KR20230157838A - Operating method of server for predicting mobility of user equipment and server thereof - Google Patents

Abstract

The server according to one embodiment may collect mobility data of user terminals and group user terminals into clusters based on the similarity of the mobility data. The server applies the identification information of the clusters to which the grouped user terminals belong and the previous movement paths of the grouped user terminals to a neural network-based prediction model, and creates a prediction model to predict the base station serving the movement locations of the user terminals belonging to each cluster. You can train.

Description

A method of operating a server for predicting mobility of a user terminal and the server {OPERATING METHOD OF SERVER FOR PREDICTING MOBILITY OF USER EQUIPMENT AND SERVER THEREOF}

The following disclosure relates to a server operation method and the server for predicting the mobility of a user terminal.

In a mobile network, a base station can repeat the process of disconnecting and reconnecting a wireless connection to reduce power consumption of a user equipment (UE) and save wireless resources. A base station in a 5G network, such as a next generation node B (gNB), predicts the mobility of a user terminal, and performs, for example, paging, handover, and/or edge computing based on the predicted mobility of the user terminal. It can perform various network functions such as (edge computing).

Machine learning models for predicting the mobility of a user terminal can predict the future location of the user terminal based on the user terminal's previous movement path. When only the previous movement path is trained in a machine learning model to predict the location of a user terminal, the prediction accuracy for the mobility of a user terminal that follows a general movement pattern is high, but the prediction accuracy for the mobility of a user terminal that does not follow a general movement pattern is low. It can be low.

In addition, in order to maintain the accuracy of the machine learning model, newly generated data must be continuously learned, and when prediction results using the machine learning model are provided for each user terminal each time, computational resources are consumed. It can get bigger.

Various embodiments may train a prediction model using a machine learning method based on the similarity of mobility data including previous movement patterns of user terminals and day of the week and time information.

Various embodiments apply the identification information of clusters to which user terminals grouped based on the similarity of mobility data and the previous movement paths of the grouped user terminals are applied to a neural network-based prediction model to determine the prediction model of the user terminals belonging to each cluster. It can be trained to predict which base station is serving the moving location.

Various embodiments determine the target cluster to which the target terminal for which mobility is to be predicted belongs, input the identification information of the target cluster and the previous movement path of the target terminal into a trained prediction model, and predict the moving location of the target terminal. Services for target terminals can be provided from mobile locations.

Various embodiments may evaluate the accuracy of the predicted movement location of the user terminal for each cluster and perform training by selecting clusters whose accuracy is lower than a certain standard.

Various embodiments may predict the mobility of a user terminal using a trained prediction model and perform various network functions such as paging, handover, and/or edge computing according to the predicted mobility.

According to one embodiment, the operating method of the server includes collecting mobility data of user terminals, grouping the user terminals into clusters based on similarity of the mobility data, and the grouping of the user terminals to which the grouped user terminals belong. By applying the identification information of clusters and the previous movement paths of the grouped user terminals to a neural network-based prediction model, the prediction model is trained to predict the base station serving the movement locations of the user terminals belonging to each cluster. It may include actions such as:

According to one embodiment, the operating method of the server includes receiving mobility data of the target terminal, determining a target cluster to which the target terminal will belong based on the mobility data, identification information of the target cluster, and the target terminal. It may include an operation of predicting the movement location of the target terminal by inputting the previous movement path of the target terminal into a trained prediction model, and an operation of providing a service for the target terminal at the predicted movement location.

According to one embodiment, the server groups the user terminals into clusters based on a communication interface for receiving mobility data of user terminals and the similarity of the mobility data, and the clusters to which the grouped user terminals belong ( By applying the identification information of the grouped user terminals and the previous movement paths of the grouped user terminals to a neural network-based prediction model, a processor trains the prediction model to predict the base station serving the movement locations of the user terminals belonging to each cluster. It can be included.

The server according to one embodiment predicts the mobility of the user terminal by learning the identification information of the clusters to which the user terminals clustered using mobility data including the user terminal's previous movement pattern, day of the week, and time information, together with the mobility data. The accuracy of the prediction model can be improved.

The server according to one embodiment predicts mobility for each cluster rather than for each user terminal, thereby reducing the computational cost and/or learning time used for training the prediction model.

The server according to one embodiment can improve prediction accuracy for user terminals with unique movement patterns by predicting mobility for each cluster rather than for individual user terminals.

According to one embodiment, instead of training all data sets when training a prediction model, the server first evaluates the accuracy of the predicted movement location of the user terminal for each cluster, and sets the accuracy to a certain standard. By selecting lower clusters and training the prediction model, training costs can be greatly reduced.

In addition, various effects that can be directly or indirectly identified through this document may be provided.

1 is a flowchart showing a method of operating a server according to an embodiment.
FIG. 2 is a diagram illustrating a method of grouping user terminals into clusters according to an embodiment.
Figure 3 is a flowchart showing a method of training a prediction model according to one embodiment.
Figure 4 is a diagram for explaining a method of training a prediction model according to an embodiment.
Figure 5 is a diagram showing the structure of a prediction model according to one embodiment.
Figure 6 is a flowchart showing a server operation method according to another embodiment.
Figure 7 is a block diagram of a server according to one embodiment.
Figure 8 is a block diagram of an electronic device in a network environment according to various embodiments.

Hereinafter, embodiments will be described in detail with reference to the attached drawings. In the description with reference to the accompanying drawings, identical components will be assigned the same reference numerals regardless of the reference numerals, and overlapping descriptions thereof will be omitted.

Electronic devices according to various embodiments disclosed in this document may be of various types. Electronic devices may include, for example, portable communication devices (e.g., smartphones), computer devices, portable multimedia devices, portable medical devices, cameras, wearable devices, or home appliances. Electronic devices according to embodiments of this document are not limited to the above-described devices.

The various embodiments of this document and the terms used herein are not intended to limit the technical features described in this document to specific embodiments, and should be understood to include various changes, equivalents, or replacements of the embodiments. In connection with the description of the drawings, similar reference numbers may be used for similar or related components. The singular form of a noun corresponding to an item may include one or more of the above items, unless the relevant context clearly indicates otherwise. As used herein, “A or B”, “at least one of A and B”, “at least one of A or B”, “A, B or C”, “at least one of A, B and C”, and “A Each of phrases such as “at least one of , B, or C” may include any one of the items listed together in the corresponding phrase, or any possible combination thereof. Terms such as "first", "second", or "first" or "second" may be used simply to distinguish one element from another, and may be used to distinguish such elements in other respects, such as importance or order) is not limited. One (e.g. first) component is said to be "coupled" or "connected" to another (e.g. second) component, with or without the terms "functionally" or "communicatively". Where mentioned, it means that any of the components can be connected to the other components directly (e.g. wired), wirelessly, or through a third component.

