Method and device for managing data in wireless communication system

Abstract

According to an embodiment of the present disclosure, a method and electronic device for managing data in a wireless communication system may be provided. The method may comprise the steps of: acquiring wireless communication system-related data; generating a knowledge graph by using the wireless communication system-related data; and adding the knowledge graph to a database.

Description

Method and device for managing data in a wireless communication system

[0001] The present disclosure relates to a method and apparatus for managing data in a wireless communication system, and more specifically, to a method of acquiring a knowledge graph using communication data, adding the knowledge graph to a database, and utilizing the knowledge graph for various tasks of a wireless communication system.

[0002] 5G mobile communication technology defines a wide frequency band to enable fast transmission speeds and new services, and can be implemented not only in frequency bands below 6 GHz ('Sub 6 GHz'), such as 3.5 gigahertz (3.5 GHz), but also in ultra-high frequency bands called millimeter waves (mmWave), such as 28 GHz and 39 GHz ('Above 6 GHz'). In addition, for 6G mobile communication technology, which is referred to as a system beyond 5G, implementation in the terahertz band (e.g., the 3 terahertz (3 THz) band at 95 GHz) is being considered to achieve transmission speeds 50 times faster and ultra-low latency reduced to one-tenth compared to 5G mobile communication technology.

[0003] In the early stages of 5G mobile communication technology, aiming to satisfy service support and performance requirements for enhanced Mobile BroadBand (eMBB), Ultra-Reliable Low-Latency Communications (URLLC), and massive Machine-Type Communications (mMTC), technologies such as beamforming and Massive MIMO to mitigate path loss and increase transmission distance in ultra-high frequency bands, support for various numerologies (such as the operation of multiple subcarrier spacings) and dynamic operation of slot formats for the efficient utilization of ultra-high frequency resources, initial access techniques to support multi-beam transmission and broadband, definition and operation of Band-Width Parts (BWP), Low Density Parity Check (LDPC) codes for high-volume data transmission, new channel coding methods such as Polar Codes for the reliable transmission of control information, and L2 pre-processing (L2 Standardization has been carried out for pre-processing, network slicing which provides a dedicated network specialized for specific services, and other methods.

[0004] Currently, discussions are underway to improve and enhance the performance of the initial 5G mobile communication technology, taking into account the services that the 5G mobile communication technology was intended to support. Additionally, standardization of the physical layer is in progress for technologies such as V2X (Vehicle-to-Everything), which helps autonomous vehicles make driving decisions and enhance user convenience based on their own location and status information transmitted by the vehicle; NR-U (New Radio Unlicensed), which aims for system operation in unlicensed bands to comply with various regulatory requirements; NR terminal low power consumption technology (UE Power Saving); Non-Terrestrial Network (NTN), which is direct terminal-satellite communication for securing coverage in areas where communication with the terrestrial network is impossible; and positioning.

[0005] In addition, standardization is underway in the field of wireless interface architecture / protocols for technologies such as the Industrial Internet of Things (IIoT) for supporting new services through linkage and convergence with other industries, Integrated Access and Backhaul (IAB) which provides nodes for expanding network service areas by integrating wireless backhaul links and access links, Mobility Enhancement including Conditional Handover and Dual Active Protocol Stack (DAPS) Handover, and 2-step Random Access (2-step RACH for NR) which simplifies random access procedures. Standardization is also underway in the field of system architecture / services for 5G baseline architectures (e.g., Service based Architecture, Service based Interface) for incorporating Network Functions Virtualization (NFV) and Software-Defined Networking (SDN) technologies, and Mobile Edge Computing (MEC), which provides services based on the location of the terminal.

[0006] When such 5G mobile communication systems are commercialized, connected devices, which are increasing explosively, will be connected to communication networks. Accordingly, it is expected that there will be a need to enhance the functionality and performance of 5G mobile communication systems and to integrate the operation of connected devices. To this end, new research is planned to be conducted on 5G performance improvement and complexity reduction, support for AI services, support for metaverse services, and drone communication using eXtended Reality (XR), Artificial Intelligence (AI), and Machine Learning (ML) to efficiently support Augmented Reality (AR), Virtual Reality (VR), and Mixed Reality (MR).

[0007] Furthermore, the advancement of these 5G mobile communication systems encompasses multi-antenna transmission technologies such as new waveforms, Full Dimensional MIMO (FD-MIMO), array antennas, and large-scale antennas to guarantee coverage in the terahertz band of 6G mobile communication technology; metamaterial-based lenses and antennas; high-dimensional spatial multiplexing technology using Orbital Angular Momentum (OAM); and Reconfigurable Intelligent Surface (RIS) technology to improve terahertz band signal coverage; as well as full-duplex technology for enhancing frequency efficiency and system networks in 6G mobile communication technology; AI-based communication technologies that realize system optimization by utilizing satellites and Artificial Intelligence (AI) from the design stage and internalizing end-to-end AI support functions; and the realization of services of complexity exceeding the limits of terminal computing capabilities by utilizing ultra-high-performance communication and computing resources. It could serve as a foundation for the development of next-generation distributed computing technologies.

[0008] According to one aspect of the present disclosure, a method for managing data in a wireless communication system may include the steps of: acquiring data related to the wireless communication system; generating a knowledge graph using the data related to the wireless communication system; and adding the knowledge graph to a database.

[0009] According to one aspect of the present disclosure, an electronic device may include a memory in which a program or at least one instruction is stored and at least one processor. By executing the program or at least one instruction stored in the memory, the electronic device may acquire data related to a wireless communication system, generate a knowledge graph using the data related to the wireless communication system, and add the knowledge graph to a database of the electronic device.

[0010] According to one aspect of the present disclosure, a computer-readable recording medium may store a program for executing at least one of the embodiments of the disclosed method on a computer.

[0011] According to one aspect of the present disclosure, a computer program may be stored in a recording medium to perform at least one of the embodiments of the disclosed method on a computer.

[0012] FIG. 1 is a drawing for explaining an electronic device according to an embodiment of the present disclosure.

[0013] FIG. 2 is a drawing for explaining modules included in an electronic device according to one embodiment of the present disclosure.

[0014] FIG. 3 is a drawing for explaining a hardware configuration included in an electronic device according to one embodiment of the present disclosure.

[0015] FIG. 4 is a drawing for explaining an AI model according to one embodiment of the present disclosure.

[0016] FIGS. 5A and 5B are drawings for illustrating an electronic device according to one embodiment of the present disclosure executing an AI model.

[0017] FIGS. 6A and 6B are drawings for illustrating an embedding according to one embodiment of the present disclosure.

[0018] FIGS. 7A and 7B are drawings for illustrating an electronic device according to one embodiment of the present disclosure that simplifies communication data.

[0019] FIG. 8 is a diagram illustrating that an electronic device (1000) according to one embodiment of the present disclosure generates a knowledge graph using communication data.

[0020] FIG. 9 is a diagram illustrating that an electronic device (1000) according to one embodiment of the present disclosure performs tokenization on communication data.

[0021] FIGS. 10A and 10B are drawings for illustrating an electronic device according to one embodiment of the present disclosure training an AI model.

[0022] FIG. 11 is a diagram illustrating an electronic device according to one embodiment of the present disclosure performing preprocessing on a token.

[0023] FIG. 12 is a diagram illustrating an electronic device according to one embodiment of the present disclosure performing preprocessing on a token.

[0024] FIG. 13 is a diagram illustrating a policy model according to one embodiment of the present disclosure performing an AI task using a knowledge graph.

[0025] FIG. 14 is a drawing for explaining a wireless communication system according to one embodiment of the present disclosure.

[0026] FIG. 15 is a flowchart illustrating a method for managing data in a wireless communication system according to an embodiment of the present disclosure.

[0027] In the present disclosure, the expression “at least one of a, b, or c” may refer to “a”, “b”, “c”, “a and b”, “a and c”, “b and c”, “a, b, and c all”, or variations thereof.

[0028] In describing the present disclosure, technical details that are well known in the technical field to which the present disclosure belongs and are not directly related to the present disclosure are omitted. This is intended to convey the essence of the present disclosure more clearly without obscuring it by omitting unnecessary explanations. Furthermore, the terms described below are defined considering their functions within the present disclosure, and these definitions may vary depending on the intentions or practices of the user or operator. Therefore, their definitions should be based on the content throughout this specification.

[0029] For the same reason, some components in the attached drawings have been exaggerated, omitted, or schematically depicted. Additionally, the dimensions of each component do not entirely reflect their actual dimensions. Identical or corresponding components in each drawing have been assigned the same reference numbers.

[0030] The advantages and features of the present disclosure, and the methods for achieving them, will become clear by referring to the embodiments described below in detail together with the accompanying drawings. However, the present disclosure is not limited to the embodiments disclosed below but may be implemented in various different forms. The disclosed embodiments are provided to ensure that the disclosure of the present disclosure is complete and to fully inform those skilled in the art of the scope of the disclosure. An embodiment of the present disclosure may be defined according to the claims. Throughout the specification, the same reference numerals indicate the same components. Furthermore, in describing an embodiment of the present disclosure, if it is determined that a detailed description of a related function or configuration might unnecessarily obscure the essence of the present disclosure, such detailed description is omitted. Additionally, terms described below are defined considering their functions in the present disclosure, and these may vary depending on the intentions or conventions of the user or operator. Therefore, their definitions should be based on the content throughout the specification.

[0031] In one embodiment, each block of the flowcharts and combinations of the flowcharts may be executed by computer program instructions. The computer program instructions may be loaded into a processor of a general-purpose computer, a computer for special purposes, or other programmable data processing equipment, and the instructions executed through the processor of the computer or other programmable data processing equipment may create means for performing the functions described in the flowchart block(s).

[0032] Computer program instructions may be stored in computer-available or computer-readable memory that can be directed toward a computer or other programmable data processing equipment to implement a function in a specific manner, and instructions stored in computer-available or computer-readable memory may also produce a manufactured item containing instruction means that performs the function described in the flowchart block(s). Computer program instructions may also be loaded onto a computer or other programmable data processing equipment.

[0033] Additionally, each block of the flowchart may represent a module, segment, or part of code containing one or more executable instructions for executing a specified logical function(s). In one embodiment, the functions mentioned in the blocks may occur out of order. For example, two blocks shown in succession may be executed substantially simultaneously or in reverse order depending on the function.

[0034] In one embodiment of the present disclosure, the terms “~ module” or “~ part” used may refer to software or hardware components such as a Field Programmable Gate Array (FPGA) or an Application Specific Integrated Circuit (ASIC), and the “~ part” may perform a specific role. Meanwhile, the meaning of “~ module” or “~ part” is not limited to software or hardware. The “~ module” or “~ part” may be configured to reside in an addressable storage medium or may be configured to run one or more processors. In one embodiment, the “~ module” or “~ part” may include components such as software components, object-oriented software components, class components, and task components, as well as processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. Functions provided through a specific component or a specific “~ module” or “~ part” may be combined or separated into additional components to reduce their number. Additionally, in one embodiment, the '~ module' or '~ part' may include one or more processors.

[0035] The term "couple" and its derivatives refer to any direct or indirect communication between two or more elements, whether or not they are in physical contact with each other. The terms "transmit," "receive," and "communicate," as well as their derivatives, include both direct and indirect communication. The terms "have" and "include," as well as their derivatives, imply inclusion without limitation.

[0036] The term "or" is inclusive and means "and / or." The phrase "related to" as well as its derivatives mean: to include, to be contained within, to interconnect with, to contain, to be contained within, to connect with or to, to be coupled with, to be communicable with, to cooperate with, to interleave, to juxtapose, to be close to, to be associated with or to, to have, to possess the characteristics of, to have a relationship with, etc.

[0037] In addition, when a component is described in the present disclosure as being "connected" or "connected" to another component, it should be understood that the component may be directly connected to or directly connected to the other component, but unless otherwise specifically stated, it may also be connected or connected through another component in between.

[0038] The meanings of the terms used in the present disclosure are explained below.

[0039] Artificial intelligence (AI) models are models that can be used in next-generation wireless communication systems, such as 5G and 6G, which have higher performance and efficiency by utilizing artificial intelligence (AI) or machine learning (ML). Instead of 'AI model', terms such as 'neural network model', 'AI / ML model', 'ML model', 'generative model', 'LLM model', 'Transformer model', 'deep learning model', 'generative model', 'LLM (Large Language Model)', 'MLM (Masked LLM)', and 'attention model' may be used.

[0040] An AI model is characterized by being created through training. Here, being created through training may mean that a basic AI model is trained by a learning algorithm using multiple training data, thereby creating predefined behavioral rules or an artificial intelligence model configured to perform a desired characteristic (or objective).

[0041] Such learning may be performed on the electronic device itself, which includes the AI ​​model according to the present disclosure, or through a separate server and / or system. The trained AI model may be distributed among network functions, servers, and electronic devices. Examples of learning algorithms include, but are not limited to, supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning.

[0042] An AI model can be composed of multiple neural network layers. Each of the multiple neural network layers has multiple weight values, and can perform neural network operations through operations between the results of previous layers and the multiple weights.

[0043] Multiple weights possessed by multiple neural network layers can be optimized based on the learning results of the AI ​​model. For example, multiple weights can be updated so that the loss value or cost value obtained from the AI ​​model during the learning process is reduced or minimized. In the embodiments of the present disclosure, a specific AI model (e.g., a Transformer model) may be described as an example, but the method according to the embodiments of the present disclosure may be applied to various other neural network models.

[0044] A 'Hidden State Matrix (hereinafter 'HSM')' can refer to the intermediate computational results output from each layer included in an AI model (e.g., self-attention layer) during the execution of the AI ​​model. In other words, the Hidden State Matrix can represent the result of performing operations on tokens included in an input sequence using the weights included in the model's hidden layer. A corresponding HSM may exist for each of the various layers included in the model (e.g., attention layer, activation layer, etc.). Therefore, the HSM may include the attention score matrix, attention value matrix, activation matrix, etc. Instead of 'Hidden State Matrix,' terms such as 'latent matrix,' 'latent variable matrix,' or 'intermediate matrix' may also be used.

[0045] Hereinafter, a base station (BS) is an entity that performs resource allocation for terminals and may be at least one of gNode B, eNode B, Node B (or xNode B (where x is an alphabet including g and e)), a radio access unit, a base station controller, a satellite, an airborn, or a node on a network. In this disclosure, the term base station may be used interchangeably with network.

[0046] User equipment (UE) may include a Mobile Station (MS), a Vehicular, a Satellite, an Airborne, a Cellular Phone, a Smartphone, a Computer, or a Multimedia System capable of performing communication functions.

[0047] Additionally, in the present disclosure, a cell may represent an area covered by a single base station in wireless communication. Cells may be classified according to size into mega cells, macro cells, micro cells, pico cells, etc., but this is merely an example, and the types of cells are not limited to those described above.

[0048] In the present disclosure, a downlink (DL) may represent a wireless transmission path for a signal transmitted by a base station to a terminal, and an uplink (UL) may represent a wireless transmission path for a signal transmitted by a terminal to a base station. Additionally, a sidelink (SL) may exist, which refers to a wireless transmission path for a signal transmitted by a terminal to another terminal.

[0049] Terms used in the following description to refer to broadcast information, control information, communication coverage, state changes (e.g., events), network entities, messages, and device components are examples provided for the convenience of explanation.

[0050] In addition, while LTE, LTE-A, or 5G systems may be described below as examples, embodiments of the present disclosure may also be applied to other communication systems having similar technical backgrounds or channel types. For example, 5G-Advance or NR-Advance or 6th generation mobile communication technology (6G or beyond 5G) developed after 5G mobile communication technology (or new radio, NR) may be included, and the 5G below may be a concept that includes existing LTE, LTE-A, and other similar services. Furthermore, the present disclosure may be applied to other communication systems with some modifications made at the discretion of a person with skilled technical knowledge, without significantly departing from the scope of the present disclosure.

[0051] For convenience of explanation below, the present disclosure uses terms and names defined in the LTE (long term evolution) and NR (new radio) specifications, which are the most recent standards defined by the 3GPP (The 3rd Generation Partnership Project) among currently existing communication standards. However, the present disclosure is not limited to the terms and names described above and may be equally applied to wireless communication systems conforming to other standards.

[0052] For example, while embodiments of the present disclosure are described using a 6th generation wireless communication technology (6G) system as an example, embodiments of the present disclosure may also be applied to other wireless communication systems having similar technical backgrounds or channel types. According to other examples, embodiments of the present disclosure may be applied to wireless communication systems prior to NR, such as NR, LTE, or LTE-A, and furthermore, embodiments of the present disclosure may be applied to wireless communication systems developed after NR.

[0053] Embodiments of the present disclosure may be supported by standard documents disclosed in at least one of wireless access systems, such as O-RAN (Open-RAN), 3GPP, or 3GPP2. That is, steps or parts among the embodiments of the present invention that are not described in order to clearly reveal the technical concept of the present invention may be supported by said documents.

[0054] In addition, in specifically describing the embodiments of the present disclosure, the focus is primarily on the New RAN (NR) wireless access network and the packet core (5G System, or 5G Core Network, or NG Core: Next Generation Core) of the 5G mobile communication standard specified by 3GPP, a mobile communication standardization organization; however, the main gist of the present disclosure may be applied to other communication systems having a similar technical background with slight modifications without significantly departing from the scope of the present disclosure, and this will be possible at the judgment of a person with technical knowledge skilled in the technical field of the present disclosure. Accordingly, the present disclosure is not limited to the terms described below, and other terms having equivalent technical meanings may be used.

[0055] Network entities of an O-RAN may include Service Management Orchestration (SMO), non-Real Time RAN Intelligent Controller (non-RT RIC), near-Real Time RAN Intelligent Controller (near-RT RIC), Central Unit (O-CU), O-DU, and O-RU. However, the components of the O-RAN structure are not limited to the examples described above. For instance, the O-RAN structure may include more or fewer components than those described above. O-RAN refers to an open wireless access network that supports openness, interoperability, vendor neutrality, and network flexibility and scalability. An O-RAN may consist of an internal RAN structure composed of O-CU, O-DU, and O-RU, and an external RAN structure represented by RICs.

[0056] RIC can refer to a component that manages, controls, and optimizes RAN functions by utilizing Artificial Intelligence (AI) and Machine Learning (ML) technologies. RIC can be responsible for AI / ML functions. RIC can be classified into non-RT RIC and near-RT RIC.

[0057] A non-RT RIC can refer to a RIC that does not operate in real time. A non-RT RIC performs the management and control of a RAN in network situations where real-time response is not required, and can manage general aspects of network operations such as network resource allocation and service optimization.

[0058] rApp (Radio Application) can refer to AI / ML applications performed on non-RT RICs. rApps are used to provide specific radio services or functions, to efficiently utilize radio resources, and to control and optimize the radio aspects of a RAN.

[0059] A near-RT RIC can refer to a RIC that operates in real time. A near-RT RIC can make decisions or perform control tasks in real time to meet rapid and dynamic demands within a network, such as fault handling, interference management, and dynamic allocation of wireless resources.

[0060] xApp (Extended Application) may be an AI / ML application that runs on a near-RT RIC (230). In one embodiment, xApp may be an application used to extend and complement various functions of the RAN. xApp performs a wider range of functions than rApp and can be utilized for network management and optimization by applying AI / ML technology to more diverse and complex network functions.

[0061] The database of the present disclosure may include at least one of a relational database and a graph database. A relational database may store and manage data in a table format, and each table may consist of rows and columns. Rows may represent individual data items (records), and columns may represent characteristics or attributes (fields) of the data.

[0062] A graph database may include graph data that includes a knowledge graph. A knowledge graph may consist of at least one of nodes, edges, or properties. Nodes may represent entities, and edges may represent relationships between nodes. Properties may represent data that further describes information about nodes or edges.

[0063] With the introduction of 5G, networks may become congested. Consequently, as operational complexity increases and OPEX (operation expenditure) rises, it may be necessary to collect and manage more diverse communication data. Instead of 'communication data,' terms such as 'relational data of wireless communication systems,' 'wireless communication network statistical data,' 'network data,' 'data related to wireless communication systems,' 'traffic data,' 'wireless communication network statistical data,' and 'wireless network statistical data' may be used.

[0064] Wireless communication systems can manage communication data by constructing a relational database to store, add, and delete data. In a relational database, the unit of communication data collection may be tabular data; however, due to the characteristics of the communication domain, the relationships between features can be dependent, and it may be difficult to graph tabular data because it is hard to identify causal relationships within the communication data. In the case of relational databases, the volume of data generated may be large in order to prevent data redundancy and inconsistencies.

[0065] The features of the communication data may include at least one of Performance Measurement (PM) data, Configuration Measurement (CM) data, or Key Performance Indicator (KPI). The features of the communication data may be further described below if necessary. Instead of 'features', terms such as 'category', 'attribute', 'field', 'parameter', 'family display name', 'family name', and 'family ID' may be used.

[0066] PM data may refer to key indicators for monitoring base station status or base station performance. PM data may include CQI (Channel Quality Indicator), MCS (Modulation and Coding Scheme), indicators representing field conditions, indicators related to signal strength between the base station and the user terminal, SINR (Signal to Interference plus Noise Ratio), terminal transmission and reception volume, indicators related to transmission quality, and indicators related to data transmission speed.

[0067] For example, communication data may be stored in the form of a table containing the MCS of PM data as shown in [Table 1], but this is merely an example for convenience of explanation and is not limited to the examples mentioned.

[0068] TimeCellDL Transmission MCS 0DL Transmission MCS 1...DL Transmission MCS 312024-02-28 07:001366165...62024-02-28 07:051438257...5...................2024-02-28 23:55123350...14

[0069] CM data may refer to base station configuration parameters and can be used to change or optimize network configuration. CM data may include response time limits between the user terminal (UE) and the base station (UE Response Timeout Time), BLER (Block Error Rate), DL target BLER, etc.

[0070] For example, communication data may be stored in the form of a table including the DL target BLER of the CM data as shown in [Table 2], but this is merely an example for convenience of explanation and is not limited to the example mentioned.

[0071] TimeCellDL Target BLER2024-02-28 07:0011202024-02-28 07:051120.......2024-02-28 23:551120

[0072] KPIs may refer to information regarding important items directly related to service management, and may include data that serve as important indicators for evaluating network performance. KPIs may include RRC (Radio Resource Control) Connection Drop Rate, DL Throughput, IP Throughput, User Throughput, Potential Throughput, Call Drop Rate, etc., but are not limited to the examples mentioned.

[0073] For example, communication data may be stored in the form of a table including the DL throughput of the KPI as shown in [Table 3], but this is merely an example for convenience of explanation and is not limited to the example mentioned.

[0074] TimeCellDL Throughput (Mbps)2024-02-28 07:001463.972024-02-28 07:051351.61......2024-02-28 23:551410.1

[0075] According to an embodiment of the present disclosure, an electronic device can acquire graph data using an AI model in a wireless communication system and train the causal relationships of the communication data. The electronic device can graph relational data to add the graph data to a database and construct a graph database. A graph database may be more efficient than a relational database in terms of resources because it manages data with new features as separate nodes even when such data is added.

[0076] Electronic devices can perform graph embedding. Graph embedding can represent the transformation of graph data (nodes, edges, subgraphs, etc.) into the same low-dimensional vector. Electronic devices can store the embedded graphs in a database. The embedded graphs can be used for the AI ​​model of a policy model to perform inference or to train the AI ​​model of the policy model.

[0077] The electronic device of the present disclosure can perform graph embedding using communication data and obtain an embedded graph. The embedded graph may include an embedding matrix. An embedding matrix may refer to a matrix that stores high-dimensional data, such as words or graph data, by converting it into low-dimensional vectors. Terms such as "knowledge graph" may be used instead of "embedded graph." For example, even if it is expressed as storing, using, or utilizing a knowledge graph, it may be understood as storing, using, or utilizing an embedded graph.

[0078] FIG. 1 is a drawing for explaining an electronic device according to an embodiment of the present disclosure.

[0079] The electronic device (1000) can acquire communication data or network data. For example, the communication data may be loaded from the database of the electronic device (1000), the electronic device (1000) may receive communication data from another electronic device (e.g., OAM), or the electronic device (1000) may directly collect communication data. The electronic device (1000) can acquire graph data using the communication data. The graph data may include a knowledge graph.

[0080] The electronic device (1000) can perform graph embedding using communication data and obtain an embedded graph. The embedded graph may include an embedding matrix. The embedding matrix may refer to a matrix that converts high-dimensional data, such as words or graph data, into low-dimensional vectors and stores them.

[0081] A node in the knowledge graph may represent an embedding vector of a token obtained by performing tokenization based on features. An edge in the knowledge graph may correspond to a feature of the communication data or a value of an attention score matrix corresponding to the node. A node in the knowledge graph may correspond to a feature of the communication data. For example, a node in the knowledge graph may correspond to at least one of Performance Measurement (PM) data, Configuration Management (CM) data, or Key Performance Indicator (KPI).

[0082] For example, nodes in the knowledge graph may correspond to CQI (Channel Quality Indicator), MCS (Modulation and Coding Scheme), indicators representing field conditions, indicators related to signal strength between the base station and the user terminal, SINR (Signal to Interference plus Noise Ratio), terminal transmission and reception volume, indicators related to transmission quality, indicators related to data transmission speed, etc. Nodes in the knowledge graph may also correspond to the UE Response Timeout Time between the user terminal (UE) and the base station, BLER (Block Error Rate), DL Target BLER, etc. Nodes in the knowledge graph may also correspond to RRC (Radio Resource Control) Connection Drop Rate, DL Throughput, etc.

[0083] The database of the electronic device (1000) disclosed in the present disclosure may include at least one of a relational database and a graph database. The database of the electronic device (1000) may store at least one of communication data, graph data (e.g., a knowledge graph) or an embedding matrix.

[0084] The electronic device (1000) can generate a knowledge graph by executing an AI model using communication data. For example, the electronic device (1000) can generate a knowledge graph by executing an AI model and performing attention. The method of generating a knowledge graph by the electronic device (1000) executing an AI model using communication data will be explained in detail with reference to FIGS. 4 to 5 and FIG. 8.

[0085] FIG. 2 is a drawing for illustrating modules included in an electronic device according to one embodiment of the present disclosure. Referring to FIG. 2, an electronic device (1000) according to one embodiment of the present disclosure may include a task management module (200), an AI model (300), and a database (500). The electronic device (1000) may further include a policy model (400).

[0086] The modules (200, 300, 400, 500) included in the electronic device (1000) of FIG. 2 are configurations classified based on function or role. The modules (200, 300, 400, 500) of the electronic device (1000) of FIG. 2 may be software configurations implemented by the processor (1300) of the electronic device (1000), which will be described later with reference to FIG. 3, executing a program stored in memory (1400), and may also be virtual configurations for which no actual matching hardware device exists. In other words, the operations performed by the processor (1300) of the electronic device (1000) by executing a program or instruction stored in memory (1400) may be classified into multiple groups according to function or purpose, and the entities performing the operations included in each classified group may be represented as the modules (200, 300, 400, 500) of FIG. 2. Accordingly, the operations described as being performed by the modules (200, 300, 400, 500) of the electronic device (1000) illustrated in FIG. 2 can actually be seen as being performed by the processor (1300) of the electronic device (1000) executing a program or instruction stored in memory (1400).

[0087] In FIG. 1, one electronic device (1000) is illustrated as including modules (200, 300, 500), but is not limited thereto. One electronic device (1000) may be implemented to include modules (200, 300, 400, 500), at least some of the modules (200, 300, 400, 500) may be implemented to be included in a separate device, and any one of the modules (200, 300, 400, 500) may be implemented to be included in another module.

[0088] As such, the modules (200, 300, 400, 500) included in the electronic device (1000) according to one embodiment of the present disclosure may be hardware configurations or software configurations and may be implemented in various forms of electronic devices (e.g., one electronic device or a combination of two or more electronic devices).

[0089] An electronic device (1000) according to one embodiment of the present disclosure may be a server that performs communication with entities of a terminal, base station, or network. For example, the electronic device (1000) may be a server (e.g., an external AI / ML server) that performs communication by being included in a base station, EMS (Element Management System), SON Manager (Self-Organizing Network Manager), OAM (Operation, Administration, and Maintenance), SON Agent (Self-Organizing Network Agent), etc., or included in a separate device from EMS (Element Management System), SON Manager (Self-Organizing Network Manager), OAM (Operation, Administration, and Maintenance), SON Agent (Self-Organizing Network Agent), etc.

[0090] The AI ​​model (300) may be implemented in the form of a transformer and may include at least one of an encoder (310) and a decoder (320). The encoder (310) and the decoder (320) of the AI ​​model (300) may each include at least one of one or more attention layers and feedforward layers.

[0091] During the process in which the AI ​​model (300) performs operations, the layers included in the encoder (310) and decoder (320) output matrices and pass them to the next layer; the matrices generated in the middle during the operation process are called Hidden State Matrices (HSM). For example, during the process in which the generative model (300) performs operations, attention layers output an attention score matrix and an attention value matrix, and feedforward layers output an activation matrix. The attention score matrix, the attention value matrix, and the activation matrix are all included in the HSM. The HSMs output from the layers of the AI ​​model (300) can be stored in the database (500) as intermediate operation results.

[0092] The task management module (200) can acquire communication data. For example, the communication data may be loaded from the database (500) of the electronic device (1000), received by the task management module (200) from another electronic device (e.g., OAM), or collected directly by the task management module (200).

[0093] The task management module (200) can obtain graph data using communication data. The graph data may include a knowledge graph. The communication data may include simplified communication data.

[0094] The task management module (200) can generate a knowledge graph by executing an AI model using communication data. The method of generating a knowledge graph by the electronic device (1000) executing an AI model using communication data will be explained in detail with reference to FIGS. 4, 5 and 8. The electronic device (1000) can train an AI model (300) using communication data. The method of training an AI model will be explained in detail with reference to FIG. 10.

[0095] The task management module (200) can simplify communication data. The task management module (200) can remove fields representing NaN (not a number) from the communication data. Communication data from which fields representing NaN (not a number) have been removed may be referred to as compressed communication data or simplified communication data. Each row of the simplified communication data may consist of a variable number of columns. A method for simplifying communication data will be described in detail with reference to FIG. 7.

[0096] The task management module (200) can perform tokenization on communication data. The task management module (200) can group the characteristics of the communication data into representative characteristics. Instead of 'representative characteristics', terms such as 'Main Characteristic', 'Representative Attribute', 'Distinctive Trait', and 'Key Aspect' may be used. The electronic device (1000) can obtain tokens by performing tokenization (grouping) according to representative characteristics. A method for performing tokenization on communication data will be explained in detail with reference to FIG. 9.

[0097] The task management module (200) can determine representative features according to a grouping rule. The grouping rule may be set in advance or defined. For example, the grouping rule may be set at the time of base station design. The task management module (200) may also determine representative features using a neural network. For example, if the electronic device (1000) performs attention and the value of the attention score matrix between features is greater than or equal to a threshold value, the features may be grouped into representative features. A method for performing tokenization on communication data will be explained in detail with reference to FIG. 9.

[0098] The task management module (200) can perform preprocessing on the token. The preprocessing may include at least one of padding or normalization. If the task management module (200) performs padding, the dimensions of the acquired token may be the same.

[0099] The task management module (200) may perform normalization on one or more acquired tokens. Normalization may include row normalization and column normalization. The task management module (200) may perform row normalization on the tokens if the tokens correspond to distribution data. The preprocessed tokens may be used by the task management module (200) to run the AI ​​model (300) or to train the AI ​​model (300). A method for performing preprocessing on tokens may be described in detail with reference to FIGS. 11 and 12.

[0100] The task management module (200) can execute the AI ​​model (300) using the acquired token. The task management module (200) can generate a knowledge graph using the execution result of the AI ​​model (300). A node in the knowledge graph may represent an embedding vector of a token acquired by performing tokenization according to a feature. A node may correspond to a feature of the communication data (e.g., a representative feature). An edge in the knowledge graph may correspond to a feature of the communication data or a value of an attention score matrix corresponding to a node.

[0101] The task management module (200) can train the AI ​​model (300) using the acquired tokens. The task management module (200) can train the AI ​​model (300) by utilizing the acquired tokens as inputs and outputs. The task management module (200) can train the AI ​​model (300) by masking at least some of the features in at least one of the inputs and outputs. In one embodiment, the task management module (200) can train the AI ​​model (300) by randomly masking some features of both the inputs and outputs.

[0102] The task management module (200) can transmit a knowledge graph to the policy model (400). The knowledge graph can be used by the policy model (400) to perform AI tasks. The policy model (400) may include AI models trained to perform a predetermined AI task.

[0103] For example, the policy model (400) can obtain outputs regarding base station parameters, CM (Configuration Measurement) data, or base station policies by executing an AI model using a knowledge graph. The AI ​​model of the policy model (400) can be trained to receive a knowledge graph as input and output base station parameters, CM (Configuration Measurement) data, or base station policies.

[0104] By executing an AI model using a knowledge graph of a policy model (400), an output regarding network performance-related KPIs (Key Performance Indicators) can be obtained. The AI ​​model of the policy model (400) can be trained to receive a knowledge graph as input and output KPIs (Key Performance Indicators).

[0105] The policy model (400) can obtain an output regarding channel quality-related indicators by executing an AI model using a knowledge graph. The AI ​​model of the policy model (400) can be trained to receive a knowledge graph as input and output channel quality-related indicators.

[0106] The policy model (400) can obtain output regarding network traffic-related metrics by executing an AI model using a knowledge graph. The AI ​​model of the policy model (400) can be trained to receive a knowledge graph as input and output traffic-related metrics.

[0107] According to one embodiment of the present disclosure, the database (500) may include a database having various structures and forms. The database (500) may include at least one of a relational database and a graph database. The database (500) may store at least one of communication data, graph data (e.g., a knowledge graph), an embedding matrix, or an attention matrix. The attention matrix of the present disclosure may include an attention output, an attention score matrix, or an attention value matrix. Information related to tasks performed by the AI ​​model (300) may be stored in the database (500). Data stored in the database (500) will be further described below as necessary.

[0108] FIG. 3 is a diagram illustrating a hardware configuration included in an electronic device according to one embodiment of the present disclosure. Referring to FIG. 2, an electronic device (1000) according to one embodiment may include a communication interface (1100), an input / output interface (1200), a processor (1300), and a memory (1400). However, the components of the electronic device (1000) are not limited to the examples described above, and the electronic device (1000) may include more components than the components described above, or fewer components. Some or all of the communication interface (1100), the input / output interface (1200), the processor (1300), and the memory (1400) may be implemented in the form of a single chip.

[0109] The communication interface (1100) is configured to transmit and receive signals (control commands and data, etc.) to and from an external device via wired or wireless means, and may be implemented to include a communication chipset that supports various communication protocols. The communication interface (1100) may receive signals from the outside and output them to the processor (1300), or transmit signals output from the processor (1300) to the outside. The electronic device (1000) may communicate with external devices through the communication interface (1100). Instead of the communication interface (1100), terms such as a transceiver unit, which collectively refers to a receiver and a transmitter, may be used.

[0110] The communication interface (1100) can transmit and receive signals with a terminal, a base station, a network element, or a network entity. The signals transmitted and received with the terminal, a base station, a network element, or a network entity may include control information and data. To this end, the communication interface (1100) may be composed of an RF transmitter that up-converts and amplifies the frequency of a transmitted signal, and an RF receiver that low-noise amplifies a received signal and down-converts the frequency. However, this is one embodiment of the communication interface (1100), and the components of the communication interface (1100) are not limited to an RF transmitter and an RF receiver.

[0111] Additionally, the communication interface (1100) can perform functions for transmitting and receiving signals through a wireless channel. For example, the communication interface (1100) can receive a signal through a wireless channel and output it to a processor (1300), and transmit the signal output from the processor (1300) through a wireless channel.

[0112] The input / output interface (1200) may include an input interface (e.g., touch screen, keyboard, microphone, etc.) for receiving commands or information from a user, and an output interface (e.g., display panel, speaker, etc.) for displaying the result of an operation according to a user's command or the status of the electronic device (1000). According to one embodiment of the present disclosure, the electronic device (1000) may receive prompts and input data from a user through the input / output interface (1200), and when the operation is completed, it may output the result of the operation (e.g., an answer to a request or question in the prompt, an image or audio edited according to the request in the prompt, etc.) through the input / output interface (1200).

[0113] A processor (1300) is a configuration that controls a series of processes to enable an electronic device (1000) to operate according to the embodiments described below, and may be composed of one or more processors. One or more processors included in the processor (1300) may be circuitry such as an SoC (System on Chip) or IC (Integrated Circuit). One or more processors included in the processor (1300) may be a general-purpose processor such as a CPU (Central Processing Unit), MPU (Micro Processor Unit), AP (Application Processor), or DSP (Digital Signal Processor); a graphics-dedicated processor such as a GPU (Graphic Processing Unit) or VPU (Vision Processing Unit); an artificial intelligence-dedicated processor such as an NPU (Neural Processing Unit); or a communication-dedicated processor such as a CP (Communication Processor). If one or more processors included in the processor (1300) are artificial intelligence-dedicated processors, said artificial intelligence-dedicated processors may be designed with a hardware structure specialized for processing a specific artificial intelligence model.

[0114] The processor (1300) can write data to memory (1400) or read data stored in memory (1400), and in particular, can process data according to a predefined operation rule or artificial intelligence model by executing a program or at least one instruction stored in memory (1400). Accordingly, the processor (1300) can perform operations described in subsequent embodiments, and operations described as being performed by the electronic device (1000) or modules (200, 300, 400, 500) included in the electronic device (1000) in subsequent embodiments can be seen as being performed by the processor (1300) unless otherwise specified.

[0115] For example, the processor (1300) can receive control signals and data signals through the transceiver (1210) and process the received control signals and data signals. The processor (1300) can transmit the processed control signals and data signals through the communication interface (1100). Additionally, the processor (1300) can write or read data to or from the memory (1400). The processor (1300) can perform the functions of the protocol stack required by the communication standard. To this end, the processor (1300) may include at least one processor or microprocessor.

[0116] Memory (1400) is configured to store various programs or data and may be composed of a storage medium or a combination of storage media such as ROM, RAM, hard disk, CD-ROM, and DVD. Memory (1400) may not exist separately but may be configured to be included in the processor (1300). Memory (1400) may be composed of volatile memory, non-volatile memory, or a combination of volatile and non-volatile memory. Memory (1400) may store a program or at least one instruction for performing operations according to the embodiments described below. Memory (1400) may provide stored data to the processor (1300) upon the request of the processor (1300).

[0117] FIG. 4 is a drawing for explaining an AI model according to one embodiment of the present disclosure.

[0118] Referring to FIG. 4, according to one embodiment of the present disclosure, an AI model (300) may be implemented in the form of a transformer and may include at least one of an encoder (310) and a decoder (320). The encoder (310) and the decoder (320) of the generative model (300) may each include one or more attention layers and feedforward layers.

[0119] Each attention layer may include one or more attention heads. For example, in the case of multi-head attention, multiple attention heads may operate in parallel, and the outputs obtained from each attention head may be combined to obtain a final output. The operation performed by the AI ​​model (300) using the attention layer will be explained in detail with reference to FIG. 4.

[0120] During the process in which the AI ​​model (300) performs operations, the layers included in the encoder (310) and decoder (320) output matrices and pass them to the next layer. The matrices generated in the middle during the operation process are called Hidden State Matrices (HSM).

[0121] For example, during the process in which the generative model (300) performs operations, attention layers may output an attention value matrix, and feedforward layers may output an activation matrix. Both the attention value matrix and the activation matrix are included in the HSM. The HSMs output from the layers of the AI ​​model (300) may be stored in the database (500) as intermediate operation results.

[0122] FIGS. 5A and FIGS. 5B are drawings for illustrating an electronic device according to one embodiment of the present disclosure executing an AI model.

[0123] The electronic device (1000) can execute the AI ​​model (300) using a token. The AI ​​model (300) can perform attention. That is, the AI ​​model (300) can perform operations using an attention layer. The attention layer may include one or more attention heads. For example, in the case of multi-head attention, multiple attention heads may operate in parallel, and the outputs obtained from each attention head may be combined to obtain a final output.

[0124] Referring to FIGS. 5A and 5B, the electronic device (1000) can obtain an embedding matrix X using n tokens. For example, the electronic device (1000) can obtain an embedding matrix X=[X1, X2, ..., X n ] can be obtained. Here, X imay represent an embedding vector of the i token. In an embodiment of the present disclosure, the i token may correspond to a specific feature of the communication data (e.g., representative feature CQI, SINR, MCS, etc.). The AI ​​model (300) can generate a query (Query, Q), a key (key, K), and a value (value, V) at each attention head using the embedding matrix X.

[0125] Q at each attention head i =W q * X i Through the key matrix Q=[Q1, Q2,.., Q n ] can be calculated. Here, Q i is X i The query vector corresponding to, Q k It may refer to a query weight matrix for calculating a query. Instead of 'query', terms such as 'query data', 'query matrix', 'query cache', 'query cache', or 'one or more query vectors' may be used.

[0126] K at each attention head i =W k * X i Through the key matrix K=[K1, K2,.., K n ] can be calculated. Here, K i is X i Key vector corresponding to, W k can refer to a key weight matrix for calculating a key. Instead of 'key', terms such as 'key data', 'key matrix', 'key cache', 'key cache', or 'one or more key vectors' may be used.

[0127] V at each attention head i =W v X i Through this, the value matrix V=[V1, V2, ..,,V n ] can be calculated. Here, X i is the embedding vector of the i token, V i is Xi The value vector corresponding to, W k It may refer to a value weight matrix for calculating values. Instead of 'value', terms such as 'value data', 'value matrix', 'value cache', or 'one or more value vectors' may be used.

[0128] The AI ​​model (300) can perform Scaled Dot Product Attention at each attention head. For example, the AI ​​model (300) can obtain an attention score matrix by calculating the dot product of the query and the key. The AI ​​model (300) can perform scaling on the attention score matrix. For example, the AI ​​model (300) can perform scaling by dividing by the square root of the dimension of the key.

[0129] The AI ​​model (300) can apply a softmax to a scaled attention score matrix and calculate a weight for each key. The AI ​​model (300) can calculate a weighted sum by multiplying the weight for each key by an attention value matrix and obtain an attention output. The attention matrix of the present disclosure may include an attention output, an attention score matrix, or an attention value matrix.

[0130] In the case of multi-head attention, the AI ​​model (300) can concatenate the attention outputs calculated from all attention heads. The AI ​​model (300) can pass the combined attention outputs as outputs to the attention layer by performing a linear transformation on them.

[0131] For example, if sentences such as “I,” “love,” and “tennis” are input as attention heads, an attention output representing the relationships between each word is produced. In the attention output, the diagonal, which shows the relationship with each word itself, can be trained to have high association, and the influence with other words can be used as an indicator representing the major correlation between words. The acquired attention output can be utilized in a graph database as an adjacency matrix.

[0132] FIGS. 6A and 6B are drawings for illustrating an embedding according to one embodiment of the present disclosure.

[0133] Representation learning can refer to a method of automatically learning representations (or features) in a format that makes data easier to interpret. Through representation learning, the optimal representation of input data can be determined, and this latent representation can be discovered.

[0134] Representation learning allows learning the structure of data without undergoing numerous feature engineering processes on raw data. Instead of 'representation learning,' terms such as 'embedding,' 'feature extraction,' 'latent representation learning,' and 'feature learning' may be used. Embeddings can include text embeddings (e.g., Word2Vec), CNN image embeddings, and graph embeddings.

[0135] Graphs can represent their structure and properties through feature matrices and adjacency matrices. If the adjacency matrix corresponding to communication data is used directly, the size of the matrix can grow exponentially as the number of nodes increases. Handling such high-dimensional data can be inefficient in terms of computational efficiency and memory usage.

[0136] According to an embodiment of the present disclosure, the electronic device (1000) can efficiently process and analyze complex graph structures by embedding a graph. Specifically, by converting complex relationships between nodes and structural information of the graph into a low-dimensional vector space, computational efficiency can be increased, and dimensionality reduction and similarity analysis between nodes can be enabled.

[0137] The electronic device (1000) can generate a feature matrix by performing tokenization on communication data and generate an adjacency matrix by performing attention. For example, the values ​​of the attention score matrix can be used as edges of a graph and can become an adjacency matrix to be utilized in a graph database.

[0138] The electronic device (1000) can obtain a hidden state matrix or an attention matrix having a lower dimension than the adjacency matrix by performing attention. The hidden state matrix or attention matrix can help perform a specific task (e.g., KPI prediction, RAN performance optimization) by extracting characteristics of network data.

[0139] Referring to Figures 6A and 6B, when the attention score matrix is ​​visualized, nodes with similar characteristics (highly related features) can be located nearby, while nodes with different characteristics (lowly related features) can be dispersed. For example, if features are highly related to the load (e.g., UE, Byte), the features can be located in adjacent places.

[0140] FIGS. 7A and 7B are drawings for illustrating an electronic device according to one embodiment of the present disclosure that simplifies communication data.

[0141] Referring to FIGS. 7A and 7B, communication data may consist of statistical data collected at predetermined intervals, such as 5 minutes or 15 minutes. For example, communication data received by an electronic device (1000) from an EMS (Element Management System) may consist of statistical data at 5-minute intervals.

[0142] Due to network characteristics, there is a high probability that communication data will contain fields representing NaN (not a number), such as QCI (Quality of Service Class Identifier), bearers, and QAM (Quadrature Amplitude Modulation). Furthermore, when new features are added, it may be difficult to reuse relational databases from prior to the addition of these features. The probability of NaN fields occurring increases as more columns or fields are added to the communication data, or as the volume of added communication data increases.

[0143] The electronic device (1000) can simplify communication data. The electronic device (1000) can remove features or fields representing NaN (not a number) from the communication data. Communication data from which fields representing NaN (not a number) have been removed may be referred to as compressed communication data or simplified communication data. Each row of the simplified communication data may consist of a variable number of columns. The fields of the simplified communication data may correspond to nodes of a knowledge graph. The electronic device (1000) can reduce information loss while configuring the graph data only with data that has values, and the processing efficiency of the communication data can be increased.

[0144] For example, in communication data at time 08:00, the UE-64QAM field and the DL_Byte_QCI6 field represent NaN (not a number). The electronic device (1000) can obtain simplified communication data by removing the UE-64QAM field and the DL_Byte_QCI6 field. The simplified communication data at time 08:00 includes (UE_256QAM: 30.5, DL_Byte_QCI9: 750, DL Throughput: 270.3) and may not include the UE-64QAM field and the DL_Byte_QCI6 field representing NaN (not a number). UE_256QAM, DL_Byte_QCI9, and DL Throughput can be used as nodes of a knowledge graph.

[0145] In the communication data at time 08:05, the UE_256QAM field and the DL_Byte_QCI9 field represent NaN (not a number). The electronic device (1000) can obtain simplified communication data by removing the UE-64QAM field and the DL_Byte_QCI6 field. The simplified communication data at time 08:00 includes (UE_64QAM: 20.4, DL_Byte_QCI6: 800, DL Throughput: 280.5) and may not include the UE-64QAM field and the DL_Byte_QCI6 field representing NaN (not a number). UE_64QAM, DL_Byte_QCI6, and DL Throughput can be used as nodes of a knowledge graph.

[0146] Simplified communication data can be used to obtain tokens, generate a knowledge graph by running an AI model (300), or train an AI model (300). The method for generating a knowledge graph will be explained in detail with reference to FIG. 8 below. The method for training an AI model (300) will be explained in detail with reference to FIG. 10 below.

[0147] FIG. 8 is a diagram illustrating that an electronic device (1000) according to one embodiment of the present disclosure generates a knowledge graph using communication data.

[0148] The electronic device (1000) can perform tokenization on communication data according to its features. According to one embodiment of the present disclosure, the communication data may include simplified communication data. A method by which the electronic device (1000) performs tokenization on communication data will be described in detail with reference to FIG. 9 below.

[0149] An electronic device (1000) can generate a knowledge graph using one or more acquired tokens. An electronic device (1000) can execute an AI model (300) using one or more acquired tokens. The contents described above with reference to FIGS. 4 and 5 may be omitted below, but they can be applied in the same way to the electronic device (1000) executing the AI ​​model.

[0150] The electronic device (1000) can generate a knowledge graph using the execution results of the AI ​​model (300). A node of the knowledge graph may represent an embedding vector of a token obtained by performing tokenization according to a feature. A node may correspond to a feature of the communication data. An edge of the knowledge graph may correspond to a feature of the communication data or a value of an attention score matrix corresponding to a node.

[0151] Referring to FIG. 8, for 08:00 communication data, UE_256QAM, DL_Byte_QCI9, and DL Throughput embedding vectors corresponding to the characteristics of the communication data can be used as nodes of a knowledge graph. An electronic device (1000) can obtain an embedding matrix including UE_256QAM, DL_Byte_QCI9, and DL Throughput embedding vectors. The electronic device (1000) can execute an AI model (300) using the obtained embedding matrix. The AI ​​model (300) can perform attention. The electronic device (1000) can generate a knowledge graph by using the execution result of the AI ​​model (300). Edges of the knowledge graph can correspond to values ​​or weights of an attention score matrix.

[0152] In the case of tabular data consisting of time-series data or statistical data at regular time intervals, it may be difficult to identify the relationships between the data and the definition of the relationships may not be clear, making it difficult to generate a knowledge graph. According to an embodiment of the present disclosure, an electronic device (1000) performs attention to generate a knowledge graph, thereby making it easy to identify the relationships between the data and to construct a graph database.

[0153] FIG. 9 is a diagram illustrating that an electronic device (1000) according to one embodiment of the present disclosure performs tokenization on communication data.

[0154] The electronic device (1000) can perform tokenization on the communication data according to the characteristics of the communication data. According to one embodiment of the present disclosure, the communication data may include simplified communication data.

[0155] The electronic device (1000) can group the characteristics of communication data into representative characteristics. Instead of 'representative characteristics', terms such as 'Main Characteristic', 'Representative Attribute', 'Distinctive Trait', and 'Key Aspect' may be used. The electronic device (1000) can perform tokenization based on the characteristics grouped into representative characteristics.

[0156] The acquired token can be used by the electronic device (1000) to execute the AI ​​model (300) or to train the AI ​​model (300). The method of executing the AI ​​model (300) will be described in detail with reference to FIGS. 4 to 5 and FIG. 8. The method of training the AI ​​model (300) will be described in detail with reference to FIG. 10 below. The acquired token may additionally include a preprocessing process. The method of preprocessing the token will be described in detail with reference to FIGS. 11 to 12 below.

[0157] The electronic device (1000) can determine representative features according to a grouping rule. The grouping rule may be set in advance or defined. For example, the grouping rule may be set at the time of base station design. When the electronic device (1000) collects communication data from the OAM, a database table may be designed according to Category, Family Display Name, etc. The columns or features of the table may be specified under the name Type Information, and data may be collected. For example, the unit of grouping may be determined by the Family Display Name, and the grouping dimension may be determined by the number of columns of Type Information.

[0158] The electronic device (1000) may determine representative features using a neural network. For example, the electronic device (1000) may group features into representative features when the value of the attention matrix between features is greater than or equal to a threshold value by performing attention.

[0159] In histogram bin count data or scalar data, features that have similar meanings, are highly related, or are of high importance can be grouped into a single representative feature.

[0160] Referring to FIG. 9, for features (CQI_1, CQI_2, … CQI 15), CQI can be determined as the representative feature, and features (CQI_1, CQI_2, … CQI 15) can be grouped as the feature CQI. A token corresponding to CQI can be obtained. For features (SINR_1, SINR_2, … SINR 15), SINR can be determined as the representative feature, and features (SINR_1, SINR_2, … SINR 15) can be grouped as the feature SINR. A token corresponding to SINR can be obtained. (Avg active Ue, Max active UE) can be determined as UE as the representative feature, and (Avg active Ue, Max active UE) can be grouped as the feature UE. A token corresponding to UE can be obtained.

[0161] FIGS. 10A and 10B are drawings for illustrating an electronic device according to one embodiment of the present disclosure training an AI model.

[0162] The electronic device (1000) can train an AI model (300) using communication data. In one embodiment, the communication data may include simplified communication data. The electronic device (1000) can obtain tokens by performing tokenization on the communication data. The details described above with reference to FIG. 9 may be omitted below, but the same applies to the electronic device (1000) performing tokenization.

[0163] Instead of training the AI ​​model (300), it can be expressed as updating the AI ​​model (300). Updating the AI ​​model (300) may include updating the attention matrix.

[0164] The electronic device (1000) can train an AI model (300) by utilizing the acquired tokens as inputs and outputs. The electronic device (1000) can train the AI ​​model (300) by masking at least some of the features among at least one of the inputs and outputs.

[0165] For example, if the communication data features include CQI, SINR, and UE, the AI ​​model (300) can be trained so that it can restore the "CQI SINR UE" from the masked part when the AI ​​model (300) takes the "CQI [MASK] UE" as input by masking the SINR of the input. For example, the AI ​​model (300) can predict the value at the masked location by performing attention using the "CQI [MASK] UE". The electronic device (1000) can calculate the loss using a loss function for the difference between the predicted value and the masked part (SINR), and update the AI ​​model (300) using the loss.

[0166] In one embodiment, the electronic device (1000) can train an AI model (300) by randomly masking some features of both the input and the output. For example, if the features of the communication data include CQI, SINR, and UE, some features (e.g., SINR) in the input can be randomly [MASK] processed, and features different from the masked features of the input can be masked in the output. The electronic device can learn the interrelationships between the features of the communication data.

[0167] FIG. 11 is a diagram illustrating an electronic device according to one embodiment of the present disclosure performing preprocessing on a token.

[0168] The electronic device (1000) can perform preprocessing on the tokens. The preprocessing may include at least one of padding or normalization. If the electronic device (1000) performs padding, the dimensions of the tokens may be the same. If the dimensions of the tokens are the same, the AI ​​model (300) can be efficiently trained or perform inference.

[0169] For example, if the tokens corresponding to CQI are [CQI_1, CQI_2, 쪋, CQI_8] and the tokens corresponding to UE are [UE_1, UE_2], the electronic device (1000) can obtain zero-padded tokens [UE_1, UE_2, 0, 0, 0, 0, 0, 0] or one-padded tokens [UE_1, UE_2, 1, 1, 1, 1, 1, 1].

[0170] When a predetermined number of communication data is secured, the electronic device (1000) can perform vector quantization. The electronic device (1000) can train a codebook using the data set. When the electronic device (1000) performs vector quantization, it can replace a given input vector with the nearest vector within the codebook. The codebook may refer to a set of vectors used to approximate the original vector data. The electronic device (1000) can use a clustering method, such as the k-means algorithm, to divide vectors with similar characteristics into groups and select the centroid of each group as the code vector. Through this process, continuous vectors can be represented as a finite number of vectors in the codebook. By converting dimensional data into a smaller number of representative vectors, compression, noise reduction, and model lightweighting can be enabled.

[0171] In one embodiment, the electronic device (1000) may perform normalization on one or more acquired tokens. The normalization of the tokens by the electronic device (1000) will be described in detail with reference to FIG. 12.

[0172] The preprocessed token can be used by the electronic device (1000) to execute the AI ​​model (300) or to train the AI ​​model (300). The method of executing the AI ​​model (300) will be described in detail with reference to FIGS. 4, 5 and FIG. 8. The method of training the AI ​​model (300) will be described in detail with reference to FIG. 10.

[0173] FIG. 12 is a diagram illustrating an electronic device according to one embodiment of the present disclosure performing preprocessing on a token.

[0174] The electronic device (1000) can perform normalization on one or more acquired tokens. Normalization may mean reducing the search space by shrinking the data within a certain range. Normalization may include row normalization and column normalization. For example, the electronic device (1000) can determine whether row normalization is required, and if row normalization is required, it can perform row normalization and then perform column normalization. If row normalization is not required, the electronic device (1000) can perform column normalization.

[0175] The electronic device (1000) can determine whether the acquired token corresponds to distribution data. Distribution data may refer to how data shows the distribution of values ​​of a specific variable. If the token corresponds to distribution data, the electronic device (1000) can perform row normalization on the token. Referring to FIG. 12, the electronic device (1000) can perform row normalization on the distribution data (CQI).

[0176] Row normalization can refer to converting a Probability Mass Function (PMF) or Probability Density Function (PDF) by comparing only the rows of columns associated with grouped features. In the case of row normalization, a training model for the normalization function may not be required.

[0177] Min-Max Scaling, Standard Scaling, and Robust Scaling can be used for column normalization. For column normalization, an AI model for the normalization function can be used. When performing inference, AI models trained for each column can be used to carry out normalization.

[0178] FIG. 13 is a diagram illustrating a policy model according to one embodiment of the present disclosure performing an AI task using a knowledge graph.

[0179] The knowledge graph can be used for the policy model (400) to perform AI tasks. The policy model (400) may include AI models trained to perform a predetermined AI task.

[0180] For example, the policy model (400) can obtain outputs regarding base station parameters, CM (Configuration Measurement) data, or base station policies by executing an AI model using a knowledge graph. The AI ​​model can be trained to receive the knowledge graph as input and output base station parameters, CM (Configuration Measurement) data, or base station policies. The base station parameters, CM (Configuration Measurement) data, or base station policies may be related to energy saving, load balancing, C-DRX (Connected Discontinuous Reception) optimization, resource scheduling optimization, etc.

[0181] By running an AI model using a knowledge graph of a policy model (400), outputs regarding network performance-related KPIs (Key Performance Indicators) can be obtained. The AI ​​model can be trained to receive a knowledge graph as input and output KPIs (Key Performance Indicators). KPIs may include RRC (Radio Resource Control) connection drop rate, DL throughput, IP throughput, UE throughput, potential throughput, or call drop rate.

[0182] The policy model (400) can obtain outputs regarding channel quality indicators by executing an AI model using a knowledge graph. The AI ​​model can be trained to receive a knowledge graph as input and output channel quality indicators. Channel quality indicators may include CQI, SINR, or BLER.

[0183] The policy model (400) can obtain outputs regarding network traffic-related metrics by executing an AI model using a knowledge graph. The AI ​​model can be trained to receive the knowledge graph as input and output traffic-related metrics. Traffic-related metrics may include AirMacByte (Downlink, Uplink) or PRB (Physical Resource Block) usage rates. KPIs, channel quality-related metrics, or traffic-related metrics may be used for root cause analysis or cell planning, etc.

[0184] FIG. 14 is a drawing for explaining a wireless communication system according to one embodiment of the present disclosure.

[0185] Referring to FIG. 14, in operation 1, a device of a wireless communication network (e.g., a Radio Unit (RU), a scheduler, a modem) can collect communication data and transmit it to an Operations Administration Maintenance (OAM). The OAM can store the communication data received from the base station.

[0186] In operation 2, the OAM can transmit communication data to the Manage plane in the Element Management System (EMS).

[0187] In operation 3, the Manage Plane can determine the application targets (e.g., Energy Saving (ES), Load Balancing (LB), scheduler) and transmit communication data to the AI ​​Server. The communication data transmitted to the AI ​​Server can be transmitted in the form of raw metadata, such as a table format, and stored in a database. According to an embodiment of the present disclosure, the electronic device (1000) (e.g., AI Server) can convert the metadata into a knowledge graph form through the AI ​​model (300) and store the knowledge graph in a graph database. The knowledge graph can be used for the AI ​​model of the policy model (400) to learn or infer.

[0188] In operation 4, the SON (Self-Organizing Network) Manager can receive decisions determined by the AI ​​server's policy model. Based on the decisions determined by the policy model, the SON Manager can perform a policy switch.

[0189] In operation 5, the SON manager can transmit the determined policy to the management plane.

[0190] In operation 6, the management plane can pass the determined policy to the OAM.

[0191] In operation 7, OAM can transmit the determined policy to the SON agent.

[0192] In operation 8, the SON agent can apply it to the application (RU / Scheduler / Modem).

[0193] Hereinafter, a method for improving the processing efficiency of a generation model according to embodiments of the present disclosure will be described with reference to the flowcharts of FIG. 15. Since the steps included in the flowchart of FIG. 15 are performed by the electronic device (1000) of FIG. 1, FIG. 2, and FIG. 3, the contents described above with reference to FIG. 1 to FIG. 14 may be omitted below and can be applied in the same way to FIG. 15.

[0194] FIG. 15 is a flowchart illustrating a method for managing data in a wireless communication system according to an embodiment of the present disclosure.

[0195] In operation S1510, the electronic device (1000) can acquire data related to a wireless communication system.

[0196] 'Data related to a wireless communication system' may refer to 'communication data' of FIGS. 1 to 14. An electronic device (1000) may acquire data related to a wireless communication system. For example, data related to a wireless communication system may be loaded from a database (500) of the electronic device (1000), or the electronic device (1000) may receive data related to a wireless communication system from another electronic device (e.g., OAM), or the electronic device (1000) may directly collect data related to a wireless communication system.

[0197] In operation S1520, the electronic device (1000) can generate a knowledge graph using wireless communication system-related data.

[0198] The electronic device (1000) can obtain simplified wireless communication system related data by removing fields indicating NaN (not a number) from the wireless communication system related data.

[0199] The electronic device (1000) can perform tokenization of the wireless communication system-related data based on the characteristics of the wireless communication system-related data. In one embodiment, the electronic device (1000) can perform tokenization using simplified wireless communication system-related data.

[0200] An electronic device (1000) can generate a knowledge graph by performing attention using one or more acquired tokens. An electronic device (1000) can execute an AI model (300) using one or more acquired tokens. A node of the knowledge graph can represent an embedding vector of a token acquired by performing tokenization according to features on data related to a wireless communication system. An edge of the knowledge graph can represent the value of an attention score matrix corresponding to the node.

[0201] The electronic device (1000) can perform normalization on one or more acquired tokens. The electronic device (1000) can generate the knowledge graph by performing attention using the normalized tokens.

[0202] The electronic device (1000) can determine whether one or more acquired tokens correspond to distribution data. The electronic device (1000) can perform row normalization on tokens corresponding to distribution data.

[0203] The electronic device (1000) can determine whether the dimensions of one or more acquired tokens are the same. If the dimensions of one or more acquired tokens are different, the electronic device (1000) can perform padding. The electronic device (1000) can generate a knowledge graph by performing attention using one or more padded tokens.

[0204] In operation S1540, the electronic device (1000) can add the knowledge graph to the database.

[0205] The electronic device (1000) can obtain output regarding base station parameters or base station policies by executing an AI model using a knowledge graph. The AI ​​model can be trained to receive the knowledge graph as input and output base station parameters, CM (Configuration Measurement) data, or base station policies.

[0206] The electronic device (1000) can obtain an output regarding a network performance-related KPI (Key Performance Indicator) by executing an AI model using a knowledge graph. The AI ​​model can be trained to receive a knowledge graph as input and output a KPI (Key Performance Indicator).

[0207] The electronic device (1000) can obtain an output regarding network traffic-related indicators by executing an AI model using a knowledge graph. The AI ​​model can be trained to receive the knowledge graph as input and output traffic-related indicators.

[0208] The electronic device (1000) can obtain an output regarding channel quality-related indicators by executing an AI model using a knowledge graph. The AI ​​model can be trained to receive a knowledge graph as input and output channel quality-related indicators.

[0209] In one embodiment, at least some of the operations performed in operations S1510, S1520, and S1540 may be omitted.

[0210] Various embodiments of the present disclosure may be implemented or supported by one or more computer programs, and computer programs may be formed from computer-readable program code and stored on a computer-readable medium. In the present disclosure, “application” and “program” may represent one or more computer programs, software components, instruction sets, procedures, functions, objects, classes, instances, related data, or parts thereof suitable for implementation in computer-readable program code. “Computer-readable program code” may include various types of computer code, including source code, object code, and executable code. “Computer-readable medium” may include various types of media accessible by a computer, such as read-only memory (ROM), random access memory (RAM), hard disk drive (HDD), compact disc (CD), digital video disc (DVD), or various types of memory.

[0211] Additionally, a device-readable storage medium may be provided in the form of a non-transitory storage medium. Here, a 'non-transitory storage medium' is a tangible device and may exclude wired, wireless, optical, or other communication links that transmit transient electrical or other signals. Meanwhile, this 'non-transitory storage medium' does not distinguish between cases where data is stored semi-permanently and cases where it is stored temporarily. For example, a 'non-transitory storage medium' may include a buffer in which data is stored temporarily. A computer-readable medium may be any available medium accessible by a computer and may include both volatile and non-volatile media, as well as removable and non-removable media. A computer-readable medium includes media in which data can be stored permanently and media in which data can be stored and subsequently overwritten, such as rewritable optical discs or erasable memory devices.

[0212] According to one embodiment, the method according to the various embodiments disclosed herein may be provided by being included in a computer program product. The computer program product may be traded between a seller and a buyer as a product. The computer program product may be distributed in the form of a device-readable storage medium (e.g., compact disc read-only memory (CD-ROM)), or distributed online (e.g., download or upload) through an application store or directly between two user devices (e.g., smartphones). In the case of online distribution, at least a portion of the computer program product (e.g., a downloadable app) may be temporarily stored or temporarily created on a device-readable storage medium, such as the memory of a manufacturer's server, an application store's server, or a relay server.

[0213] The foregoing description of the present disclosure is for illustrative purposes only, and those skilled in the art will understand that modifications can be easily made to other specific forms without altering the technical spirit or essential features of the present disclosure. For example, suitable results may be achieved even if the described techniques are performed in a different order than described, and / or components such as systems, structures, devices, circuits, etc., described are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form.

Claims

1. Regarding the method of managing data in a wireless communication system, A step of acquiring data related to a wireless communication system; A step of generating a knowledge graph using the above wireless communication system-related data; and A method comprising the step of adding the above knowledge graph to a database.

2. In Paragraph 1, The nodes of the above knowledge graph represent the embedding vectors of tokens obtained by performing tokenization according to features on the data related to the above wireless communication system, and A method in which the edge of the knowledge graph represents the value of the attention score matrix corresponding to the node.

3. In claim 1, the step of generating the knowledge graph using the wireless communication system-related data is: A step of performing tokenization of data related to the wireless communication system based on the characteristics of the data related to the wireless communication system; A method comprising the step of generating the knowledge graph by performing attention using one or more of the acquired tokens.

4. In claim 3, the step of generating the knowledge graph by performing attention using one or more acquired tokens is: A step of performing normalization on one or more of the above-mentioned acquired tokens; and A method comprising the step of generating the knowledge graph by performing attention using normalized tokens.

5. In claim 4, the step of performing normalization on one or more of the acquired tokens is: A step of determining whether one or more of the above-mentioned acquired tokens correspond to distribution data; and A method comprising the step of performing row normalization on tokens corresponding to the above distribution data.

6. In any one of claims 3 to 5, the step of generating a knowledge graph by performing attention using one or more acquired tokens is: A step of determining whether the dimensions of one or more of the above-mentioned acquired tokens are identical; If the dimensions of one or more of the above-mentioned tokens are different, a step of performing padding; and A method comprising the step of generating a knowledge graph by performing attention using one or more padded tokens.

7. In any one of paragraphs 3 to 6, the step of performing tokenization of data related to the wireless communication system is: A step of obtaining simplified wireless communication system-related data by removing a field indicating NaN (not a number) from the wireless communication system-related data; and A method comprising the step of performing tokenization using data related to a simplified wireless communication system.

8. In any one of paragraphs 1 through 7, A method further comprising the step of obtaining an output regarding base station parameters or base station policies by executing an AI model using the above knowledge graph.

9. In any one of paragraphs 1 through 8, A method further comprising the step of obtaining an output regarding a network performance-related KPI (Key Performance Indicator) by executing an AI model using the knowledge graph above.

10. In any one of paragraphs 1 through 9, A method further comprising the step of obtaining an output regarding channel quality-related indicators by executing an AI model using the above knowledge graph.

11. In any one of paragraphs 1 through 10, A method further comprising the step of obtaining an output regarding network traffic-related metrics by executing an AI model using the knowledge graph above.

12. A computer-readable recording medium having a program recorded thereon for performing the method of any one of paragraphs 1 through 11 on a computer.

13. In an electronic device for managing data in a wireless communication system, Memory in which a program or at least one instruction is stored; and It includes at least one processor, By the above at least one processor executing a program stored in the memory or at least one instruction, the electronic device, Acquire data related to wireless communication systems, and A knowledge graph is generated using the above wireless communication system-related data, and An electronic device that adds the above knowledge graph to the database of the above electronic device.

14. In Paragraph 13, The nodes of the above knowledge graph represent the embedding vectors of tokens obtained by performing tokenization according to features on the data related to the above wireless communication system, and An electronic device in which the edge of the knowledge graph represents the value of the attention score matrix corresponding to the node.

15. In claim 13, the electronic device, in generating the knowledge graph using the wireless communication system-related data, Tokenization of the wireless communication system-related data is performed based on the characteristics of the wireless communication system-related data, and An electronic device that generates the knowledge graph by performing attention using one or more of the above-mentioned acquired tokens.

Images