning model based on deep learning according to an embodiment of the present invention, as shown in fig. 2, a base station original relationship diagram is first constructed, the base station original relationship diagram and a corresponding base station feature vector set are input to an encoder to obtain a base station characterization vector, and then the base station characterization vector is input to a decoder to obtain a base station reconstruction relationship diagram and a corresponding base station reconstruction feature vector set.
Correspondingly, the invention also discloses a base station representation learning system based on deep learning, which comprises a processor and a memory, wherein the memory is stored with computer instructions, the processor is used for executing the computer instructions stored in the memory, and the system realizes the steps of the method according to any embodiment when the computer instructions are executed by the processor.
Illustratively, the base station represents the learning system specifically performing the following steps: base station data acquisition, base station relation diagram construction, diagram neural network construction and base station representation learning. Specifically, the base station data acquisition comprises base station code identification basic data acquisition, base station static characteristic data acquisition and base station dynamic characteristic data acquisition; the static characteristic data of the base station comprise information such as longitude and latitude data of geographic coordinates of the base station, operators to which the base station belongs, base station direction angles, base station coverage areas and the like; the static data of the base station can be obtained from text files and databases. The base station dynamic characteristic data comprises base station related dynamic track data which can be obtained from a database, a message queue and a text file.
The base station relation diagram construction is to construct a base station relation diagram according to the base station data, each vertex in the base station relation diagram represents a base station, and the base stations with correlation are connected with one side. Base station correlation may depend on the distance between base stations and the dynamic handoff relationship between base stations.
The graph neural network is constructed to construct a self-supervision graph neural network model, and comprises an encoder and a decoder, wherein the encoder is used for learning a base station representation vector from a base station relation graph, the decoder is used for reconstructing base station relation graph data from the learned base station representation vector, and self-supervision learning is performed through optimizing reconstruction loss.
The base station representation learning is specifically based on a base station relation diagram and base station data, model super parameters are set, the constructed graphic neural network model is trained, and the base station representation vector is obtained through learning.
The above method is described below in connection with a specific embodiment, however, it should be noted that this specific embodiment is only for better illustrating the present application and is not meant to be a undue limitation on the present application.
In this embodiment, the base station set D, d= { BS, is acquired first 1 ,BS 2 ,…,BS n -wherein BS i A unique identification code representing base station i; n represents the total number of base stations. Further, base station static characteristic data E, E= { (BS) is obtained 1 ,e 1 ),(BS 2 ,e 2 )…(BS n ,e n ) E, where e i Representing BS i The static feature vector of the corresponding base station i generally includes information such as location area code (LAC, location Area Code), cell Identifier (CI), longitude, latitude, address, belonging operator, direction angle, coverage, etc. In addition, according to the longitude and latitude information of the base station, POI (Point Of Interest) information in a circular area taking the base station as a circle center and k kilometers as a radius can be obtained by using the hundred-degree map service interface. And 3 is taken as a specific K, and the POI information in a round area with the base station as the center and the radius of 3 km is obtained at the moment.
And after the static characteristic data of the base station are acquired, further acquiring the relevant dynamic track data of the base station. Further, based on the base station related dynamic track data, base station dynamic feature data can be obtained, and if the base station dynamic feature data is recorded as F, f= { (BS) 1 ,f 1 ),(BS 2 ,f 2 ),…,(BS n ,f n ) Of f, where f i Representing BS i Dynamic feature vector of corresponding base station i. The base station dynamic feature vector typically includes information such as access time, dwell time, number of access users, etc. In this embodiment, the access time includes: the most frequent hours the base station is accessed and the least frequent hours the base station is accessed; the residence time includes: residence time of user at the base stationAverage, median value; the number of access users includes: the total number of access users on working days and non-working days, the average value of the access users on working days and non-working days, the maximum value of the access users on working days and non-working days, the median value of the access users on working days and non-working days, the minimum value of the access users on working days and non-working days, and the like.