The term “module” used in various embodiments of this document may include a unit implemented in hardware, software, or firmware, and is interchangeable with terms such as logic, logic block, component, or circuit, for example. It can be used as A module may be an integrated part or a minimum unit of the parts or a part thereof that performs one or more functions. For example, according to one embodiment, the module may be implemented in the form of an application-specific integrated circuit (ASIC).

Various embodiments of this document are stored in a storage medium (e.g., built-in memory 836 or external memory 838) that can be read by a machine (e.g., electronic device 801 of FIG. 1). It may be implemented as software (e.g., program 840) including one or more instructions. For example, a processor (e.g., processor 820) of a device (e.g., electronic device 801) may call at least one command among one or more commands stored from a storage medium and execute it. This allows the device to be operated to perform at least one function according to the at least one instruction called. The one or more instructions may include code generated by a compiler or code that can be executed by an interpreter. A storage medium that can be read by a device may be provided in the form of a non-transitory storage medium. Here, 'non-transitory' only means that the storage medium is a tangible device and does not contain signals (e.g. electromagnetic waves), and this term refers to cases where data is semi-permanently stored in the storage medium. There is no distinction between temporary storage cases.

According to one embodiment, methods according to various embodiments disclosed in this document may be included and provided in a computer program product. Computer program products are commodities and can be traded between sellers and buyers. The computer program product is distributed in the form of a machine-readable storage medium (e.g. compact disc read only memory (CD-ROM)), or through an application store (e.g. Play StoreTM) or on two user terminals (e.g. It can be distributed (e.g. downloaded or uploaded) directly between smart phones) or online. In the case of online distribution, at least a portion of the computer program product may be at least temporarily stored or temporarily created in a machine-readable storage medium, such as the memory of a manufacturer's server, an application store's server, or a relay server.

According to various embodiments, each component (e.g., module or program) of the above-described components may include a single or plural entity, and some of the plurality of entities may be separately placed in other components. there is. According to various embodiments, one or more of the components or operations described above may be omitted, or one or more other components or operations may be added. Alternatively or additionally, multiple components (eg, modules or programs) may be integrated into a single component. In this case, the integrated component may perform one or more functions of each component of the plurality of components in the same or similar manner as those performed by the corresponding component of the plurality of components prior to the integration. . According to various embodiments, operations performed by a module, program, or other component may be executed sequentially, in parallel, iteratively, or heuristically, or one or more of the operations may be executed in a different order, or omitted. Alternatively, one or more other operations may be added.

1 is a flowchart showing a method of operating a server according to an embodiment. In the following embodiments, each operation may be performed sequentially, but is not necessarily performed sequentially. For example, the order of each operation may be changed, and at least two operations may be performed in parallel.

Referring to the flowchart 100 of FIG. 1, the server (e.g., the server 700 of FIG. 7 and/or the server 808 of FIG. 8) according to an embodiment creates a prediction model (e.g., through operations 110 to 130). : The prediction model 500 of FIG. 5) can be trained.

In operation 110, the server may collect mobility data of user terminals (eg, electronic device 801 in FIG. 8). For example, the server may collect mobility data by sampling the movement paths of user terminals at fixed time intervals, but is not necessarily limited thereto.

In operation 120, the server may group user terminals into clusters (eg, clusters 210, 230, and 250 in FIG. 2) based on the similarity of the mobility data collected in operation 110. The server may group user terminals into clusters, for example, by similarity between one or a combination of movement times and movement patterns based on mobility data. Travel time may include, but is not necessarily limited to, any one or combination of time, day of the week, date, and season, for example.

Grouping user terminals into clusters can be called 'clustering'. Clustering is one of the unsupervised learning techniques in machine learning, and can be understood as classifying similar data in a data set by grouping them into groups.

The server uses machine learning-based clustering techniques, including, for example, k-mean clustering techniques, K-nearest neighbor clustering techniques, and mean-shift clustering techniques. Thus, user terminals can be grouped into clusters, but the method is not necessarily limited to this. The server can use various clustering techniques to group user terminals according to movement patterns, time, and day of the week.

For example, when a certain data set is classified into k clusters, there may be k centroids in the data set. At this time, assigning the data included in the data set to a nearby center based on the Euclidean distance and grouping the data gathered at the same center into one cluster may correspond to the k-means clustering technique. In k-means clustering, 'mean' can be understood as the average distance between the center of each cluster and the data. The method by which the server groups user terminals into clusters will be described in more detail with reference to FIG. 2 below.

In operation 130, the server serves the movement locations of user terminals belonging to each cluster by applying the identification information of the clusters to which the user terminals grouped in operation 120 and the previous movement paths of the grouped user terminals belong to a neural network-based prediction model. A prediction model can be trained to predict base stations. Hereinafter, 'training' and 'learning' have the same meaning and can be used interchangeably. The method by which the server trains the prediction model will be described in more detail with reference to FIGS. 3 and 4 below. Additionally, the structure and operation of the prediction model according to one embodiment will be described in more detail with reference to FIG. 5 below.

FIG. 2 is a diagram illustrating a method of grouping user terminals into clusters according to an embodiment. Referring to FIG. 2, a server (e.g., server 700 in FIG. 7 and/or server 808 in FIG. 8) according to an embodiment clusters user terminals (e.g., electronic devices 801 in FIG. 8). A diagram 200 is shown to explain a method of grouping into groups.

Before training the prediction model, which is a machine learning model for predicting the mobility of the user terminal, the server subjects the user terminals to various clustering techniques, such as k-means clustering technique, K-nearest neighbor clustering technique, and mean moving clustering technique. It can be classified by:

The server may group mobility data of user terminals into a number of clusters (210, 230, 250) according to similarity using various clustering techniques. The server may group user terminals into clusters, for example, based on similarity between one or a combination of movement time and mobility pattern based on mobility data. Travel time may include, but is not necessarily limited to, any one or combination of time, day of the week, date, and season, for example. Travel patterns may include, but are not necessarily limited to, various travel routes, as well as, for example, travel within a city center, travel outside a city, and/or travel on a highway.

For example, in a 5G network, various IoT (internet of things) terminals may be added, and IoT terminals may have very diverse movement patterns. In addition, even if it is the same user terminal, for example, it may have different movement patterns depending on whether it is a weekday from Monday to Friday or a weekend such as Saturday or Sunday, and/or whether it is dawn, morning, afternoon, or night even on the same day of the week. It may be possible.

Based on the mobility data of user terminals, the server may group, for example, user terminals that show a large amount of movement within the city center during the week into a first cluster 210. The server may group user terminals that always show a high amount of movement not only in the city center but also on the outskirts of the city, regardless of the day of the week, into the second cluster 230. Additionally, the server may group user terminals that show a large amount of movement within the city center on weekends into a third cluster 250.

For example, the server assigns '1' as identification information (ID) to the first cluster 210, '2' as identification information to the second cluster 230, and '2' as identification information to the third cluster 250. '3' can be assigned as identification information, but is not necessarily limited to this. In addition, the server can assign identification information to clusters in various ways, such as 'A', 'B', 'C', or '1-1'-'1-2'. The server may grant identification information of the clusters to which the user terminal belongs to each of the user terminals grouped into each cluster.

Figure 3 is a flowchart showing a method of training a prediction model according to one embodiment. In the following embodiments, each operation may be performed sequentially, but is not necessarily performed sequentially. For example, the order of each operation may be changed, and at least two operations may be performed in parallel.

Referring to the flowchart 300 of FIG. 3, the server (e.g., the server 700 of FIG. 7 and/or the server 808 of FIG. 8) according to one embodiment creates a prediction model (e.g., through operations 310 to 330). : The prediction model 500 of FIG. 5) can be trained.

In operation 310, the server may obtain, for example, the movement location of the user terminal (e.g., the electronic device 801 in FIG. 8) predicted by a prediction model for each cluster.

In operation 320, the server may evaluate the accuracy of the movement location of the user terminal predicted for each cluster (eg, clusters 210, 230, and 250 in FIG. 2) in operation 310. The server can evaluate accuracy based on whether the movement location of the user terminal predicted for each cluster corresponds to an area served by the predicted base station. For example, if the movement position (a) of the user terminal predicted for each cluster corresponds to the area (A) served by the predicted base station, the server evaluates the accuracy as 'high' among high, medium, and low, or Alternatively, the accuracy can be evaluated as 95% or 99%. On the other hand, if the movement location (b) of the user terminal predicted for each cluster does not correspond to the area (A) served by the predicted base station, the server determines whether the predicted movement location (b) of the user terminal and the predicted base station are Depending on the distance difference or distance proportion between serving areas (A), the accuracy can be evaluated as 'medium' or 'low', or the accuracy can be evaluated as '60%' or '45%'.

In operation 330, the server may train a prediction model based on the accuracy of each cluster evaluated in operation 320. The server can determine a target cluster among clusters to use for training a prediction model based on the accuracy of each cluster. For example, if the accuracy of the corresponding cluster among the clusters is lower than a preset standard, the server may determine the corresponding cluster as the target cluster. For example, if the accuracy of the corresponding cluster among the clusters is higher than or equal to the preset standard, the server does not determine the corresponding cluster as the target cluster. At this time, the number of target clusters determined may be one or multiple depending on the accuracy. The server can train the prediction model by applying the identification information of the target cluster and the movement paths of user terminals belonging to the target cluster to the prediction model.

In one embodiment, the accuracy of the prediction model for corner cases can be improved by training the prediction model for each cluster grouped according to the various movement patterns described above.

Figure 4 is a diagram for explaining a method of training a prediction model according to an embodiment. Referring to FIG. 4, a server (e.g., server 700 of FIG. 7 and/or server 808 of FIG. 8) according to an embodiment is configured to use clusters (e.g., clusters 210, 230, and 250 of FIG. 2). )) A diagram for explaining a method of training a prediction model (e.g., the prediction model 500 in FIG. 5) according to the accuracy of the movement position of the user terminal (e.g., the electronic device 801 in FIG. 8) predicted by each. 400 is shown.

In order to maintain a certain level of accuracy in predicting the mobility of the user terminal, the server according to one embodiment may continuously learn the path that the user terminal recently moved and update the prediction model.

The server may perform clustering on a data set containing mobility data of user terminals. The server may group user terminals into clusters based on the similarity of mobility data. In one embodiment, clustering is performed using a machine learning method using the user terminal's previous movement pattern, day of the week, and/or time information, and the prediction accuracy of the prediction model is improved by training the identification information of the cluster when training the prediction model. It can be improved.

The server is connected to each grouped cluster (e.g., 1st cluster (cluster#1), 2nd cluster (cluster#2), 3rd cluster (cluster#3), .., n-th cluster (cluster#n)) The moving location of the user terminal can be predicted separately. The server can evaluate the accuracy of the movement location of the user terminal predicted for each cluster. For example, the server may evaluate accuracy based on whether the movement location of the user terminal predicted for each cluster corresponds to the area served by the base station predicted by the prediction model. If the user terminal is located in an area served by the base station, the server may evaluate accuracy highly. The server can reduce the computational cost and/or learning time used for training the prediction model by delivering prediction results for each cluster rather than for individual user terminals.

Based on the evaluated accuracy, the server can reduce the overall learning amount of the prediction model by, for example, not learning clusters with accuracy higher than the preset standard, but selecting and learning clusters with accuracy lower than or equal to the preset standard. there is.

For example, the accuracy of the first cluster predicted by the prediction model may be 98.3%, the accuracy of the second cluster may be 45.7%, the accuracy of the third cluster may be 63.1%, and the accuracy of the 100th cluster may be 80.70%. . At this time, the preset accuracy standard may be, for example, 70%.

The server compares the accuracy of each cluster predicted by the prediction model with a preset accuracy standard (e.g., 70%) and creates a prediction model based on mobility data belonging to the second cluster and the third cluster, which show accuracy lower than the standard. You can train.

In one embodiment, mobility data with similar patterns overlap in data sets classified into the same cluster, so the accuracy of the prediction model can be maintained even if only some data sets within the cluster are learned. In addition, before training the prediction model, an operation to predict the moving position of the user terminal is performed for each cluster, but in this prediction operation, the back propagation procedure of the prediction model is not performed, so the processing time and computational amount for prediction are not performed. It may not take much.

In one embodiment, when training a prediction model, instead of learning the mobility data included in all data sets, the prediction accuracy is first pre-tested for each cluster, and clusters with low accuracy are selected and learned to increase the learning amount. And/or the cost required for learning can be significantly reduced while maintaining the prediction accuracy of the prediction model.

Figure 5 is a diagram showing the structure of a prediction model according to one embodiment. Referring to FIG. 5, a simplified configuration diagram of a prediction model 500 according to one embodiment is shown.

The prediction model 500 may be, for example, a machine learning model or a deep neural network (DNN)-based model.

Prediction model 500 may include, for example, a stacked deep neural network 510 and fully connected layers 530 .

The deep neural network cells 515 included in the stacked deep neural network 510 may refer to, for example, an AI model of the RNN (recurrent neural network) series, but are not necessarily limited thereto. Deep neural network (RNN) cells 515 refer to the recurrent layer of a deep neural network, and the output results of the cells can be called a 'hidden state'. The deep neural network cells 515 may be, for example, but are not necessarily limited to recurrent neural network (RNN) cells.

The stacked deep neural network 510 receives the identification information (C) of the clusters (e.g., the clusters 210, 230, and 250 in FIG. 2) and the previous movement paths (Sn) of the grouped user terminals and Spatiotemporal characteristics of the mobility of the electronic device 801 (e.g., FIG. 8) can be learned. The stacked deep neural network 510 may be, for example, an RNN-type neural network such as long short-term memory (LSTM) or gated recurrent units (GRU), but is not necessarily limited thereto. At this time, the identification information (C) of the clusters may be determined by the server before machine learning of the prediction model 500.

The stacked deep neural network 510 may include an input layer composed of deep neural network cells 515 and hidden layers. As the identification information of the clusters (C) and the previous movement paths (S) of the grouped user terminals are applied to the input layer of the stacked deep neural network 510, the hidden layers are applied to the user terminals based on the previous movement paths (S). The moving location where the terminal is expected to move can be learned.

The stacked deep neural network 510 may be composed of, for example, three layers of recurrent neural network (RNN) cells 515, but is not necessarily limited thereto. In Figure 5, for convenience of explanation, a structure in which three layers are stacked is shown, but the structure is not necessarily limited thereto. Prediction model 500 may be a multi-layer neural network with a depth greater than three layers.

The first layer of the stacked deep neural network 510 is the identification information (C) of the clusters and the previous movement paths of the grouped user terminals. It may be an input layer for receiving input. The previous location path (Sn) of the user terminal input to each of the deep neural network cells 515 in the input layer represents previous location information at a fixed time interval, for example, sequentially as the length of the input sequence from S ₁ to Sn. can be entered. Additionally, since the identification information (C) of the clusters input to the input layer is not time series data, it is input equally to all deep neural network cells 515 and can be learned regardless of input order or location. In one embodiment, one prediction model is used by using the identification information (C) of the clusters as an input to the prediction model 500, but it provides differentiated learning and prediction for each cluster and improves the performance of the prediction model. You can do it.

The second and third layers of the stacked deep neural network 510 may operate as hidden layers that are not involved in input and output. At this time, the number of layers and the number of nodes used as hidden layers in the stacked deep neural network 510 may be set experimentally and/or empirically to values suitable for the application environment.

The output of the deep neural network cells 515 is delivered to the next neighboring deep neural network cell of the same layer and the deep neural network cells of the next layer, so the output of the last deep neural network cell of the highest layer is the final output of the stacked deep neural network 510 ( O) It can be.

The fully connected layers 530 allow user terminals to be located for each base station (gNB) from the output (O) of the stacked deep neural network 510 by an activation function, for example, a softmax activation function. A base station list 550 including probabilities can be output. In one embodiment, the softmax activation function is described as an example for convenience of explanation, but the present invention is not necessarily limited thereto, and various activation functions other than the softmax activation function may be used.

The fully connected layers 530 may receive i outputs (O) of the stacked deep neural network 510 and output hidden state values equal to j, which is the number of base stations to be predicted. Here, i may be the same as or different from n, which is the number of input values of the stacked deep neural network 510. For example, the softmax activation function can output normalized values between 0 and 1, and the sum of the output values can always be 1. Accordingly, each of the output values of the prediction model 500 may represent the probability that the user terminal will be located at the corresponding base station.

The server (e.g., the server 700 in FIG. 7 and/or the server 808 in FIG. 8) sorts the probability that the user terminal is located for each base station output through the fully connected layers 530, and the sorted probability A base station list 550 including base stations corresponding to a probability of a certain percentage (e.g., top 20%) can be output. At this time, base stations included in the base station list 550 may be base stations serving locations or areas where the user terminal is expected to move.

Figure 6 is a flowchart showing a server operation method according to another embodiment. In the following embodiments, each operation may be performed sequentially, but is not necessarily performed sequentially. For example, the order of each operation may be changed, and at least two operations may be performed in parallel.

Referring to the flowchart 600 of FIG. 6, a server (e.g., server 700 of FIG. 7 and/or server 808 of FIG. 8) according to an embodiment performs operations 610 to 640 to connect a target terminal (e.g., : The mobility of the electronic device 801 in FIG. 8 can be predicted, and services for the target terminal can be provided at the predicted movement location.

In operation 610, the server may receive mobility data of the target terminal. Mobility data may include, for example, location information including the path taken by the user terminal, and/or time information, but is not necessarily limited thereto. The location information includes identification information of the user terminal (e.g., ID of the user terminal), information about the first base station corresponding to the first location (e.g., source) from which the user terminal departs in the activation mode, and the user terminal. In the activation mode, it may include any one or a combination of information about the second base station corresponding to the second location (eg, destination) arrived at by moving from the first location, but is not necessarily limited thereto.

In operation 620, the server may determine the target cluster to which the target terminal will belong based on the mobility data received in operation 610. The server may determine the target cluster to which the target terminal belongs by comparing the result of analyzing any one or a combination of movement time and movement pattern based on mobility data with a preset cluster type.

In operation 630, the server inputs the identification information of the target cluster determined in operation 620 and the previous movement path of the target terminal into a trained prediction model (e.g., prediction model 500 in FIG. 5) to predict the movement location of the target terminal. there is. The prediction model is, for example, a neural network-based prediction model trained to predict base stations serving the location where the target terminal is expected to move based on type information of clusters and previous movement paths of user terminals belonging to each cluster. It can be included.

In operation 640, the server may provide a service for the target terminal at the movement location predicted in operation 630. Services for the target terminal may include, but are not necessarily limited to, paging, handover, and/or user plane function (UPF) re-selection in edge computing, for example. No.

Figure 7 is a block diagram of a server according to one embodiment. Referring to FIG. 7 , a server 700 (eg, server 808 in FIG. 8 ) according to an embodiment may include a communication interface 710, a processor 730, and a memory 750.

The communication interface 710, processor 730, and memory 750 may be connected to each other through a communication bus 705.

The communication interface 710 may receive mobility data of user terminals (eg, the electronic device 801 of FIG. 8).

The processor 730 may group user terminals into clusters based on the similarity of mobility data received through the communication interface 710. The processor 730 uses the identification information of the clusters to which the grouped user terminals belong (e.g., clusters 210, 230, and 250 in FIG. 2) and the previous movement paths of the grouped user terminals using a neural network-based prediction model (e.g., By applying this to the prediction model 500 of FIG. 5, the prediction model can be trained to predict the base station serving the moving locations of user terminals belonging to each cluster.

The processor 730 may obtain the movement location of the user terminal predicted by the prediction model for each cluster. The processor 730 may evaluate the accuracy of the movement location of the user terminal predicted for each cluster. The processor 730 may train a prediction model based on the accuracy of each cluster.

The processor 730 may evaluate accuracy based on whether the movement location of the user terminal predicted for each cluster corresponds to an area served by the predicted base station. The processor 730 may determine a target cluster among clusters to be used for training a prediction model based on the accuracy of each cluster. The processor 730 may train the prediction model by applying the identification information of the target cluster and the movement paths of user terminals belonging to the target cluster to the prediction model.

Depending on the embodiment, the processor 730 determines the target cluster to which the target terminal belongs based on mobility data, inputs the identification information of the target cluster and the previous movement path of the target terminal into the trained prediction model, and moves the target terminal. Location can be predicted. The processor 730 may provide services for the target terminal at the predicted movement location.

Memory 750 may store a prediction model trained by processor 730.

Additionally, the processor 730 can execute programs and control the server 700. Program code executed by the processor 730 may be stored in the memory 750.

The memory 750 may store information received from the communication interface 710. The memory 750 stores executable instructions to be executed by the processor 730. Additionally, the memory 750 can store various information generated during processing by the processor 730. In addition, the memory 750 can store various data and programs. Memory 750 may include volatile memory or non-volatile memory. The memory 750 may be equipped with a high-capacity storage medium such as a hard disk to store various data.

Additionally, the processor 730 may perform at least one method or a technique corresponding to at least one method described above with reference to FIGS. 1 to 6 . The processor 730 may be a device or server implemented in hardware that has a circuit with a physical structure for executing desired operations. For example, the intended operations may include code or instructions included in the program. For example, the server 700 implemented as hardware includes a microprocessor, a central processing unit (CPU), a graphic processing unit (GPU), a processor core, and a multi- It may include a multi-core processor, multiprocessor, application-specific integrated circuit (ASIC), field programmable gate array (FPGA), and/or neural processing unit (NPU).

Figure 8 is a block diagram of an electronic device in a network environment according to various embodiments. Referring to FIG. 8, in a network environment 800 corresponding to an example of a user terminal according to an embodiment, an electronic device 801 is connected to an electronic device (e.g., a short-range wireless communication network) through a first network 898 (e.g., a short-range wireless communication network). 802), or may communicate with the electronic device 804 or the server 808 through a second network 899 (eg, a long-distance wireless communication network). According to one embodiment, the electronic device 801 may communicate with the electronic device 804 through the server 808. According to one embodiment, the electronic device 801 includes a processor 820, a memory 830, an input module 850, an audio output module 855, a display module 860, an audio module 870, and a sensor module ( 876), interface 877, connection terminal 878, haptic module 879, camera module 880, power management module 888, battery 889, communication module 890, subscriber identification module 896 , or may include an antenna module 897. In some embodiments, at least one of these components (eg, the connection terminal 878) may be omitted, or one or more other components may be added to the electronic device 801. In some embodiments, some of these components (e.g., sensor module 876, camera module 880, or antenna module 897) are integrated into one component (e.g., display module 860). It can be.

The processor 820, for example, executes software (e.g., program 840) to operate at least one other component (e.g., hardware or software component) of the electronic device 801 connected to the processor 820. It can be controlled and various data processing or calculations can be performed. According to one embodiment, as at least part of data processing or computation, the processor 820 stores commands or data received from another component (e.g., sensor module 876 or communication module 890) in volatile memory 832. The commands or data stored in the volatile memory 832 can be processed, and the resulting data can be stored in the non-volatile memory 834. According to one embodiment, the processor 820 includes a main processor 821 (e.g., a central processing unit or an application processor) or an auxiliary processor 823 that can operate independently or together (e.g., a graphics processing unit, a neural network processing unit ( It may include a neural processing unit (NPU), an image signal processor, a sensor hub processor, or a communication processor). For example, if the electronic device 801 includes a main processor 821 and a auxiliary processor 823, the auxiliary processor 823 may be set to use lower power than the main processor 821 or be specialized for a designated function. You can. The auxiliary processor 823 may be implemented separately from the main processor 821 or as part of it.

The auxiliary processor 823 may, for example, act on behalf of the main processor 821 while the main processor 821 is in an inactive (e.g., sleep) state, or while the main processor 821 is in an active (e.g., application execution) state. ), together with the main processor 821, at least one of the components of the electronic device 801 (e.g., the display module 860, the sensor module 876, or the communication module 890) At least some of the functions or states related to can be controlled. According to one embodiment, co-processor 823 (e.g., image signal processor or communication processor) may be implemented as part of another functionally related component (e.g., camera module 880 or communication module 890). there is. According to one embodiment, the auxiliary processor 823 (eg, neural network processing unit) may include a hardware structure specialized for processing artificial intelligence models. Artificial intelligence models can be created through machine learning. For example, such learning may be performed in the electronic device 801 itself, where artificial intelligence is performed, or may be performed through a separate server (e.g., server 808). Learning algorithms may include, for example, supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning, but It is not limited. An artificial intelligence model may include multiple artificial neural network layers. Artificial neural networks include deep neural network (DNN), convolutional neural network (CNN), recurrent neural network (RNN), restricted boltzmann machine (RBM), belief deep network (DBN), bidirectional recurrent deep neural network (BRDNN), It may be one of deep Q-networks or a combination of two or more of the above, but is not limited to the examples described above. In addition to hardware structures, artificial intelligence models may additionally or alternatively include software structures.

The memory 830 may store various data used by at least one component (eg, the processor 820 or the sensor module 876) of the electronic device 801. Data may include, for example, input data or output data for software (e.g., program 840) and instructions related thereto. Memory 830 may include volatile memory 832 or non-volatile memory 834.

The program 840 may be stored as software in the memory 830 and may include, for example, an operating system 842, middleware 844, or application 846.

The input module 850 may receive commands or data to be used in a component of the electronic device 801 (e.g., the processor 820) from outside the electronic device 801 (e.g., a user). The input module 850 may include, for example, a microphone, mouse, keyboard, keys (eg, buttons), or digital pen (eg, stylus pen).

The sound output module 855 may output sound signals to the outside of the electronic device 801. The sound output module 855 may include, for example, a speaker or receiver. Speakers can be used for general purposes such as multimedia playback or recording playback. The receiver can be used to receive incoming calls. According to one embodiment, the receiver may be implemented separately from the speaker or as part of it.

The display module 860 can visually provide information to the outside of the electronic device 801 (eg, a user). The display module 860 may include, for example, a display, a hologram device, or a projector, and a control circuit for controlling the device. According to one embodiment, the display module 860 may include a touch sensor configured to detect a touch, or a pressure sensor configured to measure the intensity of force generated by the touch.

The audio module 870 can convert sound into an electrical signal or, conversely, convert an electrical signal into sound. According to one embodiment, the audio module 870 acquires sound through the input module 850, the sound output module 855, or an external electronic device (e.g., directly or wirelessly connected to the electronic device 801). Sound may be output through an electronic device 802 (e.g., speaker or headphone).

The sensor module 876 detects the operating state (e.g., power or temperature) of the electronic device 801 or the external environmental state (e.g., user state) and generates an electrical signal or data value corresponding to the detected state. can do. According to one embodiment, the sensor module 876 includes, for example, a gesture sensor, a gyro sensor, an air pressure sensor, a magnetic sensor, an acceleration sensor, a grip sensor, a proximity sensor, a color sensor, an IR (infrared) sensor, a biometric sensor, It may include a temperature sensor, humidity sensor, or light sensor.

The interface 877 may support one or more designated protocols that can be used to connect the electronic device 801 directly or wirelessly with an external electronic device (eg, the electronic device 802). According to one embodiment, the interface 877 may include, for example, a high definition multimedia interface (HDMI), a universal serial bus (USB) interface, an SD card interface, or an audio interface.

The connection terminal 878 may include a connector through which the electronic device 801 can be physically connected to an external electronic device (eg, the electronic device 802). According to one embodiment, the connection terminal 878 may include, for example, an HDMI connector, a USB connector, an SD card connector, or an audio connector (eg, a headphone connector).

The haptic module 879 can convert electrical signals into mechanical stimulation (e.g., vibration or movement) or electrical stimulation that the user can perceive through tactile or kinesthetic senses. According to one embodiment, the haptic module 879 may include, for example, a motor, a piezoelectric element, or an electrical stimulation device.

The camera module 880 can capture still images and moving images. According to one embodiment, the camera module 880 may include one or more lenses, image sensors, image signal processors, or flashes.

The power management module 888 can manage power supplied to the electronic device 801. According to one embodiment, the power management module 888 may be implemented as at least a part of, for example, a power management integrated circuit (PMIC).

Battery 889 may supply power to at least one component of electronic device 801. According to one embodiment, the battery 889 may include, for example, a non-rechargeable primary cell, a rechargeable secondary cell, or a fuel cell.

Communication module 890 is configured to provide a direct (e.g., wired) communication channel or wireless communication channel between the electronic device 801 and an external electronic device (e.g., electronic device 802, electronic device 804, or server 808). It can support establishment and communication through established communication channels. Communication module 890 operates independently of processor 820 (e.g., an application processor) and may include one or more communication processors that support direct (e.g., wired) communication or wireless communication. According to one embodiment, the communication module 890 is a wireless communication module 892 (e.g., a cellular communication module, a short-range wireless communication module, or a global navigation satellite system (GNSS) communication module) or a wired communication module 894 (e.g., : LAN (local area network) communication module, or power line communication module) may be included. Among these communication modules, the corresponding communication module is a first network 898 (e.g., a short-range communication network such as Bluetooth, wireless fidelity (WiFi) direct, or infrared data association (IrDA)) or a second network 899 (e.g., legacy It may communicate with an external electronic device 804 through a telecommunication network such as a cellular network, a 5G network, a next-generation communication network, the Internet, or a computer network (e.g., LAN or WAN). These various types of communication modules may be integrated into one component (e.g., a single chip) or may be implemented as a plurality of separate components (e.g., multiple chips). The wireless communication module 892 uses subscriber information (e.g., International Mobile Subscriber Identifier (IMSI)) stored in the subscriber identification module 896 within a communication network such as the first network 898 or the second network 899. The electronic device 801 can be confirmed or authenticated.

The wireless communication module 892 may support 5G networks after 4G networks and next-generation communication technologies, for example, NR access technology (new radio access technology). NR access technology provides high-speed transmission of high-capacity data (eMBB (enhanced mobile broadband)), minimization of terminal power and access to multiple terminals (mMTC (massive machine type communications)), or high reliability and low latency (URLLC (ultra-reliable and low latency). -latency communications)) can be supported. The wireless communication module 892 may support high frequency bands (e.g., mmWave bands), for example, to achieve high data rates. The wireless communication module 892 uses various technologies to secure performance in high frequency bands, such as beamforming, massive MIMO (multiple-input and multiple-output), and full-dimensional multiplexing. It can support technologies such as input/output (FD-MIMO: full dimensional MIMO), array antenna, analog beam-forming, or large scale antenna. The wireless communication module 892 may support various requirements specified in the electronic device 801, an external electronic device (e.g., electronic device 804), or a network system (e.g., second network 899). According to one embodiment, the wireless communication module 892 supports Peak data rate (e.g., 20 Gbps or more) for realizing eMBB, loss coverage (e.g., 164 dB or less) for realizing mmTC, or U-plane latency (e.g., 164 dB or less) for realizing URLLC. Example: Downlink (DL) and uplink (UL) each of 0.5 ms or less, or round trip 1 ms or less) can be supported.

The antenna module 897 may transmit or receive signals or power to or from the outside (e.g., an external electronic device). According to one embodiment, the antenna module 897 may include an antenna including a radiator made of a conductor or a conductive pattern formed on a substrate (eg, PCB). According to one embodiment, the antenna module 897 may include a plurality of antennas (eg, an array antenna). In this case, at least one antenna suitable for the communication method used in the communication network, such as the first network 898 or the second network 899, is connected to the plurality of antennas by, for example, the communication module 890. can be selected. Signals or power may be transmitted or received between the communication module 890 and an external electronic device through the selected at least one antenna. According to some embodiments, in addition to the radiator, other components (eg, radio frequency integrated circuit (RFIC)) may be additionally formed as part of the antenna module 897.

According to various embodiments, antenna module 897 may form a mmWave antenna module. According to one embodiment, a mmWave antenna module includes: a printed circuit board, an RFIC disposed on or adjacent to a first side (e.g., bottom side) of the printed circuit board and capable of supporting a designated high frequency band (e.g., mmWave band); And a plurality of antennas (e.g., array antennas) disposed on or adjacent to the second side (e.g., top or side) of the printed circuit board and capable of transmitting or receiving signals in the designated high frequency band. can do.

At least some of the components are connected to each other through a communication method between peripheral devices (e.g., bus, general purpose input and output (GPIO), serial peripheral interface (SPI), or mobile industry processor interface (MIPI)) and signal ( (e.g. commands or data) can be exchanged with each other.

According to one embodiment, commands or data may be transmitted or received between the electronic device 801 and the external electronic device 804 through the server 808 connected to the second network 899. Each of the external electronic devices 802 or 804 may be of the same or different type as the electronic device 801. According to one embodiment, all or part of the operations performed in the electronic device 801 may be executed in one or more of the external electronic devices 802, 804, or 808. For example, when the electronic device 801 needs to perform a certain function or service automatically or in response to a request from a user or another device, the electronic device 801 may perform the function or service instead of executing the function or service on its own. Alternatively, or additionally, one or more external electronic devices may be requested to perform at least part of the function or service. One or more external electronic devices that have received the request may execute at least part of the requested function or service, or an additional function or service related to the request, and transmit the result of the execution to the electronic device 801. The electronic device 801 may process the result as is or additionally and provide it as at least part of a response to the request. For this purpose, for example, cloud computing, distributed computing, mobile edge computing (MEC), or client-server computing technology can be used. The electronic device 801 may provide an ultra-low latency service using, for example, distributed computing or mobile edge computing. In another embodiment, the external electronic device 804 may include an Internet of Things (IoT) device. Server 808 may be an intelligent server using machine learning and/or neural networks. According to one embodiment, the external electronic device 804 or server 808 may be included in the second network 899. The electronic device 801 may be applied to intelligent services (e.g., smart home, smart city, smart car, or healthcare) based on 5G communication technology and IoT-related technology.

According to one embodiment, the operating method of the servers 700 and 808 includes an operation 110 of collecting mobility data of user terminals 801, and dividing the user terminals 801 into clusters 210 based on the similarity of the mobility data. , 230, 250, operation 120, and identification information of the clusters 210, 230, 250 to which the grouped user terminals 801 belong and previous movement of the grouped user terminals 801. By applying paths to the neural network-based prediction model 500, training the prediction model 500 to predict the base station serving the moving locations of the user terminals 801 belonging to each of the clusters 210, 230, and 250. Operation 130 may be included.

According to one embodiment, the operation of training the prediction model 500 involves obtaining the movement position of the user terminal 801 predicted by the prediction model 500 for each of the clusters 210, 230, and 250. Operation 310, evaluating the accuracy of the movement position of the user terminal 801 predicted for each cluster (210, 230, 250), and operation 320 for evaluating the accuracy for each cluster (210, 230, 250). As a basis, operation 330 of training the prediction model 500 may be included.

According to one embodiment, the operation of evaluating the accuracy is based on whether the movement location of the user terminal 801 predicted for each cluster 210, 230, and 250 corresponds to an area served by the predicted base station. This may include an operation to evaluate the accuracy.

According to one embodiment, the operation of training the prediction model 500 based on the accuracy for each cluster (210, 230, 250) is performed based on the accuracy for each cluster (210, 230, 250). An operation of determining a target cluster to be used for training the prediction model 500 among clusters 210, 230, and 250, and identification information of the target cluster and movement paths of user terminals 801 belonging to the target cluster. It may include an operation of training the prediction model 500 by applying it to the prediction model 500.

According to one embodiment, the operation of determining the target cluster includes determining the corresponding cluster as the target cluster when the accuracy of the corresponding cluster among the clusters 210, 230, and 250 is lower than a preset standard. can do.

According to one embodiment, the operation of training the prediction model 500 may include training the prediction model 500 to predict base stations serving the movement locations of user terminals 801 belonging to the target cluster. You can.

According to one embodiment, the prediction model 500 receives identification information of the clusters 210, 230, and 250 and previous movement paths of the grouped user terminals 801 to determine the mobility of the user terminal 801. Stacked deep neural networks 510 that learn spatio-temporal features, and a base station list 550 including the probability that the user terminal 801 is located for each base station from the output of the stacked deep neural networks 510. Output may include fully connected layers 530 (fully connected layers).

According to one embodiment, the stacked deep neural network 510 includes an input layer composed of deep neural network cells, and hidden layers, and the input layer includes identification information of the clusters 210, 230, and 250 and the grouping. As the previous movement paths of one user terminal 801 are applied, the hidden layers can learn the movement location where the user terminal 801 is expected to move based on the previous movement paths.

According to one embodiment, the servers 700 and 808 sort the probability that the user terminal 801 is located for each of the base stations output through the fully connected layers 530, and select a certain probability among the sorted probabilities. The base station list 550 including base stations corresponding to the probability of the ratio can be output.

According to one embodiment, the grouping operation divides the user terminals 801 into the clusters 210, 230, and 250 based on similarity between one or a combination of movement time and movement pattern based on the mobility data. May include grouping operations.

According to one embodiment, the travel time may include any one or a combination of time, day of the week, date, and season.

According to one embodiment, the grouping operation is based on machine learning including a k-mean clustering technique, a K-nearest neighbor clustering technique, and a mean-shift clustering technique. It may include grouping the user terminal 801 into the clusters 210, 230, and 250 using a clustering technique.

According to one embodiment, the operation of collecting the mobility data may include collecting the mobility data by sampling the movement paths of the user terminals 801 at fixed time intervals.

According to one embodiment, the operating method of the servers 700 and 808 includes operation 610 of receiving mobility data of a target terminal, operation 620 of determining a target cluster to which the target terminal will belong based on the mobility data, and the target cluster. Operation 630 of predicting the moving location of the target terminal by inputting the identification information and the previous movement path of the target terminal into the trained prediction model 500, and providing a service for the target terminal at the predicted moving location. Operation 640 may be included.

According to one embodiment, the operation of determining the target cluster is performed by comparing the result of analyzing any one or a combination of movement time and movement pattern based on the mobility data with a preset cluster type to determine the target cluster to which the target terminal belongs. It may include an operation to determine .

According to one embodiment, the prediction model 500 is based on type information of the clusters 210, 230, and 250 and previous movement paths of the user terminals 801 belonging to each of the clusters 210, 230, and 250. It may include a neural network-based prediction model 500 trained to predict base stations serving the location where the target terminal is expected to move.

According to one embodiment, the servers 700 and 808 organize the user terminals 801 into clusters ( 210, 230, 250), identification information of the clusters 210, 230, 250 to which the grouped user terminals 801 belong, and previous movement paths of the grouped user terminals 801. A processor ( 730).

According to one embodiment, the processor 730 obtains the movement position of the user terminal 801 predicted by the prediction model 500 for each of the clusters 210, 230, and 250, and Evaluate the accuracy of the movement position of the user terminal 801 predicted for each cluster (210, 230, 250), and train the prediction model 500 based on the accuracy for each cluster (210, 230, 250). can do.

According to one embodiment, the processor 730 based on whether the movement location of the user terminal 801 predicted for each cluster 210, 230, and 250 corresponds to an area served by the predicted base station. , evaluate the accuracy, and determine a target cluster to be used for training the prediction model 500 among the clusters 210, 230, 250 based on the accuracy for each cluster 210, 230, 250, and the target The prediction model 500 can be trained by applying the cluster identification information and the movement paths of the user terminals 801 belonging to the target cluster to the prediction model 500.

Claims (20)

An operation of collecting mobility data of user terminals;
Grouping the user terminals into clusters based on similarity of the mobility data; and
By applying the identification information of the clusters to which the grouped user terminals belong and the previous movement paths of the grouped user terminals to a neural network-based prediction model, the base station serving the movement locations of the user terminals belonging to each cluster is predicted. Actions to train a prediction model
A method of operating a server, including. According to paragraph 1,
The operation of training the prediction model is
Obtaining, for each cluster, a moving location of the user terminal predicted by the prediction model;
Evaluating the accuracy of the movement location of the user terminal predicted for each cluster; and
An operation of training the prediction model based on the accuracy for each cluster
A method of operating a server, including. According to paragraph 2,
The operation of evaluating the accuracy is
An operation of evaluating the accuracy based on whether the movement location of the user terminal predicted for each cluster corresponds to an area served by the predicted base station.
A method of operating a server for predicting mobility of a user terminal, including. According to paragraph 2,
The operation of training the prediction model based on the accuracy for each cluster is
determining a target cluster to be used for training the prediction model among the clusters based on the accuracy for each cluster; and
An operation of training the prediction model by applying identification information of the target cluster and movement paths of user terminals belonging to the target cluster to the prediction model.
A method of operating a server, including. According to paragraph 4,
The operation of determining the target cluster is
If the accuracy of the corresponding cluster among the clusters is lower than a preset standard, determining the corresponding cluster as the target cluster
A method of operating a server, including. According to clause 5,
The operation of training the prediction model is
An operation of training the prediction model to predict base stations serving the movement locations of user terminals belonging to the target cluster.
A method of operating a server, including. According to paragraph 1,
The prediction model is
Layered deep neural networks that receive identification information of the clusters and previous movement paths of the grouped user terminals and learn spatiotemporal characteristics of the mobility of the user terminals; and
Fully connected layers that output a base station list including the probability of the user terminal being located for each base station from the output of the stacked deep neural networks.
A method of operating a server, including. In clause 7,
The layered deep neural network is
Input layer composed of deep neural network cells, and hidden layers
Including,
As the identification information of the clusters and the previous movement paths of the grouped user terminals are applied to the input layer, the hidden layers learn the movement location where the user terminal is expected to move based on the previous movement paths. , how the server operates. In clause 7,
The server is
A server operation method of sorting the probability of the user terminal being located for each of the base stations output through the fully connected layers, and outputting the base station list including base stations corresponding to a certain percentage of the sorted probabilities. . According to paragraph 1,
The grouping operation is
An operation of grouping the user terminals into the clusters based on similarity between any one or a combination of movement time and movement pattern based on the mobility data.
A method of operating a server, including. According to clause 10,
The travel time is
A method of operating a server, including any one or a combination of time, day of the week, date, and season. According to paragraph 1,
The grouping operation is
The user terminal is configured using a machine learning-based clustering technique including a k-mean clustering technique, a K-nearest neighbor clustering technique, and a mean-shift clustering technique. Grouping operations into clusters
A method of operating a server, including. According to paragraph 1,
The operation of collecting the mobility data is
An operation of collecting the mobility data by sampling the movement paths of the user terminals at fixed time intervals
A method of operating a server, including. An operation of receiving mobility data of a target terminal;
An operation of determining a target cluster to which the target terminal will belong based on the mobility data;
Predicting the moving location of the target terminal by inputting the identification information of the target cluster and the previous movement path of the target terminal into a trained prediction model; and
An operation of providing a service for the target terminal at the predicted movement location
A method of operating a server, including. According to clause 14,
The operation of determining the target cluster is
An operation of determining a target cluster to which the target terminal will belong by comparing the result of analyzing any one or a combination of movement time and movement pattern based on the mobility data with a preset cluster type.
A method of operating a server, including. According to clause 14,
The prediction model is
A neural network-based prediction model trained to predict base stations serving the location where the target terminal is expected to move based on type information of clusters and previous movement paths of user terminals belonging to each cluster.
A method of operating a server, including. A computer program combined with hardware and stored in a computer-readable recording medium to execute the method of any one of claims 1 to 16. a communication interface for receiving mobility data of user terminals; and
Based on the similarity of the mobility data, the user terminals are grouped into clusters, and the identification information of the clusters to which the grouped user terminals belong and the previous movement paths of the grouped user terminals are applied to a neural network-based prediction model. , a processor that trains the prediction model to predict the base station serving the movement locations of user terminals belonging to each cluster.
Server, including. According to clause 18,
The processor is
For each of the clusters, obtain the moving position of the user terminal predicted by the prediction model, evaluate the accuracy of the moving position of the user terminal predicted for each cluster, and based on the accuracy for each cluster, A server that trains the prediction model. According to clause 19,
The processor is
Evaluating the accuracy based on whether the movement location of the user terminal predicted for each cluster corresponds to an area served by the predicted base station,
Determining a target cluster among the clusters to be used for training the prediction model based on the accuracy of each cluster,
A server that trains the prediction model by applying identification information of the target cluster and movement paths of user terminals belonging to the target cluster to the prediction model.

Images