Method and apparatus for predicting network performance indicator

Abstract

A method by which an electronic device predicts a network performance indicator may comprise the steps of: acquiring context data from an existing value of state-related performance information of a base station, by using a first artificial intelligence model; acquiring a converted value of the state-related performance information on the basis of the context data and a target value of configuration information about the base station, by using a second artificial intelligence model; and acquiring a predicted value of a network performance indicator on the basis of the converted value of the state-related performance information and the target value of the configuration information about the base station, by using a third artificial intelligence model.

Description

Method and device for predicting network performance indicators

[0001] The present disclosure relates to a method and electronic device for predicting network performance indicators in a wireless network system.

[0002] Looking back at the evolution of wireless communication through successive generations, technologies have been developed primarily for human-oriented services, such as voice, multimedia, and data. Following the commercialization of 5G (5th-generation) communication systems, connected devices, which have been increasing explosively, are expected to be connected to communication networks. Examples of networked objects include vehicles, robots, drones, home appliances, displays, smart sensors installed in various infrastructures, construction machinery, and factory equipment. Mobile devices are expected to evolve into various form factors, such as augmented reality glasses, virtual reality headsets, and holographic devices. In the 6G (6th-generation) era, efforts are underway to develop improved 6G communication systems to connect hundreds of billions of devices and objects to provide diverse services. For this reason, 6G communication systems are being referred to as "beyond 5G" systems.

[0003] In the 6G communication system predicted to be realized around 2030, the maximum transmission speed is tera (i.e., 1,000 gigabit) bps, and the wireless latency is 100 microseconds (μsec). In other words, compared to the 5G communication system, the transmission speed in the 6G communication system is 50 times faster, and the wireless latency is reduced to one-tenth.

[0004] To achieve such high data transmission speeds and ultra-low latency, 6G communication systems are being considered for implementation in the terahertz band (e.g., the 95 GHz to 3 terahertz (3 THz) band). In the terahertz band, due to more severe path loss and atmospheric absorption compared to the millimeter wave (mmWave) band introduced in 5G, the importance of technology capable of guaranteeing signal reach, or coverage, is expected to increase. As key technologies to ensure coverage, radio frequency (RF) devices, antennas, new waveforms that offer better coverage than orthogonal frequency division multiplexing (OFDM), beamforming, and multi-antenna transmission technologies such as massive multiple-input and multiple-output (massive MIMO), full-dimensional MIMO (FD-MIMO), array antennas, and large-scale antennas must be developed. In addition, new technologies such as metamaterial-based lenses and antennas, high-dimensional spatial multiplexing technology using orbital angular momentum (OAM), and reconfigurable intelligent surface (RIS) are being discussed to improve coverage of terahertz band signals.

[0005] In addition, to improve frequency efficiency and system network, development is underway in 6G communication systems for full duplex technology, in which uplink and downlink simultaneously utilize the same frequency resources at the same time; network technology that integrates satellites and HAPS (high-altitude platform stations); network structure innovation technology that supports mobile base stations and enables network operation optimization and automation; dynamic spectrum sharing technology through collision avoidance based on spectrum usage prediction; AI-based communication technology that utilizes AI (artificial intelligence) from the design stage and internalizes end-to-end AI support functions to realize system optimization; and next-generation distributed computing technology that realizes services of complexity exceeding the limits of terminal computing capabilities by utilizing ultra-high performance communication and computing resources (mobile edge computing (MEC), cloud, etc.). In addition, attempts are continuing to further strengthen connectivity between devices, further optimize networks, promote the softwareization of network entities, and increase the openness of wireless communication through the design of new protocols to be used in 6G communication systems, the implementation of hardware-based security environments, the development of mechanisms for the safe utilization of data, and the development of technologies regarding privacy maintenance methods.

[0006] Due to the research and development of such 6G communication systems, it is expected that a new dimension of hyper-connected experience will become possible through the hyper-connectivity of 6G communication systems, which encompasses not only connections between objects but also connections between people and objects. Specifically, it is projected that 6G communication systems will enable the provision of services such as truly immersive extended reality (truly immersive XR), high-fidelity mobile holograms, and digital replicas. Furthermore, services such as remote surgery, industrial automation, and emergency response, which are provided through 6G communication systems with enhanced security and reliability, will be applied in various fields including industry, healthcare, automotive, and home appliances.

[0007] The present disclosure may be implemented in various ways, including methods, systems, devices, or computer programs stored on computer-readable storage media.

[0008] In one embodiment of the present disclosure, a method for an electronic device to predict a network performance indicator may include the step of obtaining context data from an existing value of state-related performance information of a base station using a first artificial intelligence model. In one embodiment of the present disclosure, a method for an electronic device to predict a network performance indicator may include the step of obtaining a transformed value of the state-related performance information based on the context data and a target value of configuration information for the base station using a second artificial intelligence model. In one embodiment of the present disclosure, a method for an electronic device to predict a network performance indicator may include the step of obtaining a predicted value of the network performance indicator based on the transformed value of the state-related performance information and a target value of configuration information for the base station using a third artificial intelligence model.

[0009] In one embodiment of the present disclosure, a program for performing a method of predicting network performance indicators on a computer may be recorded on a computer-readable recording medium.

[0010] In one embodiment of the present disclosure, an electronic device may include at least one processor comprising a processing circuit and a memory storing one or more instructions. In one embodiment of the present disclosure, by the at least one processor executing the one or more instructions individually or collectively, the electronic device may obtain context data from an existing value of state-related performance information of a base station using a first artificial intelligence model. In one embodiment of the present disclosure, by the at least one processor executing the one or more instructions individually or collectively, the electronic device may obtain a transformed value of the state-related performance information based on the context data and a target value of configuration information for the base station using a second artificial intelligence model. In one embodiment of the present disclosure, by the at least one processor executing the one or more instructions individually or collectively, the electronic device may obtain a predicted value of the network performance indicator based on the transformed value of the state-related performance information and a target value of configuration information for the base station using a third artificial intelligence model.

[0011] FIG. 1 is a diagram illustrating an example of predicting a key performance indicator (KPI) according to one embodiment of the present disclosure.

[0012] FIG. 2 is a drawing showing an example of a base station and an Element Management System (EMS) according to one embodiment of the present disclosure.

[0013] FIG. 3 is a diagram illustrating an example of predicting KPIs in one embodiment of the present disclosure.

[0014] FIG. 4 is a diagram illustrating an example of converting state-related performance information in one embodiment of the present disclosure.

[0015] FIG. 5 is a diagram illustrating an example of predicting KPIs in one embodiment of the present disclosure.

[0016] FIG. 6 is a diagram illustrating an example of extracting context data in one embodiment of the present disclosure.

[0017] FIG. 7 is a diagram illustrating an example of converting state-related performance information in one embodiment of the present disclosure.

[0018] FIG. 8 is a diagram illustrating an example of predicting KPIs in one embodiment of the present disclosure.

[0019] FIG. 9 is a diagram showing an example of the operation flow of an EMS in one embodiment of the present disclosure.

[0020] FIG. 10 is a diagram illustrating an example of a method for an electronic device to predict network performance indicators in one embodiment of the present disclosure.

[0021] FIG. 11 is a drawing showing an example of an electronic device according to one embodiment of the present disclosure.

[0022] The present disclosure is capable of various modifications and may have various embodiments, and specific embodiments are illustrated in the drawings and described in detail. However, this is not intended to limit the present disclosure to specific embodiments, and it should be understood that it includes all modifications, equivalents, and substitutions that fall within the spirit and scope of the present disclosure.

[0023] In describing the embodiments, detailed descriptions of related prior art are omitted if it is determined that such descriptions would unnecessarily obscure the gist of the matter. Additionally, numbers used in the description of the embodiments (e.g., First, Second, etc.) are merely identification symbols to distinguish one component from another. Furthermore, unless the context clearly indicates otherwise, the singular forms 'a', 'an', and 'the' may be understood to include multiple objects.

[0024] It should be understood that the blocks in each flowchart and combinations of flowcharts can be executed by one or more computer programs containing computer-executable instructions. One or more computer programs may be stored all in a single memory or may be partitioned and stored in multiple different memories.

[0025] All functions or operations described in this document may be processed by a single processor or a combination of processors. A single processor or a combination of processors is a circuitry that performs processing and may include circuitry such as an AP (Application Processor), CP (Communication Processor), GPU (Graphical Processing Unit), NPU (Neural Processing Unit), MPU (Microprocessor Unit), SoC (System on Chip), IC (Integrated Chip), etc.

[0026] Embodiments of the present disclosure are described below with reference to the attached drawings so that those skilled in the art can easily implement them. However, the present disclosure may be embodied in various different forms and is not limited to the embodiments described herein. Prior to a detailed description of the invention, terms used in this specification below may be defined or understood as follows.

[0027] In the present specification, when one component is described as being 'connected' to or 'connected' to another component, it should be understood that the one component may be directly connected to or directly connected to the other component, but unless specifically stated otherwise, it may also be connected or connected through another component in between. Additionally, 'connection' may include wireless connection or wired connection.

[0028] In addition, components expressed as '~part (unit)', 'module', etc. in this specification may consist of two or more components combined into a single component, or a single component may be divided into two or more components according to more detailed functions. Furthermore, each component described below may additionally perform some or all of the functions performed by other components in addition to the primary function it is responsible for, and it is obvious that some of the primary functions performed by each component may be exclusively performed by other components.

[0029] 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', 'all of a, b, and c', or variations thereof. In the present disclosure, the expression 'a or b' may refer to 'a', 'b', 'a and b', or variations thereof. In the present disclosure, the expression 'a (or, b, c)' or the expression 'a, b, or c' may refer to 'a', 'b', 'c', 'a and b', 'a and c', 'b and c', 'all of a, b, and c', or variations thereof.

[0030] In the present disclosure, 'Performance Management (PM)' may include data related to the performance of a base station. For example, the performance management may include various data collected for measuring the performance of a base station.

[0031] In one embodiment of the present disclosure, 'performance information' may include 'environment-related performance information (Environment PM)' and 'state-related performance information (State PM).' For example, 'environment-related performance information' may include performance information derived from the surrounding environment of a base station that is not significantly affected by base station settings. For example, 'environment-related performance information' may include Channel Quality Indicator (CQI), Signal to Noise Ratio (SINR), Rank Indicator (RI), terminal information, etc. For example, 'state-related performance information' may include performance information that is significantly affected by base station settings (e.g., settings). For example, 'state-related performance information' may include Modulation Coding Scheme (MCS), Block Error Rate (BLER), etc.

[0032] In one embodiment of the present disclosure, 'performance information' may be classified into state-related performance information and environment-related performance information based on sensitivity to changes in the value of configuration information for a base station (e.g., the setting value of a control parameter). For example, 'environment-related performance information' may include performance information that is insensitive to changes in the value of configuration information. For example, 'state-related performance information' may include performance information that is sensitive to changes in the value of configuration information.

[0033] In one embodiment of the present disclosure, whether each piece of information is included in state-related performance information or environment-related performance information may be pre-set, determined, or defined. In one embodiment of the present disclosure, whether each piece of information is included in state-related performance information or environment-related performance information may be determined based on a specific algorithm or method. For example, an electronic device may determine whether said specific performance information is included in state-related performance information or environment-related performance information by quantifying the degree of change of specific performance information according to a change in the value of setting information.

[0034] In the present disclosure, 'existing state-related performance information' or 'existing values ​​of state-related performance information' are not limited to data of state-related performance information collected at an 'operation time'. For example, 'existing state-related performance information' or 'existing values ​​of state-related performance information' may include data of state-related performance information collected at a specific time or within a specific time interval. For example, 'existing state-related performance information' or 'existing values ​​of state-related performance information' are not limited to data of state-related performance information actually collected, but may include data of state-related performance information obtained by simulation or data of artificially generated state-related performance information.

[0035] In the present disclosure, 'existing environment-related performance information' or 'existing values ​​of environment-related performance information' are not limited to data of environment-related performance information collected at an 'operation time'. For example, 'existing environment-related performance information' or 'existing values ​​of environment-related performance information' may include data of environment-related performance information collected at a specific time or within a specific time interval. For example, 'existing environment-related performance information' or 'existing values ​​of environment-related performance information' are not limited to data of environment-related performance information actually collected, but may include data of environment-related performance information obtained by simulation or data of environment-related performance information artificially generated.

[0036] In the present disclosure, 'Configuration Management (CM)' may include control parameters for a base station as items subject to parameter optimization. For example, configuration management may include UE (user equipment) response timeout time, DL (downlink) target BLER, etc.

[0037] In the present disclosure, a 'key performance indicator (KPI)' may include an indicator representing network performance. For example, the performance indicator may include information used for communication service management, evaluation indicators closely associated with communication service management, or information associated with communication service quality. For example, the 'performance indicator' may include Radio Resource Control (RRC) Connection Drop Rate and DL Throughput.

[0038] In one embodiment of the present disclosure, whether certain information is included in performance information, configuration information, or performance indicators may be predefined, set, or determined. In one embodiment of the present disclosure, whether certain information is included in performance information, configuration information, or performance indicators may be defined, set, or determined by a specific method or algorithm.

[0039] FIG. 1 is a drawing illustrating an example of predicting KPIs according to one embodiment of the present disclosure.

[0040] For efficient and reliable wireless network operation, it is necessary to properly set and manage base station control parameters (e.g., configuration parameters) related to network elements. Currently, adjusting base station control parameters requires manual work by specialized engineers, which can lead to issues regarding human resource costs and management. Furthermore, there is a possibility of errors occurring during the manual adjustment process.

[0041] To automate the setting of base station control parameters, a method of determining the setting values ​​by repeating optimization cycles can be used. For example, the optimal control parameter setting can be determined by repeatedly modifying the control parameter settings and evaluating network performance at each setting. However, the method of determining control parameter setting values ​​by repeating optimization cycles can be time-consuming and may result in excessive costs.

[0042] Reinforcement learning can be utilized to recommend optimized control parameters. However, since network data corresponds to various types of high-dimensional data (e.g., hundreds of dimensions, thousands of dimensions) and network operations are very complex, applying reinforcement learning to the field of network technology requires high-performance simulators, and there may be time or cost limitations regarding the use of high-performance simulators. Therefore, it is difficult to apply the aforementioned method of repeating optimization cycles or the method of utilizing reinforcement learning to large-scale commercial networks.

[0043] In the present disclosure, a method for predicting performance indicators based on base station settings may be provided. Since configuration information for a base station directly affects the network state (e.g., state-related performance information), the actual performance of the network may vary significantly depending on the base station settings. As an embodiment of the present disclosure, to optimize a wireless network by automatically deriving and setting values ​​of wireless network parameters (e.g., control parameters for a base station) during wireless network operation, a technology and method of operation may be provided for predicting the state of a wireless network that changes as the setting values ​​of wireless network parameters are changed, and utilizing the predicted state of the wireless network for predicting performance indicators (KPIs). For example, an electronic device may transform the state of a wireless network based on the settings for a base station and predict performance indicators of the wireless network based on the transformed state of the wireless network.

[0044] In one embodiment of the present disclosure, an electronic device may use an artificial intelligence model to identify, determine, or obtain predicted data, inferred data, or estimated data of performance indicators. For example, the electronic device may use an artificial intelligence model to predict, infer, or estimate the value of a performance indicator corresponding to a specific setting condition for a base station. For example, the electronic device may use an artificial intelligence model to predict, infer, or estimate the value of a performance indicator when the setting information of a base station is set to specific data (or a specific value).

[0045] In one embodiment of the present disclosure, an artificial intelligence model used by an electronic device for predicting performance indicators may include a plurality of mutually dependent models or a plurality of mutually independent models. For example, referring to FIG. 1, the artificial intelligence model used for predicting performance indicators may include a context extraction model (110), a state transition model (120), and a KPI prediction model (130). For example, at least some of the context extraction model (110), the state transition model (120), and the KPI prediction model (130) may be connected end-to-end and trained as a single artificial intelligence model. For example, at least some of the context extraction model (110), the state transition model (120), and the KPI prediction model (130) may be trained individually as independent artificial intelligence models.

[0046] In one embodiment of the present disclosure, the context extraction model (110) may include a model trained to extract context data from state-related performance information (140) for a base station. In one embodiment of the present disclosure, the state conversion model (120) may include a model trained to output data (e.g., conversion data) of state-related performance information when the configuration information is set to specific target data. In one embodiment of the present disclosure, the KPI prediction model (130) may include a model trained to output predicted data of a performance indicator based on input data.

[0047] Referring to FIG. 1, an electronic device can extract context data based on state-related performance information (140) of a base station using a context extraction model (110). In one embodiment of the present disclosure, the electronic device can obtain context data from existing values ​​of state-related performance information (140) of a base station using a context extraction model (110). For example, the electronic device can obtain context data corresponding to the current state of a network based on current state-related performance information using a context extraction model (110).

[0048] In one embodiment of the present disclosure, context data may include data in which input data is embedded into data of a specific dimension by an artificial intelligence model (e.g., a context extraction model (110)). For example, context data may include feature data extracted by an artificial intelligence model from input data, which may or may not be understood by a user. For example, context data may include vector data regarding a situation (e.g., user traffic patterns, software version, weather, etc.) in which a base station is in a specific time (e.g., time of information collection) by utilizing artificial intelligence techniques from information (e.g., performance information) collected for base station performance measurement.

[0049] Referring to FIG. 1, the electronic device can obtain transformed state-related performance information based on context data (or existing state-related performance information) and base station configuration information (150) using a state transformation model (120). For example, the transformed state-related performance information may include transformed information in which configuration information of a specific value (e.g., target value) is reflected in the existing state-related performance information. For example, the transformed state-related performance information may include information that predicts, infers, or estimates the value of state-related performance information under conditions where the base station configuration information is set to a specific value.

[0050] In one embodiment of the present disclosure, an electronic device may obtain a transformed value of state-related performance information based on a target value of context data and configuration information (150) for a base station by using a state transformation model (120). For example, the target value is a value of configuration information used to predict performance indicators or performance information, and the electronic device may predict, infer, or estimate the performance information or performance indicator of a base station when the configuration information (150) for a base station is set to the target value. For example, the transformed value of state-related performance information may include a value of state-related performance information predicted, inferred, or estimated when the configuration information for a base station is set to the target value. For example, the state transformation model (120) may include a model trained to predict, infer, or estimate the value of state-related performance information when the configuration information (150) for a base station is set to the target value based on a target value of context data and configuration information (150) for a base station corresponding to an existing value of state-related performance information.

[0051] Referring to FIG. 1, an electronic device can predict, infer, identify, determine, or obtain a value of KPI based on at least some of the configuration information (150) for a base station, performance information related to a transformed state, or performance information related to an environment (160) using a KPI prediction model (130). For example, the electronic device can obtain a predicted value (170) of KPI under conditions where the configuration information (150) for a base station is set to a target value using a KPI prediction model (130). For example, the KPI prediction model (130) may include a model trained to predict the value of KPI when the configuration information for a base station is set to a specific value based on the configuration information (150) for a base station. For example, the KPI prediction model (130) may include a model trained to predict the value of KPI when the performance information related to the state of a base station is a specific value (e.g., a transformed value) based on the performance information related to the state of a base station. For example, the KPI prediction model (130) may include a model trained to predict the value of the KPI when the environment-related performance information (160) of the base station is a specific value (e.g., existing value, target value) based on the environment-related performance information (160) of the base station.

[0052] In one embodiment of the present disclosure, an electronic device may obtain a predicted value of a performance indicator based on a target value of configuration information for a base station and a converted value of state-related performance information using a KPI prediction model (130). For example, the electronic device may obtain a predicted value of a performance indicator based on a target value of configuration information for a base station, a converted value of state-related performance information, and environment-related performance information (160) using a KPI prediction model (130). For example, the predicted value of a performance indicator may include a value of a performance indicator predicted when the configuration information for a base station is set to a target value.

[0053] In one embodiment of the present disclosure, the environment-related performance information (160) of a base station that serves as the basis for predicting the value of a performance indicator may include existing environment-related performance information corresponding to existing state-related performance information. For example, the environment-related performance information (160) of a base station that serves as the basis for predicting the value of a performance indicator may include information collected together when existing state-related performance information is collected. For example, an electronic device may obtain a predicted value of a performance indicator based on a target value of configuration information for a base station, a converted value of state-related performance information, and an existing value of environment-related performance information (160) using a KPI prediction model (130).

[0054] In one embodiment of the present disclosure, data input to the model or data output by the model may be time-series data. In one embodiment of the present disclosure, data input to the model or data output by the model may be data in the form of a structured table. For example, at least some of the configuration information (150), state-related performance information (140), environment-related performance information (160), or performance indicators may be structured as shown in the tables below.

[0055]

[0056] <Table 1: Example of data format for state-related performance information input or output to the model>

[0057]

[0058] <Table 2: Examples of data formats for environment-related performance information input or output to the model>

[0059]

[0060] <Table 3: Examples of data formats for configuration information input or output to the model>

[0061]

[0062] <Table 4: Example of data format for KPIs output from the model>

[0063] For example, the data in Tables 1 to 4 above may be input into the model or output from the model all at once, but are not limited thereto. For example, in Tables 1 to 4 above, data corresponding to each time point (e.g., data in each row) may be input into the model or output from the model individually. For example, in Tables 1 to 4 above, time series data of each piece of information (e.g., data in each column) may be input into the model or output from the model individually.

[0064] FIG. 2 is a drawing showing an example of a base station and an Element Management System (EMS) according to one embodiment of the present disclosure.

[0065] In explaining Fig. 2, descriptions that overlap with the descriptions in Fig. 1 may be omitted.

[0066] In a wireless network system, a base station (210) can perform the function of connecting a terminal and a core network and may include, for example, an eNB or a gNB. Referring to FIG. 2, the base station (210) may include, but is not limited to, an OAM (Operations Administration Maintenance) (212), a SON (Self-Organization Network) agent (214), a RU (Radio Unit) (216), a scheduler (218), and a modem (220). For example, the base station (210) may include additional elements not shown in FIG. 2, or may not include some elements shown in FIG. 2. At least some of the OAM (212), SON agent (214), RU (216), scheduler (218), or modem (220) included in the base station (210) may be elements that are logically, functionally, software-wise, or hardware-wise distinct from other elements.

[0067] The EMS (230) is connected to the base station (210) via wired or wireless connection and can perform functions such as collecting information on the status, performance, and errors of the base station (210), managing the settings of the base station (210), or solving problems on the network. Referring to FIG. 2, the EMS (230) may include a Manage Plane (232), an Artificial Intelligence (AI) server (240), and a SON management module (260), but is not limited thereto. For example, the EMS (230) may include additional elements not shown in FIG. 2, or may not include some elements shown in FIG. 2. For example, the AI ​​server (240) may be a module, server, or device outside the EMS and may be connected to or communicate with the EMS (230) via wired or wireless connection. At least some of the management plane (232), AI server (240), or SON management module (260) included in the EMS (230) may be elements that are logically, functionally, software-wise, or hardware-wise distinct from other elements.

[0068] In one embodiment of the present disclosure, the OAM (212) may collect, acquire, or store statistical data regarding the wireless network of the base station (210). For example, the OAM (212) may collect or acquire statistical information of the communication network from a wireless communication network device, such as a RU (216), a scheduler (218), or a modem (220). The OAM (212) may transmit the collected or acquired statistical data to the management plane (232) of the EMS (230).

[0069] In one embodiment of the present disclosure, the management plane (232) may determine a target application among a plurality of applications. For example, the plurality of applications may include an application for Energy Saving (ES), an application for Load Balancing (LB), or an application for a Scheduler. For example, the management plane (232) may determine, among the plurality of applications, an application for a function or effect to be achieved or improved through base station settings as the target application. The management plane (232) may transmit information regarding the determined target application to the AI ​​server (240).

[0070] In one embodiment of the present disclosure, the AI ​​server (240) may store data or information received from the management plane (232) in the database (242). For example, the AI ​​server (240) may receive data generated from the OAM (212) via the management plane (232). In one embodiment of the present disclosure, the AI ​​server (240) may determine the training or inference of an AI model (e.g., context extraction model, state transition model, KPI prediction model) based on the data or information received from the management plane (232).

[0071] In one embodiment of the present disclosure, the AI ​​server (240) may acquire data extracted by a Distributed Unit (DU) base station equipment, an X86-based DU emulator, or a system-level simulator. In one embodiment of the present disclosure, the AI ​​server (240) may determine the training or inference of an AI model (e.g., a context extraction model, a state transition model, a KPI prediction model) based on the acquired data.

[0072] In one embodiment of the present disclosure, the AI ​​server (240) may include a trained or updated AI model. For example, referring to FIG. 2, the AI ​​server (240) may include a trained or updated context extraction model (248), a state transition model (250), and a KPI prediction model (252). For example, the AI ​​server (240) may include or store information about the structure of each AI model or weight values ​​of each AI model.

[0073] In one embodiment of the present disclosure, the AI ​​server (240) may include an AI model corresponding to each application. For example, the AI ​​server (240) may include a context extraction model corresponding to a first application and a context extraction model corresponding to a second application. For example, the context extraction model corresponding to the first application and the context extraction model corresponding to the second application may be composed of the same structure and different weight values. For example, the context extraction model corresponding to the first application and the context extraction model corresponding to the second application may be models that are composed of the same structure but are trained or updated with different training data. For example, the context extraction model corresponding to the first application and the context extraction model corresponding to the second application may be models with different structures.

[0074] In one embodiment of the present disclosure, the AI ​​server (240) may include a policy model (244). For example, the policy model (244) may be a model trained to determine or optimize a policy for a communication network (e.g., the configuration of a base station) based on specific data. For example, the AI ​​server (240) may include a policy model (244) corresponding to each application.

[0075] In one embodiment of the present disclosure, the AI ​​server (240) may transmit a policy model (244) or an AI model corresponding to a target application to the SON management module (260) according to instructions, directives, or controls of the AI ​​management module (246). For example, the AI ​​server (240) may transmit data of an AI model trained for a target application (e.g., weight values, etc.) to the SON management module (260). In one embodiment of the present disclosure, the AI ​​server (240) may transmit data received from the management plane (232) or data stored in the database (242) to the SON management module (260).

[0076] In one embodiment of the present disclosure, the SON management module (260) can perform inference of an AI model based on acquired data. For example, the SON management module (260) can perform inference of an AI model corresponding to a target application (e.g., context extraction model (262), state transformation model (264), KPI prediction model (266)) based on acquired data.

[0077] In one embodiment of the present disclosure, a context extraction model (262) can extract information about a specific situation (e.g., context data) from data received from an AI server (240) (e.g., existing data of state-related performance information). For example, the specific situation may be designated or determined by a policy model (268). In one embodiment of the present disclosure, a state transformation model (264) can predict or transform state-related performance information from data received from an AI server (240) and context data received from the context extraction model (262). In one embodiment of the present disclosure, a KPI prediction model (266) can predict KPIs from data received from an AI server (240) and transformed state-related performance information received from the state transformation model (264). For example, the KPI predicted by the KPI prediction model (266) may be designated or determined by a policy model (268). In one embodiment of the present disclosure, the policy model (268) can determine the setting value of the communication network based on data received from the AI ​​server (240) and the predicted KPI received from the KPI prediction model (266).

[0078] In one embodiment of the present disclosure, the SON management module (260) transmits a policy (e.g., a setting value of a communication network) determined by the policy model (268) to the management plane (232), and the management plane (232) can transmit the policy to the OAM (212). In one embodiment of the present disclosure, the OAM (212) provides the policy to the SON agent (214), and the SON agent (214) can apply or reflect the policy to the RU (216), scheduler (218), or modem (220).

[0079] In one embodiment of the present disclosure, the SON management module (260) may include an analysis module (e.g., an analyzer) (270) to analyze and utilize result values ​​inferred from an AI model or a policy model (268). Referring to FIG. 2, the analysis module (270) may include a result visualization module (272), a network simulation module (274), or a root cause analysis module (276). For example, the result visualization module (272) may visualize result data of the AI ​​model (e.g., final result data, intermediate result data). For example, the network simulation module (274) may reproduce and analyze various wireless network situations resulting from changes in setting values ​​for base stations. For example, the root cause analysis module (276) may utilize result data of the AI ​​model or analysis data of the analysis module (270) to identify (or discover) problems occurring in specific situations or to analyze the cause of the problem.

[0080] Without performing operations to convert state-related performance information based on base station configuration data, electronic devices may determine wireless network configuration values ​​by directly predicting KPIs, or use optimization methods that directly predict wireless network configuration values. When using such optimization methods, it may be impossible to infer or analyze wireless network states other than the KPI, which is the final inference result of the AI ​​model. Consequently, additional analysis may not be easy when specific problem situations occur or are anticipated in the wireless network.

[0081] According to one embodiment of the present disclosure, an electronic device can identify or obtain intermediate inference results as well as the final inference result, the KPI, by predicting the KPI using an AI model structured in three stages. Therefore, both the final inference result and the intermediate inference result can be used for analysis. For example, an AI model structured separately for context information extraction, wireless network state transformation, and KPI prediction can derive useful intermediate inference results (e.g., user-perceivable results or explainable results), and analysis results useful for wireless network configuration, etc., can be generated from the intermediate inference results.

[0082] FIG. 2 illustrates each element in block form, and at least some of the elements illustrated in block form may be hardware modules, software modules, or a combination thereof. For example, an operation described as being performed by a specific module or a specific model may be performed by an electronic device containing the specific module or a specific model. For example, an operation described as being performed by a specific module or a specific model may be performed by an electronic device by executing or driving the specific module or a specific model under the control of at least one processor of the electronic device containing the specific module or a specific model. For example, an operation described as being performed by a specific module may be performed independently by the specific module.

[0083] At least some of the elements illustrated in block form in FIG. 2 may be included in the same electronic device. At least some of the elements illustrated in block form in FIG. 2 may be included in different electronic devices.

[0084] FIG. 2 illustrates that the AI ​​server (240) is included in the EMS (230), but is not limited thereto. For example, a method for constructing the AI ​​server (240) to reduce the resource burden of the EMS (230) may be used.

[0085] In one embodiment of the present disclosure, the AI ​​server (240) may be constructed as a separate hardware device from the EMS (230). For example, the AI ​​server (240) may include hardware devices with an enhanced environment, such as an acceleration device (e.g., a GPU (Graphics Processing Unit), TPU (Tensor Processing Unit), NPU (Neural Processing Unit), VPU (Vision Processing Unit), etc.) or a many-core, to handle learning such as RL (Reinforcement Learning) and ML (Machine Learning). For example, the AI ​​server (240) may be a hardware device that exists physically separately from the EMS (230) and may be connected via a wired network such as Ethernet or Infiniband, or a wireless network such as WiFi (Wireless Fidelity). For example, the AI ​​server (240) can be a lightweight accelerator (e.g., USB (Universal Serial Bus) Type TPU, Jetson Nano, etc.) that can be connected to or communicate with existing DU base station equipment, thereby creating an environment that minimizes transmission time.

[0086] In one embodiment of the present disclosure, the AI ​​server (240) may perform training of the AI ​​model, etc., using hardware equipment separate from the EMS (230). For example, the AI ​​server (240) may perform training of the AI ​​model, etc., using hardware equipment in an enhanced environment such as an acceleration device or a multi-core. For example, the AI ​​server (240) may use hardware equipment that is connected to or capable of communicating via a wired network or a wireless network. For example, a lightweight acceleration device (e.g., USB Type TPU, Jetson Nano, etc.) that is connected to or capable of communicating with existing DU base station equipment may be used. In this case, the AI ​​server (240) may perform operations using a hardware environment that minimizes transmission time.

[0087] FIG. 3 is a diagram illustrating an example of predicting KPIs in one embodiment of the present disclosure.

[0088] In describing Fig. 3, descriptions that overlap with the descriptions described in either Fig. 1 or Fig. 2 may be omitted.

[0089] The electronic device can predict, infer, identify, determine, or obtain the value of a KPI by using a context extraction model (110) and a KPI prediction model (130). For example, the electronic device can obtain the predicted value (170) of a KPI by using only at least a portion of an artificial intelligence model that includes the context extraction model (110), the state transformation model, and the KPI prediction model (130). For example, if intermediate inference results are not required for predicting the KPI, the electronic device can use an artificial intelligence model from which the state transformation model is excluded. For example, if the transformation results of state-related performance information for result visualization or providing a user-friendly solution are not required, the electronic device can obtain the predicted value (170) of a KPI by using an artificial intelligence model that includes the context extraction model (110) and the KPI prediction model (130).

[0090] In one embodiment of the present disclosure, the electronic device can predict the value of a KPI by performing inference of a context extraction model (110) and, without inference for transforming state-related performance information, immediately performing inference of a KPI prediction model (130). For example, the electronic device can obtain context data based on state-related performance information (140) using the context extraction model (110), and obtain a predicted value (170) of a KPI using the KPI prediction model (130) based on the obtained context data, setting information (150), and environment-related performance information (160). Referring to FIG. 3, the electronic device can obtain context data output from the context extraction model (110) by inputting the existing value of state-related performance information (140) into the context extraction model (110). The electronic device can obtain the predicted value (170) of the KPI output from the KPI prediction model (130) by inputting the existing value of context data, environment-related performance information (160), and the target value of setting information (150) into the KPI prediction model (130).

[0091] In one embodiment of the present disclosure, the context extraction model (110) and the KPI prediction model (130) may be connected end-to-end and trained. For example, the context extraction model (110) and the KPI prediction model (130) may be connected end-to-end and trained through another model (e.g., a state transition model). For example, the context extraction model (110) and the KPI prediction model (130) may be trained based on training data including input data and correct values ​​of KPIs.

[0092] In one embodiment of the present disclosure, the context extraction model (110) and the KPI prediction model (130) may each be trained individually or independently. For example, the context extraction model (110) may be trained based on first training data containing state-related performance information, and the KPI prediction model (130) may be trained based on second training data containing correct values ​​of KPIs. For example, the context extraction model (110) or the KPI prediction model (130) may be trained by being connected end-to-end with another model (e.g., a state transformation model).

[0093] FIG. 4 is a diagram illustrating an example of converting state-related performance information in one embodiment of the present disclosure.

[0094] In describing Fig. 4, any description that overlaps with the description in any one of Figs. 1 to 3 may be omitted.

[0095] The electronic device can obtain transformed values ​​of context data and state-related performance information by using a context extraction model (110) and a state transformation model (120). For example, the electronic device can obtain intermediate inference results by using only at least a portion of an artificial intelligence model that includes a context extraction model (110), a state transformation model (120), and a KPI prediction model. For example, if a final inference result (e.g., KPI prediction result) is not required and only an intermediate inference result is required, the electronic device can use an artificial intelligence model from which the KPI prediction model is excluded. For example, for applications that require transformed results of state-related performance information rather than KPI prediction results (e.g., data visualization, simulation, etc.), the electronic device can use an artificial intelligence model that includes a context extraction model (110) and a state transformation model (120).

[0096] In one embodiment of the present disclosure, the electronic device may obtain a transformed value of state-related performance information by performing inference of the context extraction model (110) and inference of the state transformation model (120), and may not perform inference for KPI prediction. For example, the electronic device may obtain context data based on state-related performance information (140) using the context extraction model (110), and obtain transformed state-related performance information (410) using the state transformation model (120) based on the obtained context data and setting information (150). Referring to FIG. 4, the electronic device may obtain context data output from the context extraction model (110) by inputting the existing value of state-related performance information (140) into the context extraction model (110). The electronic device may obtain a transformed value of state-related performance information by inputting the target value of the context data and setting information (150) into the state transformation model (120).

[0097] In one embodiment of the present disclosure, the context extraction model (110) and the state transition model (120) may be connected end-to-end and trained. For example, the context extraction model (110) and the state transition model (120) may be connected end-to-end and trained through another model (e.g., a KPI prediction model). In one embodiment of the present disclosure, the context extraction model (110) and the state transition model (120) may each be trained individually or independently. For example, the context extraction model (110) or the state transition model (120) may be connected end-to-end and trained with another model (e.g., a KPI prediction model).

[0098] FIG. 5 is a diagram illustrating an example of predicting KPIs in one embodiment of the present disclosure.

[0099] In explaining Fig. 5, any description that overlaps with the description in any one of Figs. 1 to 4 may be omitted.

[0100] The electronic device can predict, infer, identify, determine, or obtain the value of a KPI by using a state transformation model (120) and a KPI prediction model (130). For example, the electronic device can obtain the predicted value (170) of a KPI by using only at least a portion of an artificial intelligence model that includes a context extraction model, a state transformation model (120), and a KPI prediction model (130). For example, if the electronic device already possesses context information for a specific state, it may use an artificial intelligence model from which the context extraction model is excluded. For example, if context data (510) corresponding to performance information related to a specific state is obtained, the electronic device can obtain the predicted value (170) of a KPI by using an artificial intelligence model that includes a state transformation model (120) and a KPI prediction model (130).

[0101] In one embodiment of the present disclosure, the electronic device can predict the value of a KPI by performing inference of a state transformation model (120) directly without inference for context extraction and by performing inference of a KPI prediction model (130). For example, the electronic device can obtain state-related performance information (e.g., transformed state-related performance information) based on context data (510) and setting information (150) using the state transformation model (120), and obtain a predicted value (170) of a KPI using the KPI prediction model (130) based on the obtained state-related performance information, setting information (150), and environment-related performance information (160). Referring to FIG. 5, the electronic device can obtain a transformed value of state-related performance information output from the state transformation model (120) by inputting target values ​​of context data (510) and setting information (150) into the state transformation model (120). The electronic device can obtain a predicted value (170) of the KPI output from the KPI prediction model (130) by inputting the converted value of the state-related performance information, the target value of the setting information (150), and the existing value of the environment-related performance information (160) into the KPI prediction model (130).

[0102] In one embodiment of the present disclosure, the state transition model (120) and the KPI prediction model (130) may be connected end-to-end and trained. For example, the state transition model (120) and the KPI prediction model (130) may be connected end-to-end and trained through another model (e.g., a context extraction model). In one embodiment of the present disclosure, the state transition model (120) and the KPI prediction model (130) may each be trained individually or independently. For example, the state transition model (120) or the KPI prediction model (130) may be connected end-to-end and trained with another model (e.g., a context extraction model).

[0103] FIG. 6 is a diagram illustrating an example of extracting context data in one embodiment of the present disclosure.

[0104] In describing Fig. 6, any description that overlaps with the description in any one of Figs. 1 to 5 may be omitted.

[0105] The electronic device can obtain context information for a specific state by using a context extraction model (110). For example, the electronic device can obtain context information for a specific state by using only at least a portion of an artificial intelligence model that includes a context extraction model (110), a state transformation model, and a KPI prediction model. For example, if only an intermediate inference result is needed rather than a final inference result (e.g., a KPI prediction result), the electronic device can use an artificial intelligence model from which the state transformation model and the KPI prediction model are excluded. For example, for applications that require context information for a specific state rather than a KPI prediction result and a transformation result of state-related performance information (e.g., data analysis, simulation, etc.), the electronic device can use an artificial intelligence model that includes a context extraction model (110).

[0106] In one embodiment of the present disclosure, an electronic device may obtain context data (610) corresponding to specific state-related performance information (140) by performing inference of a context extraction model (110), and may not perform inference for state transformation and inference for KPI prediction. For example, the electronic device may obtain context data (610) from state-related performance information (140) using a context extraction model (110). Referring to FIG. 6, the electronic device may obtain context data (610) output from a context extraction model (110) by inputting a specific value of state-related performance information (140) (e.g., a value at a specific point in time, a value in a specific time interval, a target value, etc.) into the context extraction model (110). The context data (610) may include data in which state-related performance information (140) input to the context extraction model (110) is embedded in a specific dimension.

[0107] In one embodiment of the present disclosure, the context extraction model (110) may be trained individually or independently of other models. In one embodiment of the present disclosure, the context extraction model (110) may be trained by being connected end-to-end with other models.

[0108] FIG. 7 is a diagram illustrating an example of converting state-related performance information in one embodiment of the present disclosure.

[0109] In describing Fig. 7, any description that overlaps with the description in any one of Figs. 1 to 6 may be omitted.

[0110] The electronic device can predict, infer, identify, determine, or obtain state-related performance information (720) in a specific setting by using a state transformation model (120). For example, the electronic device can obtain state-related performance information (720) in a specific setting by using only at least a portion of an artificial intelligence model that includes a context extraction model, a state transformation model (120), and a KPI prediction model. For example, if context information for a specific state is already possessed and a final inference result (e.g., KPI prediction result) is not required, and only an intermediate inference result is required, the electronic device may use an artificial intelligence model from which the context transformation model and the KPI prediction model are excluded. For example, for applications that require a transformation result of state-related performance information rather than a KPI prediction result and a context extraction result (e.g., data visualization, simulation, etc.), the electronic device may use an artificial intelligence model that includes a state transformation model (120).

[0111] In one embodiment of the present disclosure, the electronic device may obtain performance information related to the transformed state by performing inference of the state transformation model (120) without performing inference for context extraction, and may not perform inference for KPI prediction. For example, the electronic device may obtain context data (710) corresponding to specific state-related performance information and obtain performance information related to the transformed state (720) based on the context data (710) and setting information (150) using the state transformation model (120). Referring to FIG. 7, the electronic device may obtain a predicted value (or transformed value) of the state-related performance information (720) output from the state transformation model (120) by inputting target values ​​of the context data (710) and setting information (150) into the state transformation model (120). The converted value of the state-related performance information (720) or the converted state-related performance information (720) may include the predicted value or predicted result of the state-related performance information when the setting information (150) is set to the target value.

[0112] In one embodiment of the present disclosure, the state transition model (120) may be learned individually or independently of other models. In one embodiment of the present disclosure, the state transition model (120) may be learned by being connected end-to-end with other models.

[0113] FIG. 8 is a diagram illustrating an example of predicting KPIs in one embodiment of the present disclosure.

[0114] In describing Fig. 8, any description that overlaps with the description in any one of Figs. 1 to 7 may be omitted.

[0115] The electronic device can predict, infer, identify, determine, or obtain the value of a KPI by using a KPI prediction model (130). For example, the electronic device can obtain the predicted value (170) of a KPI by using only at least a portion of an artificial intelligence model that includes a context extraction model, a state transformation model, and a KPI prediction model (130). For example, when predicting a KPI, if the transformation of state-related performance information is not required or if the electronic device possesses already transformed state-related performance information, the electronic device may use an artificial intelligence model from which the context extraction model and the state transformation model are excluded. For example, if the electronic device possesses state-related performance information (140) in a specific setting (e.g., target setting), the electronic device can obtain the predicted value (170) of a KPI in a specific setting by using an artificial intelligence model that includes the KPI prediction model (130).

[0116] In one embodiment of the present disclosure, an electronic device can predict the value of a KPI by performing inference of a KPI prediction model (130) without inference for context extraction and inference for state transformation. For example, the electronic device can obtain a predicted value (170) of a KPI based on state-related performance information (140), setting information (150), and environment-related performance information (160) using the KPI prediction model (130). Referring to FIG. 8, the electronic device can obtain a predicted value (170) of a KPI output from the KPI prediction model (130) by inputting the target value of the state-related performance information (140) and setting information (150), and the existing value of the environment-related performance information (160) into the KPI prediction model (130).

[0117] In one embodiment of the present disclosure, the KPI prediction model (130) may be trained individually or independently of other models. In one embodiment of the present disclosure, the KPI prediction model (130) may be trained by being connected end-to-end with other models.

[0118] FIG. 9 is a diagram showing an example of the operation flow of an EMS in one embodiment of the present disclosure.

[0119] In describing Fig. 9, any description that overlaps with the description in any one of Figs. 1 to 8 may be omitted.

[0120] Referring to FIG. 9, a method of operation (900) of an EMS according to one embodiment of the present disclosure may include steps 910 to 972. In one embodiment of the present disclosure, steps 910 to 972 may be performed by at least one of an electronic device performing an EMS function, an electronic device included in the EMS, or an EMS module included in the electronic device. For example, at least some of steps 910 to 972 may be performed by a module included in the EMS. For example, at least some of steps 910 to 972 may be executed by at least one processor included in the electronic device. The method of operation (900) of an EMS is not limited to that illustrated in FIG. 9, and in one or more embodiments, additional steps not illustrated in FIG. 9 may be included, or some steps may be omitted.

[0121] In step 910, the management plane can determine the target application. For example, the management plane can select at least one of a plurality of applications as the target application.

[0122] In one embodiment of the present disclosure, the EMS may identify performance information, configuration information, or performance indicators that serve as the basis or target for the training or inference of an artificial intelligence (AI) model corresponding to a target application. For example, the EMS may identify (or determine) which information to use as performance information, which information to use as configuration information, or which indicator to use as a performance indicator, based on the target application. For example, the EMS may identify (or determine) the values ​​of performance information, configuration information, or performance indicators to be used in the training or inference of the AI ​​model.

[0123] At step 920, the AI ​​server may identify or determine whether the AI ​​model needs initial setup or update. In one embodiment of the present disclosure, the AI ​​server may identify or determine whether the AI ​​model needs to be trained (e.g., initial training or update). For example, the AI ​​server may determine the training of the AI ​​model periodically or non-periodically. For example, the AI ​​server may determine the training of the AI ​​model when certain conditions are met. For example, the AI ​​server may determine the training of the AI ​​model when the wireless network environment changes.

[0124] In one embodiment of the present disclosure, if it is identified that an initial setup or update of the AI ​​model is required, the AI ​​server may train the AI ​​model. For example, if it determines that an initial setup or update of the AI ​​model is required, the AI ​​server may initially train at least one of a context extraction model, a state transition model, or a KPI prediction model by performing at least one of steps 922 to 952. For example, if it determines that an initial setup or update of the AI ​​model is required, the AI ​​server may update at least one of a context extraction model, a state transition model, or a KPI prediction model that has been previously created or stored by performing at least one of steps 922 to 952.

[0125] In step 922, the AI ​​server can identify data for training the AI ​​model. For example, the AI ​​server can identify data for training at least one of a context extraction model, a state transition model, or a KPI prediction model. In one embodiment of the present disclosure, the AI ​​server can use network data obtained from at least one of an OAM or a management plane as data for training the AI ​​model. For example, the AI ​​server can store network data provided by at least one of an OAM or a management plane in a database. The AI ​​server can load network data stored in the database for training the AI ​​model.

[0126] In step 930, the AI ​​server can train a context extraction model. In one embodiment of the present disclosure, the AI ​​server can train a context extraction model for embedding base station situations. In one embodiment of the present disclosure, the AI ​​server can train a context extraction model corresponding to a target application based on data corresponding to a target application.

[0127] In step 932, the AI ​​server can store the trained context extraction model. For example, the AI ​​server can store the trained context extraction model in a database. For example, the AI ​​server can store a context extraction model corresponding to a target application.

[0128] In step 934, the AI ​​server can extract context information from performance information through inference during the training process of the AI ​​model. In one embodiment of the present disclosure, the AI ​​server can extract context data in a form recognizable by the AI ​​model from state-related performance information. In one embodiment of the present disclosure, the AI ​​server can extract context data from state-related performance information through inference during the training process of the context extraction model. FIG. 9 illustrates step 934 as a separate step following step 930, but is not limited thereto. For example, step 934 may be performed together with step 930 as part of step 930.

[0129] In one embodiment of the present disclosure, the context data extracted in step 934 may be used for training an AI model. For example, the extracted context data may be used to calculate a loss for training at least one of a context extraction model, a state transition model, or a KPI prediction model. For example, the extracted context data may be used as input data in inference during the training process of the state transition model or the KPI prediction model.

[0130] In step 940, the AI ​​server may train a state transition model. In one embodiment of the present disclosure, the AI ​​server may train a state transition model for inserting (or reflecting) configuration information. In one embodiment of the present disclosure, the AI ​​server may train a state transition model corresponding to a target application based on data corresponding to a target application.

[0131] In step 942, the AI ​​server can store the learned state transition model. For example, the AI ​​server can store the learned state transition model in a database. For example, the AI ​​server can store a state transition model corresponding to a target application.

[0132] In step 944, the AI ​​server can transform the state based on configuration information through inference during the training process of the AI ​​model. In one embodiment of the present disclosure, the AI ​​server can acquire or determine state-related performance information (e.g., transformed state-related performance information) into which configuration information to be applied is inserted (or reflected) through inference during the training process of the AI ​​model. For example, the AI ​​server can acquire or determine a transformed value of the state-related performance information based on a target value of the configuration information through inference during the training process of the AI ​​model. FIG. 9 illustrates step 944 as a separate step following step 940, but is not limited thereto. For example, step 944 may be performed together with step 940 as part of step 940.

[0133] In one embodiment of the present disclosure, the transformed state-related performance information in step 944 may be used for training an AI model. For example, the transformed state-related performance information may be used to calculate a loss for training at least one of a context extraction model, a state transformation model, or a KPI prediction model. For example, the transformed state-related performance information may be used as input data in inference during the training process of the context extraction model or the KPI prediction model.

[0134] In step 950, the AI ​​server can train a KPI prediction model. In one embodiment of the present disclosure, the AI ​​server can train a KPI prediction model for predicting base station performance. In one embodiment of the present disclosure, the AI ​​server can train a KPI prediction model corresponding to a target application based on data corresponding to a target application.

[0135] In step 952, the AI ​​server can store the trained KPI prediction model. For example, the AI ​​server can store the trained KPI prediction model in a database. For example, the AI ​​server can store a KPI prediction model corresponding to a target application.

[0136] In step 954, the AI ​​server may select an AI model corresponding to the target application and provide the selected AI model to the SON management module. In one embodiment of the present disclosure, the AI ​​server may select or identify a context extraction model corresponding to the target application among a plurality of context extraction models, a state transition model corresponding to the target application among a plurality of state transition models, and a KPI prediction model corresponding to the target application among a plurality of KPI prediction models, and provide them to the SON management module.

[0137] In one embodiment of the present disclosure, at least some of the operations of the AI ​​server for the context extraction model (e.g., steps 930 to 934), the operations of the AI ​​server for the state transition model (e.g., steps 940 to 944), or the operations of the AI ​​server for the KPI prediction model (950 and 952) may be performed independently or in parallel. In one embodiment of the present disclosure, at least some of the operations of the AI ​​server for the context extraction model (e.g., steps 930 to 934), the operations of the AI ​​server for the state transition model (e.g., steps 940 to 944), or the operations of the AI ​​server for the KPI prediction model (950 and 952) may be performed sequentially or simultaneously.

[0138] In one embodiment of the present disclosure, if it is determined that initial setup or updating of the AI ​​model is required, the AI ​​server may perform at least one of steps 922 to 952 to train or update at least a part of the AI ​​model, and then in step 954, select an AI model corresponding to the target application and provide it to the SON management module. In one embodiment of the present disclosure, if it is determined that initial setup or updating of the AI ​​model is unnecessary, the AI ​​server may omit steps 922 to 952, and in step 954, select an AI model corresponding to the target application and provide it to the SON management module.

[0139] In step 960, the SON management module can identify data for inference of an AI model. In one embodiment of the present disclosure, the SON management module can identify at least one of performance information or configuration information as input data for inference of an AI model. For example, the SON management module can identify an existing value (e.g., a current value) of performance information or a target value of configuration information to be used as input data for an AI model. For example, the SON management module can receive at least one of performance information or configuration information from at least one of a management plane or an AI server. For example, the SON management module can load at least one of performance information or configuration information stored in a database.

[0140] In step 962, the SON management module can identify an AI model corresponding to a target application. In one embodiment of the present disclosure, the SON management module can obtain an AI model corresponding to a target application from an AI server or a database. For example, the SON management module can load at least one of a context extraction model, a state transition model, or a KPI prediction model corresponding to a target application.

[0141] In step 964, the SON management module can extract base station situation information from performance information using a context extraction model. In one embodiment of the present disclosure, the SON management module can extract context data from existing state-related performance information using a context extraction model.

[0142] In step 966, the SON management module can transform state-related performance information by using a state transformation model to insert (or reflect) configuration information to be applied. In one embodiment of the present disclosure, the SON management module can obtain transformed state-related performance information by using a state transformation model to transform existing state-related performance information based on configuration information to be applied. For example, the SON management module can obtain transformed values ​​of state-related performance information based on target values ​​of configuration information and context data (e.g., context data corresponding to existing state-related performance information) using a state transformation model.

[0143] In step 968, the SON management module can predict KPIs based on performance information related to the transformed state using a KPI prediction model. In one embodiment of the present disclosure, the SON management module can determine, infer, or obtain predicted values ​​of KPIs based on performance information related to the transformed state, performance information related to the environment, and configuration information using a KPI prediction model. For example, the SON management module can determine, infer, or obtain predicted values ​​of KPIs based on the transformed value of the performance information related to the state, the existing value of the performance information related to the environment, and the target value of the configuration information using a KPI prediction model.

[0144] At step 970, the SON management module may select or determine parameters of the wireless network based on the predicted KPIs. In one embodiment of the present disclosure, the SON management module may determine parameter values ​​of the wireless network (e.g., base station setting values) based on the KPI prediction results (e.g., predicted values). In one embodiment of the present disclosure, the SON management module may determine base station settings based on the KPI prediction results using a policy model. For example, the SON management module may determine base station settings that optimize performance for the target application using a policy model corresponding to the target application.

[0145] At step 972, the SON management module can analyze the inference results of the AI ​​model (e.g., final inference results, intermediate inference results). In one embodiment of the present disclosure, the SON management module can analyze the inference results of at least one of a context extraction model, a state transition model, or a KPI prediction model. For example, the SON management module can perform at least one of visualization, network simulation, or root cause analysis on at least one of the final inference results or intermediate inference results. Accordingly, user-friendly real-time wireless network performance analysis or wireless network data analysis may be possible.

[0146] In one embodiment of the present disclosure, at least some of the management plane performing step 910, the AI ​​server performing steps 920 to 954, or the SON management module performing steps 960 to 972 may be included in the same electronic device. In one embodiment of the present disclosure, at least some of the management plane performing step 910, the AI ​​server performing steps 920 to 954, or the SON management module performing steps 960 to 972 may be included in different electronic devices.

[0147] FIG. 10 is a diagram illustrating an example of a method for an electronic device to predict network performance indicators in one embodiment of the present disclosure.

[0148] In describing Fig. 10, any description that overlaps with the description in any one of Figs. 1 to 9 may be omitted.

[0149] Referring to FIG. 10, a network performance indicator prediction method (1000) according to one embodiment of the present disclosure may include steps 1010 to 1030. In one embodiment of the present disclosure, steps 1010 to 1030 may be executed by at least one processor included in an electronic device. In one embodiment of the present disclosure, steps 1010 to 1030 may be performed by a SON management module included in an electronic device. The network performance indicator prediction method (1000) is not limited to that illustrated in FIG. 10, and in one or more embodiments, additional steps not illustrated in FIG. 10 may be included, or some steps may be omitted.

[0150] In step 1010, the electronic device can obtain context data from existing values ​​of state-related performance information of the base station using a first artificial intelligence model. For example, the state-related performance information may include at least one of an MCS (Modulation Coding Scheme) or a BLER (Block Error Rate).

[0151] In step 1020, the electronic device can obtain a transformed value of state-related performance information based on context data and a target value of configuration information for a base station using a second artificial intelligence model. For example, the configuration information for a base station may include at least one of a base station data retransmission period, a base station response timeout, a terminal data retransmission period, a terminal response timeout, or a target BLER (Block Error Rate).

[0152] In step 1030, the electronic device may use a third artificial intelligence model to obtain a predicted value of a network performance indicator (e.g., KPI) based on a converted value of state-related performance information and a target value of configuration information for a base station. For example, the network performance indicator may include at least one of an RRC (Radio Resource Control) Connection Drop Rate, DL (Downlink) IP (Internet Protocol) Throughput, UL (Uplink) IP Throughput, or VoLTE (Voice over LTE) Quality Defect Rate. In one embodiment of the present disclosure, the electronic device may use a third artificial intelligence model to obtain a predicted value of a network performance indicator based on a converted value of state-related performance information, a target value of configuration information, and environment-related performance information. For example, the environment-related performance information may include at least one of a CQI (Channel Quality Indicator), SINR (Signal to Noise Ratio), RI (Rank Indicator), or terminal information.

[0153] In one embodiment of the present disclosure, at least some of the first artificial intelligence model, the second artificial intelligence model, or the third artificial intelligence model used in the method (1000) may be connected end-to-end and trained. In one embodiment of the present disclosure, an electronic device may identify the first artificial intelligence model, the second artificial intelligence model, and the third artificial intelligence model corresponding to a target application among a plurality of applications. The electronic device may perform the method (1000) using the identified first artificial intelligence model, the second artificial intelligence model, and the third artificial intelligence model. In one embodiment of the present disclosure, the electronic device may perform at least one of visualization, simulation, or cause analysis on at least one of context data, transformed values ​​of state-related performance information, or predicted values ​​of network performance indicators.

[0154] FIG. 11 is a drawing showing an example of an electronic device according to one embodiment of the present disclosure.

[0155] In describing Fig. 11, any description that overlaps with the description in any one of Figs. 1 to 10 may be omitted.

[0156] The electronic device (1100) illustrated in FIG. 11 may be a computing device or a server device that predicts network performance indicators. For example, the electronic device (1100) may include a device that predicts network performance indicators using an artificial intelligence model. For example, the electronic device (1100) may be a communication device that constitutes a RAN, such as a base station or an EMS, or a server device that constitutes an existing RAN, or a separate server device that predicts network performance indicators.

[0157] In one embodiment of the present disclosure, an electronic device for learning or updating an artificial intelligence model for predicting network performance indicators may be the same as or different from an electronic device (1100) that performs prediction or inference using the artificial intelligence model. For example, if the electronic device for learning or updating the artificial intelligence model and the electronic device (1100) (i.e., the electronic device performing the prediction or inference operation) are different, the electronic device (1100) may receive the learned or updated artificial intelligence model from the electronic device for learning or updating the artificial intelligence model. One or more embodiments described below with respect to the electronic device (1100) may also be applied to the electronic device for learning or updating the artificial intelligence model for predicting network performance indicators.

[0158] In one embodiment of the present disclosure, the artificial intelligence model may be dynamically updated as a prediction or inference operation is performed. For example, at least one sub-model of the artificial intelligence model or at least a portion of the weights of the artificial intelligence model may be dynamically updated as a prediction or inference operation is performed. For example, if the electronic device updating the artificial intelligence model and the electronic device (1100) are the same, the electronic device (1100) may perform a prediction or inference using the artificial intelligence model and simultaneously update the artificial intelligence model. For example, if the electronic device updating the artificial intelligence model and the electronic device (1100) are different, the electronic device updating the artificial intelligence model may receive data identified, generated, or produced by the electronic device (1100) as it performs a prediction or inference using the artificial intelligence model, and update the artificial intelligence model based on the received data.

[0159] In one embodiment of the present disclosure, the electronic device (1100) may include at least one processor (1110) and a memory (1120), but is not limited thereto.

[0160] The processor (1110) is electrically connected to the components included in the electronic device (1100) and can perform operations or data processing regarding the control and / or communication of the components included in the electronic device (1100). In one embodiment of the present disclosure, the processor (1110) can load a request, command, or data received from at least one of the other components into memory for processing and store the processing result data in memory. In one embodiment of the present disclosure, the processor (1110) can process input data or control other components to process it according to data, operation rules, algorithms, methods, or models stored in memory (1120). For example, the processor (1110) can perform operations of a predefined operation rule, algorithm, method, module, or artificial intelligence model (e.g., neural network model) stored in memory (1120) using input data.

[0161] According to various embodiments, the processor (1110) may include at least one of a general-purpose processor such as a CPU (central processing unit), MPU (Micro Processor Unit), AP (application processor), 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). For example, if the processor (1110) is an artificial intelligence-dedicated processor, the artificial intelligence-dedicated processor may be designed with a hardware structure specialized for processing a specific artificial intelligence model.

[0162] The processor (1110) may include various processing circuits and / or multiple processors. For example, the term "processor" as used in the present disclosure, including in the claims, may include various processing circuits including at least one processor. One or more of the at least one processor may be configured to perform one or more functions of the present disclosure individually and / or collectively in a distributed manner. Where the "processor," "at least one processor," or "one or more processors" are described in the present disclosure as being configured to perform multiple functions, this may include a situation where one processor performs some of the functions and other processor(s) perform other parts of the functions, and a situation where a single processor performs all functions. Additionally, the at least one processor may include a combination of processors performing various functions in a distributed manner. The at least one processor may execute program instructions to achieve or perform various functions.

[0163] The memory (1120) is electrically connected to the processor (1110) and can store one or more modules, algorithms, operating rules, models (e.g., machine learning models, artificial intelligence models), programs, instructions, or data related to the operation of components included in the electronic device (1100). For example, the memory (1120) may include any non-transient computer-readable recording medium. For example, the memory (1120) can store one or more modules, algorithms, operating rules, models, programs, instructions, or data for processing and control of the processor (1110). The memory (1120) may include at least one type of storage medium among flash memory type, hard disk type, multimedia card micro type, card type memory (e.g., SD or XD memory, etc.), RAM (Random Access Memory), SRAM (Static Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM (Programmable Read-Only Memory), magnetic memory, magnetic disk, and optical disk, but is not limited thereto. The memory (1120) may not exist separately and may be configured to be included in the processor (1110). The memory (1120) may be composed of volatile memory, non-volatile memory, or a combination of volatile memory and non-volatile memory. A program or at least one instruction for performing operations according to the embodiments described above may be stored in the memory (1120). The memory (1120) may also provide stored data to the processor (1110) upon the request of the processor (1110).

[0164] In one embodiment of the present disclosure, the memory (1120) may store data and / or information identified, acquired, generated, or determined by the electronic device (1100). For example, the memory (1120) may store network data or weights of an artificial intelligence model. For example, the memory (1120) may store inference results of an artificial intelligence model (e.g., final inference results, intermediate inference results). For example, the memory (1120) may store data and / or information identified, acquired, generated, or determined by the electronic device (1100) in a compressed form. In one embodiment of the present disclosure, the memory (1120) may store predefined or determined information.

[0165] In one embodiment of the present disclosure, an electronic device (1100) may include a module that performs (or is used to perform) at least one operation. Some of the modules that perform at least one operation of the electronic device (1100) may be composed of a plurality of sub-modules or may constitute a single module. The modules that perform at least one operation of the electronic device (1100) may be implemented as a hardware module, a software module, and / or a combination thereof.

[0166] The memory (1120) may include software modules that perform at least some of the operations of the electronic device (1100) described above. In one embodiment of the present disclosure, the module included in the memory (1120) may perform operations by being executed by the processor (1110). For example, the module included in the memory (1120) (i.e., the software module) may include a program, model, operation rule, or algorithm configured to perform operations that derive output data for input data, which are executed according to the control or command of the processor (1110).

[0167] In one embodiment of the present disclosure, the memory (1120) may include a program, instructions, a neural network model, an artificial intelligence model, a machine learning model, a statistical model, an operation rule, or an algorithm for processing network data. For example, the memory (1120) may include an artificial intelligence model trained to predict or infer the value of a network performance indicator in a specific setting. For example, the memory (1120) may include the values ​​(weights) of a plurality of parameters constituting the artificial intelligence model.

[0168] The model contained in the memory (1120) can be created through learning. Here, being created through learning may mean that a basic artificial intelligence model is trained using multiple learning data by a learning algorithm, thereby creating an artificial intelligence model configured to perform a desired characteristic (or purpose). Such learning may be performed on the device itself where the artificial intelligence according to the present disclosure is performed, or it may be performed through a separate server and / or system. Examples of learning algorithms include supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning, but are not limited to the examples described above.

[0169] An artificial intelligence model may 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 operations of the previous layer and the multiple weights. The multiple weights possessed by the multiple neural network layers can be optimized by the learning results of the artificial intelligence model. For example, the multiple weights can be updated so that the loss value or cost value obtained by the artificial intelligence model during the learning process is reduced or minimized. For example, the artificial intelligence model may include, but is not limited to, a Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), Deep Neural Network (DNN), Recurrent Neural Network (RNN), Restricted Boltzmann Machine (RBM), Deep Belief Network (DBN), Bidirectional Recurrent Deep Neural Network (BRDNN), Transformer, or Deep Q-Networks. For example, artificial intelligence models may include, but are not limited to, statistical method models such as logistic regression, Gaussian Mixture Model (GMM), Support Vector Machine (SVM), Latent Dirichlet Allocation (LDA), or decision tree.

[0170] In one embodiment of the present disclosure, an artificial intelligence model may be composed of one or more sub-models. For example, the artificial intelligence model may include one or more sub-models that can be distinguished based on a function, input data, output data, loss function, or structure. For example, the artificial intelligence model may include one or more sub-models that are arbitrarily or logically distinguished.

[0171] In one embodiment of the present disclosure, at least one sub-model included in an artificial intelligence model may be trained or updated together with other sub-models. For example, at least one sub-model included in an artificial intelligence model may be trained or updated dependently with other sub-models. In one embodiment of the present disclosure, at least one sub-model included in an artificial intelligence model may be trained or updated independently of other sub-models. In one embodiment of the present disclosure, at least one sub-model included in an artificial intelligence model may not be trained or updated even with the training or updating of other sub-models. For example, at least one sub-model may be a model that has already completed training and may maintain a fixed weight without being updated even with the training or updating of the artificial intelligence model.

[0172] The electronic device (1100) may include more components than those shown in FIG. 11. In one embodiment of the present disclosure, the electronic device (1100) may further include a communication interface (or, communication module) for communicating with another device, server, or system. In one embodiment of the present disclosure, the electronic device (1100) may further include an input / output device and / or an input / output interface.

[0173] In the present disclosure, descriptions that overlap in FIGS. 1 to 11 may be omitted, and one or more embodiments described above in at least one of FIGS. 1 to 11 may be applied or implemented in combination with each other.

[0174] According to one embodiment of the present disclosure, a wireless communication environment reflecting specific base station settings can be predicted, and such prediction can assist in understanding changes in the user's wireless network status and enable interpretation of changes in the user's wireless network status.

[0175] According to one embodiment of the present disclosure, the effect of setting control parameters on the wireless communication environment can be determined by changing the state information of the wireless network according to the change of control parameters of the base station.

[0176] According to one embodiment of the present disclosure, a technology for predicting key performance indicators can be improved. For example, multiple performance indicators can be predicted by converting state information of a wireless network to pre-reflect the impact of a base station's control parameters on the wireless communication environment.

[0177] According to one embodiment of the present disclosure, a user can be supported through an interpretable AI. For example, the intermediate inference results of an AI model may be provided in a form that is easy for the user to perceive, use, or analyze. For example, in utilizing or analyzing the inference results of an AI model, technical usability and scalability can be ensured.

[0178] According to one embodiment of the present disclosure, an automatic learning and update function may be provided. For example, as the wireless network environment changes, an AI model automatic update function may be provided to continuously train the AI ​​model and optimize the algorithm.

[0179] In one embodiment of the present disclosure, a method for an electronic device to predict a network performance indicator may include the step of obtaining context data from an existing value of state-related performance information of a base station using a first artificial intelligence model. In one embodiment of the present disclosure, a method for an electronic device to predict a network performance indicator may include the step of obtaining a transformed value of the state-related performance information based on the context data and a target value of configuration information for the base station using a second artificial intelligence model. In one embodiment of the present disclosure, a method for an electronic device to predict a network performance indicator may include the step of obtaining a predicted value of the network performance indicator based on the transformed value of the state-related performance information and a target value of configuration information for the base station using a third artificial intelligence model.

[0180] In one embodiment of the present disclosure, the step of obtaining a predicted value of the network performance indicator may include the step of obtaining a predicted value of the network performance indicator based on the transformed value of the state-related performance information, the target value of the setting information, and the environment-related performance information using the third artificial intelligence model.

[0181] In one embodiment of the present disclosure, a method for an electronic device to predict network performance indicators may include the step of identifying the first artificial intelligence model, the second artificial intelligence model, and the third artificial intelligence model corresponding to a target application among a plurality of applications.

[0182] In one embodiment of the present disclosure, a method for an electronic device to predict a network performance indicator may include the step of determining a setting value of setting information for the base station based on the prediction result of the network performance indicator.

[0183] In one embodiment of the present disclosure, a method for an electronic device to predict a network performance indicator may include the step of performing at least one of visualization, simulation, or cause analysis on at least one of the context data, the transformed value of the state-related performance information, or the predicted value of the network performance indicator.

[0184] In one embodiment of the present disclosure, a program for performing a method of predicting network performance indicators by an electronic device on a computer may be recorded on a computer-readable recording medium.

[0185] In one embodiment of the present disclosure, an electronic device may include at least one processor comprising a processing circuit and a memory storing one or more instructions. In one embodiment of the present disclosure, by the at least one processor executing the one or more instructions individually or collectively, the electronic device may obtain context data from an existing value of state-related performance information of a base station using a first artificial intelligence model. In one embodiment of the present disclosure, by the at least one processor executing the one or more instructions individually or collectively, the electronic device may obtain a transformed value of the state-related performance information based on the context data and a target value of configuration information for the base station using a second artificial intelligence model. In one embodiment of the present disclosure, by the at least one processor executing the one or more instructions individually or collectively, the electronic device may obtain a predicted value of the network performance indicator based on the transformed value of the state-related performance information and a target value of configuration information for the base station using a third artificial intelligence model.

[0186] In one embodiment of the present disclosure, by having the at least one processor execute the one or more instructions individually or collectively, the electronic device can obtain a predicted value of the network performance indicator based on the transformed value of the state-related performance information, the target value of the setting information, and the environment-related performance information using the third artificial intelligence model.

[0187] In one embodiment of the present disclosure, the environment-related performance information may include at least one of CQI (Channel Quality Indicator), SINR (Signal to Noise Ratio), RI (Rank Indicator), or terminal information.

[0188] In one embodiment of the present disclosure, the state-related performance information may include at least one of an MCS (Modulation Coding Scheme) or a BLER (Block Error Rate).

[0189] In one embodiment of the present disclosure, the configuration information for the base station may include at least one of a base station data retransmission period, a base station response timeout, a terminal data retransmission period, a terminal response timeout, or a target BLER (Block Error Rate).

[0190] In one embodiment of the present disclosure, the network performance indicator may include at least one of RRC (Radio Resource Control) Connection Drop Rate, DL (Downlink) IP (Internet Protocol) Throughput, UL (Uplink) IP Throughput, or VoLTE (Voice over LTE) Quality Defect Rate.

[0191] In one embodiment of the present disclosure, at least some of the first artificial intelligence model, the second artificial intelligence model, or the third artificial intelligence model may be connected end-to-end and learned.

[0192] In one embodiment of the present disclosure, by having the at least one processor execute the one or more instructions individually or collectively, the electronic device can identify the first artificial intelligence model, the second artificial intelligence model, and the third artificial intelligence model corresponding to a target application among a plurality of applications.

[0193] In one embodiment of the present disclosure, by having the at least one processor execute the one or more instructions individually or collectively, the electronic device can determine a setting value of the setting information for the base station based on the predicted result of the network performance indicator.

[0194] In one embodiment of the present disclosure, by having the at least one processor execute the one or more instructions individually or collectively, the electronic device may perform at least one of visualization, simulation, or cause analysis of at least one of the context data, the transformed value of the state-related performance information, or the predicted value of the network performance indicator.

[0195] A device-readable storage medium may be provided in the form of a non-transitory storage medium. Here, 'non-transitory storage medium' simply means that it is a tangible device and does not contain a signal (e.g., electromagnetic waves), and the term 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.

[0196] 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., 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.

Claims

1. In a method for an electronic device to predict network performance indicators, A step of obtaining context data from existing values ​​of state-related performance information of a base station using a first artificial intelligence model; A step of obtaining a transformed value of the state-related performance information based on the target value of the context data and the configuration information for the base station using a second artificial intelligence model; and A method comprising the step of obtaining a predicted value of the network performance indicator based on the transformed value of the state-related performance information and the target value of the configuration information for the base station using a third artificial intelligence model.

2. In Paragraph 1, The step of obtaining the predicted value of the above network performance indicator is, A method comprising the step of obtaining a predicted value of the network performance indicator based on the transformed value of the state-related performance information, the target value of the setting information, and the environment-related performance information, using the third artificial intelligence model.

3. In Paragraph 2, A method in which the above environment-related performance information includes at least one of CQI (Channel Quality Indicator), SINR (Signal to Noise Ratio), RI (Rank Indicator), or terminal information.

4. In any one of paragraphs 1 through 3, A method in which the above state-related performance information includes at least one of MCS (Modulation Coding Scheme) or BLER (Block Error Rate).

5. In any one of paragraphs 1 through 4, A method comprising at least one of the following: a base station configuration information, the base station response timeout, the terminal data retransmission period, the terminal response timeout, or a target BLER (Block Error Rate).

6. In any one of paragraphs 1 through 5, A method in which the above network performance indicators include at least one of RRC (Radio Resource Control) Connection Drop Rate, DL (Downlink) IP (Internet Protocol) Throughput, UL (Uplink) IP Throughput, or VoLTE (Voice over LTE) Quality Defect Rate.

7. In any one of paragraphs 1 through 6, A method in which at least some of the first artificial intelligence model, the second artificial intelligence model, or the third artificial intelligence model are connected end-to-end and learned.

8. In any one of paragraphs 1 through 7, A method comprising the step of identifying the first artificial intelligence model, the second artificial intelligence model, and the third artificial intelligence model corresponding to a target application among a plurality of applications.

9. In any one of paragraphs 1 through 8, A method comprising the step of determining a setting value of setting information for the base station based on the predicted result of the network performance indicator.

10. In any one of paragraphs 1 through 9, A method comprising the step of performing at least one of visualization, simulation, or cause analysis on at least one of the context data, the transformed value of the state-related performance information, or the predicted value of the network performance indicator.

11. A computer-readable recording medium having a program recorded thereon for performing the method of any one of paragraphs 1 through 10 on a computer.

12. In an electronic device (1100), At least one processor (1110) including a processing circuit; and It includes a memory (1120) that stores one or more instructions, and By the above at least one processor (1110) executing the above one or more instructions individually or collectively, the electronic device (1100) Using a first artificial intelligence model, context data is obtained from existing values ​​of base station status-related performance information, and Using a second artificial intelligence model, a transformed value of the state-related performance information is obtained based on the target value of the context data and the configuration information for the base station, and An electronic device that obtains a predicted value of a network performance indicator based on a converted value of the state-related performance information and a target value of the configuration information for the base station, using a third artificial intelligence model.

13. In Paragraph 12, By the above at least one processor (1110) executing the above one or more instructions individually or collectively, the electronic device (1100) An electronic device that obtains a predicted value of the network performance indicator based on the transformed value of the state-related performance information, the target value of the setting information, and the environment-related performance information, using the third artificial intelligence model.

14. In Paragraph 13, An electronic device wherein the above environment-related performance information includes at least one of CQI (Channel Quality Indicator), SINR (Signal to Noise Ratio), RI (Rank Indicator), or terminal information.

15. In any one of paragraphs 12 through 14, An electronic device wherein the above-mentioned state-related performance information includes at least one of an MCS (Modulation Coding Scheme) or a BLER (Block Error Rate).

Images