Further based on the longitude and latitude data of the base stations and the dynamic characteristic data of the base stations obtained in the steps, the adjacent matrixes M and M between the base stations are obtained through calculation i，j =1 means BS i Corresponding base station i to BS j The distance of the corresponding base station j is smaller than k km, and the base station BS i With base station BS j The frequency of occurrence in adjacent track points of the same track in track data is not less than n times; otherwise M i，j =0; in this embodiment, the values of k and n are 3 and 20 respectively. Further, based on the adjacency matrix M obtained in this step, a first base station relation graph G is constructed.
Meanwhile, a first feature vector set V of the base station is constructed based on the obtained static feature data of the base station and the related dynamic track data of the base station, and the first feature vector of one base station comprises the following components: the operator code to which the base station belongs; a base station location area code; the number of POIs of various types in a circular area with the base station as a circle center and k kilometers as a radius; the most frequent hours the base station is visited; the least frequent hours the base station is visited; average value and median value of residence time of user in the base station; the total number of access users on working days and non-working days, the average value of the access users on working days and non-working days, the maximum value of the access users on working days and non-working days, the median value of the access users on working days and non-working days and the minimum value of the access users on working days and non-working days.
Based on the determined first base station relation graph G and the first eigenvector V (V e R N*d N represents the number of base stations and d represents the dimension of the first eigenvector of the base station), a graph attention self-encoder neural network model is constructed. The model comprises an encoder and a decoder; wherein the input of the encoder is a first base station relationship graph G and the top of the graph GA first set of eigenvectors V of the point (i.e., base station), the output of the encoder being a set h (hεR N*d′ N is the number of base stations, d' is the learned dimension of the base station token vector, and in this embodiment, the value is 128). The input of the decoder is a vector set h obtained by learning the base station representation, and the output is a reconstructed second base station relation diagramAnd a second set of feature vectors->Two training targets of the model are provided, namely, the cost of the reconstruction loss of the characteristic vector of the base station is minimum, and the model is +.>The other is that the standard vector similarity of the base station and the neighbor base station is maximized, and the base station is +.>The total loss function is expressed as:where N is the total number of base stations, h i Is the expression vector of the ith base station, M is the total number of neighbor base stations of the ith base station, h j For the representation vector of the neighbor base station j of the ith base station, λ is a model hyper-parameter, and λ takes 0.5 in this embodiment.
In this embodiment, the number of encoder and decoder layers is set to 3, and the information of the neural network of layer k-1 is transferred to layer k by a self-attention mechanism. Specifically, the query Q, key K, value V in attention are set to: q (Q) k ,K k ,V k ＝σ(W k h k-1 ) The weights of the base station i and other base stations are as follows:the expression vector of base station i is updated as:where Nei (i) represents the set of neighbor base stations of base station i and j represents the neighbor base stations of base station i.
Further, setting relevant super parameters to train the self-attention encoder neural network model, wherein the base station representation vector dimension is 128, namely the hidden layer size is 128, the number of layers of the coder and decoder is 3, the number of model training iterations is 100, and the learning rate is 10 -3 。
According to the base station representation learning method and system based on deep learning, the representation vector of the base station is obtained through self-encoder neural network model learning based on the first base station relation diagram and the first feature vector, namely, the base station features are projected to the low-dimensional space vector, so that the representation learning of the base station can be efficiently realized, and the base station information can be better applied in various data mining processes.
In addition, the invention also discloses a computer readable storage medium, on which a computer program is stored, which program, when being executed by a processor, implements the steps of the method according to any of the embodiments above.
Those of ordinary skill in the art will appreciate that the various illustrative components, systems, and methods described in connection with the embodiments disclosed herein can be implemented as hardware, software, or a combination of both. The particular implementation is hardware or software dependent on the specific application of the solution and the design constraints. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention. When implemented in hardware, it may be, for example, an electronic circuit, an Application Specific Integrated Circuit (ASIC), suitable firmware, a plug-in, a function card, or the like. When implemented in software, the elements of the invention are the programs or code segments used to perform the required tasks. The program or code segments may be stored in a machine readable medium or transmitted over transmission media or communication links by a data signal carried in a carrier wave. A "machine-readable medium" may include any medium that can store or transfer information. Examples of machine-readable media include electronic circuitry, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio Frequency (RF) links, and the like. The code segments may be downloaded via computer networks such as the internet, intranets, etc.
It should also be noted that the exemplary embodiments mentioned in this disclosure describe some methods or systems based on a series of steps or devices. However, the present invention is not limited to the order of the above-described steps, that is, the steps may be performed in the order mentioned in the embodiments, or may be performed in a different order from the order in the embodiments, or several steps may be performed simultaneously.
In this disclosure, features that are described and/or illustrated with respect to one embodiment may be used in the same way or in a similar way in one or more other embodiments and/or in combination with or instead of the features of the other embodiments.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, and various modifications and variations can be made to the embodiments of the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (8)
1. A method for deep learning-based base station representation learning, the method comprising:
acquiring base station information, wherein the base station information comprises unique identification code information of a base station, base station static characteristic data and base station related dynamic track data, the base station static characteristic data comprises position area codes, cell codes, longitudes, latitudes, addresses, affiliated operators, direction angles and POI information in a specific range, and the base station related dynamic track data comprises the accessed time of the base station, the stay time of users in the base station and the number of access users of the base station;
determining an adjacency matrix between base stations based on the acquired longitude and latitude of the base stations and the related dynamic track data of the base stations, and constructing a first base station relation diagram based on the adjacency matrix; the adjacency matrix is used for representing the distance between the base stations and the dynamic switching relation;
constructing a first feature vector set based on the base station static feature data and the base station related dynamic track data;
constructing a self-encoder neural network model, inputting the first base station relation diagram and the first characteristic vector set into an encoder to obtain a base station expression vector set, and inputting the base station expression vector set into a decoder to obtain a reconstructed second base station relation diagram and second characteristic vector set;
each element in the adjacency matrix is 0 or 1;
when the distance between the first base station and the second base station is smaller than a preset distance, and the total number of times of the first base station and the second base station serving as adjacent track points is not smaller than the preset number of times, the corresponding element in the adjacent matrix is 1;
the decoder includes self-attention modules Q, K, V each of which is σ (W k h k-1 ) The method comprises the steps of carrying out a first treatment on the surface of the Wherein σ is a nonlinear activation function; w (W) k Is a trainable weight coefficient matrix with the size of N multiplied by L, wherein N is the number of base stations, and L is h k-1 Size W of (2) k The initial value is randomly initialized to be a real number between 0 and 1; h is a k-1 Is the output of the neural network of the k-1 layer.
2. The deep learning-based base station representation learning method according to claim 1, wherein the accessed time of the base station includes a time period in which the number of times the base station is accessed is the largest and a time period in which the number of times the base station is accessed is the smallest;
the time for the user to stay in the base station comprises the average value and the median value of the time for the user to stay in the base station; and/or
The access user quantity of the base station comprises total number of access users of the base station on working days and non-working days, average number of access users of the base station on working days and non-working days, maximum number of access users of the base station on working days and non-working days, median number of access users of the base station on working days and non-working days and minimum number of access users of the base station on hours.
3. The deep learning-based base station representation learning method of claim 1, further comprising:
constructing a loss function;
and updating weight parameters of the self-encoder neural network model.
4. A deep learning based base station representation learning method according to claim 3, characterized in that the loss function is:wherein V is i For the first eigenvector of the i-th base station,the second eigenvector of the ith base station, N is the total number of base stations, h i Is the expression vector of the ith base station, lambda is the supermodel parameter, M is the total number of neighbor base stations of the ith base station, h j Is a representation vector of a neighbor base station j of the i-th base station.
5. The deep learning based base station representation learning method of any one of claims 1 to 4, wherein the encoder and the decoder each comprise a three-layer graph attention neural network layer.
6. The deep learning based base station representation learning method of claim 4, further comprising:
setting the iteration times and the learning rate of the self-encoder neural network model;
training the self-encoder neural network model.
7. A deep learning based base station representation learning system comprising a processor and a memory, wherein said memory has stored therein computer instructions for executing the computer instructions stored in said memory, which system when executed by the processor implements the steps of the method according to any of claims 1 to 6.
8. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210548996.4A CN115169526B (en) | 2022-05-20 | 2022-05-20 | Base station representation learning method, system and storage medium based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210548996.4A CN115169526B (en) | 2022-05-20 | 2022-05-20 | Base station representation learning method, system and storage medium based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115169526A CN115169526A (en) | 2022-10-11 |
| CN115169526B true CN115169526B (en) | 2023-08-01 |
Family
ID=83483544
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210548996.4A Active CN115169526B (en) | 2022-05-20 | 2022-05-20 | Base station representation learning method, system and storage medium based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115169526B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114007228A (en) * | 2021-11-01 | 2022-02-01 | 天津大学 | Intelligent base station control method based on heterogeneous graph neural network flow prediction |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110972062B (en) * | 2019-12-24 | 2020-10-09 | 邑客得（上海）信息技术有限公司 | Base station position parameter calibration method and system based on mobile phone signaling data |
| CN111209261B (en) * | 2020-01-02 | 2020-11-03 | 邑客得（上海）信息技术有限公司 | User travel track extraction method and system based on signaling big data |
| US11696153B2 (en) * | 2020-08-13 | 2023-07-04 | Samsung Electronics Co., Ltd. | Transfer learning of network traffic prediction model among cellular base stations |
| CN113065649B (en) * | 2021-02-22 | 2024-01-05 | 中国互联网络信息中心 | Complex network topology graph representation learning method, prediction method and server |
| CN113806463B (en) * | 2021-09-06 | 2023-04-28 | 北京信息科技大学 | Track similarity calculation method based on space-time pyramid matching |
| CN114067069A (en) * | 2021-11-16 | 2022-02-18 | 武汉中海庭数据技术有限公司 | Track representation method and system based on deep learning |
-
2022
- 2022-05-20 CN CN202210548996.4A patent/CN115169526B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114007228A (en) * | 2021-11-01 | 2022-02-01 | 天津大学 | Intelligent base station control method based on heterogeneous graph neural network flow prediction |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115169526A (en) | 2022-10-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111932036B (en) | Fine spatio-temporal scale dynamic population prediction method and system based on position big data | |
| US10034141B2 (en) | Systems and methods to identify home addresses of mobile devices | |
| US7107038B2 (en) | Method for determining if the location of a wireless communication device is within a specified area | |
| CN110414732B (en) | Travel future trajectory prediction method and device, storage medium and electronic equipment | |
| CN105554704A (en) | Fake-locus-based location privacy protection method for use in recommendation system | |
| CN109949063B (en) | Address determination method and device, electronic equipment and readable storage medium | |
| CN106951828B (en) | Urban area function attribute identification method based on satellite images and network | |
| Xu et al. | Urban noise mapping with a crowd sensing system | |
| EP3644580A1 (en) | Human mobility measuring method | |
| Hu | Rainstorm flash flood risk assessment using genetic programming: a case study of risk zoning in Beijing | |
| Zhuang et al. | Integrating a deep forest algorithm with vector‐based cellular automata for urban land change simulation | |
| Li et al. | Research on remote sensing dynamic monitoring of ecological resource environment based on GIS | |
| CN109996178B (en) | Positioning method, apparatus, device and medium | |
| CN112214677B (en) | Point of interest recommendation method and device, electronic equipment and storage medium | |
| CN115169526B (en) | Base station representation learning method, system and storage medium based on deep learning | |
| CN110992230B (en) | Full-scale demographic method, device and server based on terminal signaling data | |
| CN116528282B (en) | Coverage scene recognition method, device, electronic equipment and readable storage medium | |
| CN117079148A (en) | Urban functional area identification method, device, equipment and medium | |
| Loyola et al. | Extending battery lifetime of mobile devices with geofence capabilities on dynamic-speed urban environments | |
| KR102401823B1 (en) | System and method for computing thermal environment application index by disaster type, and weather factor of intense heat and freeze | |
| Li et al. | Outdoor location estimation in changeable environments | |
| US10989786B2 (en) | Outdoor localization using hotspot labels | |
| CN111314850B (en) | Matching method, device, equipment and medium for store and user | |
| Adegboyega et al. | GIS-based site suitability and vulnerability assessment of telecommunication base transceiver station facilities in Ibadan metropolis, Nigeria | |
| CN117042125B (en) | Base station position calibration method and system based on graph variation self-encoder |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |