5G Communication System and Method for Providing Analysis Information in 5G Communication System

Abstract

According to various embodiments, a 5G communication system including a network data analysis function and an operation, management, and maintenance system is described, wherein the operation, management, and maintenance system is configured to specify a machine learning model used by the network data analysis function to generate analysis information, and the network data analysis function is configured to receive the specification of the machine learning model and generate analysis information using the machine learning model.

Description

Technical Field

[0001] The present disclosure relates to a 5G communication system and a method for providing analysis information in a 5G communication system.

Background Art

[0002] In a communication network such as a 5G mobile communication network, it is important to ensure that a certain quality of service can be maintained. This includes monitoring network statistical analysis and prediction information (hereinafter also referred to as analysis information) generated from information such as load, resource usage, available components, user mobility, and component status, so that, for example, when an overload is imminent, a degradation in service quality can be avoided. The prediction information can be provided using a machine learning model. The quality of the prediction, which can be measured as the accuracy of the model, is an important factor since the prediction is used to optimize network operation. To achieve high accuracy, an approach that enables the use of a machine learning model appropriate for that purpose, for example, a machine learning model adjusted according to the requirements and / or characteristics of the communication system (e.g., how it is used by end users, i.e., subscribers) used to generate the analysis information, is desirable.

Summary of the Invention

[0003] According to various embodiments, a 5G communication system is provided that includes a network data analysis function and an operation, administration, and maintenance system, the operation, administration, and maintenance system being configured to specify a machine learning model used by the network data analysis function to generate analysis information, and the network data analysis function being configured to receive the specification of the machine learning model and generate analysis information using the machine learning model.

[0004] In the figures, like reference numerals generally represent like parts throughout the different figures. The drawings are not necessarily to scale, but rather are generally emphasized when explaining the principles of the present invention. In the following description, various aspects are described with reference to the following drawings.

Brief Description of the Drawings

[0005]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

[0006] The following detailed description refers to the accompanying drawings, which show specific details and the manner in which the present invention may be implemented, using the figures. Other aspects may be utilized and structural, logical, and electrical changes may be made without departing from the scope of the present invention. The various aspects of the present disclosure are not necessarily mutually exclusive, and some aspects of the present disclosure may be combined with one or more other aspects of the present disclosure to form new aspects.

[0007] Various examples corresponding to aspects of the present disclosure are described below.

[0008] Example 1 is a 5G communication system including a network data analytics function and an operation, administration, and maintenance (OAM) system, wherein the OAM system is configured to specify a machine learning model for use by the network data analytics function to generate analysis information, and the network data analytics function is configured to receive the specification of the machine learning model and generate analysis information using the machine learning model.

[0009] Example 2 is the 5G communication system according to Example 1, wherein the OAM system is configured to transmit the specification of the machine learning model to the network data analytics function.

[0010] Example 3 is the 5G communication system according to Example 1 or 2, wherein the operation, management, and maintenance system is configured to transmit information regarding the temporal validity, spatial validity, and / or access rights of the machine learning model to the network data analysis function, and the network data analysis function is configured to use the machine learning model to generate analysis information and / or provide access to the machine learning model in accordance with the temporal validity, spatial validity, and / or access rights of the machine learning model.

[0011] Example 4 is the 5G communication system according to any one of Examples 1 to 3, wherein the operation, management, and maintenance system is configured to transmit information regarding the machine learning model, including the storage location information of the machine learning model, to the network data analysis function.

[0012] Example 5 is the 5G communication system according to any one of claims 1 to 4, wherein the specification of the machine learning model is model identification, and the network data analysis function is configured to obtain the values of the trainable parameters of the machine learning model from the storage location of the machine learning model.

[0013] Example 6 further includes an analysis data repository function. The operation, management, and maintenance system is configured to specify the machine learning model to the analysis data repository function. The analysis data repository function is configured to obtain information regarding the values of the trainable parameters of the machine learning model from the storage location and store the information regarding the values of the trainable parameters of the machine learning model, and is configured to provide the information regarding the values of the parameters of the machine learning model to the network data analysis function in response to a request from the network data analysis function. This is the 5G communication system according to Example 1.

[0014] Example 7 is a 5G communication system according to any one of Examples 1 to 6, wherein the operation, management, and maintenance system is configured to transmit the specifications of one or more cases for generating analysis information using the machine learning model to the network data analysis function by the network data analysis function, and the network data analysis function is configured to generate analysis information in the one or more specified cases using the machine learning model.

[0015] Example 8 is a 5G communication system according to any one of Examples 1 to 6, including a network function, wherein the operation, management, and maintenance system is configured to specify the machine learning model for the network function, the network function provides the specifications of the machine learning model to the network data analysis function, and is configured to request the network data analysis function to provide analysis information to the network function using the specified machine learning model.

[0016] Example 9 is a 5G communication system according to Example 8, wherein the operation, management, and maintenance system is configured to transmit the specifications of one or more cases for generating analysis information using the machine learning model to the network function by the network data analysis function, the network function checks whether analysis information is required for one of the one or more cases, and when it is determined that analysis information is required for one of the one or more cases, is configured to request the network data analysis function to provide analysis information to the network function using the specified machine learning model.

[0017] Example 10 is that the operation, management, and maintenance system is configured to determine the value of the training parameter for training the machine learning model and notify the network data analysis function. The network data analysis function is configured to train the machine learning model to generate analysis information according to the value of the training parameter, and is the 5G communication system according to any one of Examples 1 to 9.

[0018] Example 11 is the 5G communication system according to Example 10, wherein the operation, management and maintenance system is configured to receive information about the training parameter supported by the network data analysis function, and determine the value of the training parameter for training the machine learning model supported by the network data analysis function and notify the network data analysis function.

[0019] Example 12 is a method for providing analysis information in a 5G communication system, wherein the operation, management and maintenance system of the 5G communication system designates a machine learning model for the network data analysis function of the 5G communication system to be used to generate analysis information, and the network data analysis function receives the specification of the machine learning model and uses the machine learning model to generate the analysis information.

[0020] Example 13 is the method according to Example 12, including determining information about the use of the 5G communication system by an end user and selecting the machine learning model according to the determined information about the use.

[0021] It should be noted that one or more of the features of any of the above examples may be lacking in any one of the other examples. In particular, the examples described in the context of a communication system are equally valid for methods, and vice versa.

[0022] According to a further embodiment, there is provided a computer-readable medium including a computer program and instructions that, when executed by a computer, cause the computer to execute the method according to any one of the above examples.

[0023] In the following, various examples will be described in more detail.

[0024] FIG. 1 shows a mobile radio communication system 100 configured in accordance with 5G (Fifth Generation) as specified by 3GPP (Third Generation Partnership Project), that is, a 5G communication system.

[0025] The mobile radio communication system 100 includes mobile radio terminal devices 102 such as UE (user equipment), nano equipment (NE), etc. The mobile radio terminal device 102 is also called a subscriber terminal and forms the terminal side. On the other hand, the other components of the mobile radio communication system 100 described below are parts of the mobile communication network side, that is, parts of a mobile network (for example, a Public Land Mobile Network (PLMN)).

[0026] Furthermore, the mobile radio communication system 100 includes a radio access network (RAN) 103 including a plurality of radio access network nodes, that is, base stations configured to provide radio access in accordance with 5G (Fifth Generation) radio access technology (5G New Radio). It should be noted that the mobile radio communication system 100 may be configured in accordance with LTE (Long Term Evolution) or another mobile radio communication standard (for example, non-3GPP access such as WiFi), but 5G is used as an example in this specification. Each radio access network node may provide wireless communication with the mobile radio terminal device 102 via a wireless interface. It should be noted that the radio access network 103 may include any number of radio access network nodes.

[0027] The mobile wireless communication system 100 further includes an Access and Mobility Management Function (AMF) 101 connected to the RAN 103, a Unified Data Management (UDM) 104, and a Network Slice Selection Function (NSSF) 105. Here and in the following examples, the UDM may further include an actual UE subscription database known, for example, as a Unified Data Repository (UDR). The core network 119 further includes an Authentication Server Function (AUSF) 114 and a Policy Control Function (PCF) 115.

[0028] The core network 119 can have a plurality of core network slices 106, 107, and for each core network slice 106, 107, an operator (also called a Mobile Network Operator (MNO)) can create a plurality of core network slice instances 108, 109. For example, the core network 119 includes a first core network slice 106 having three core network slice instances (CNIs) 108 for providing Enhanced Mobile Broadband (eMBB), and a second core network slice 107 having three core network slice instances (CNIs) 109 for providing Vehicle-to-Everything (V2X).

[0029] Typically, when a core network slice is deployed (i.e., generated), a network function (NF) is instantiated or (if already instantiated) referenced to form a core network slice instance, and the network functions belonging to the core network slice instance are configured using the identification of the core network slice instance.

[0030] Specifically, in the example shown, each instance 108 of the first core network slice 106 includes a first session management function (SMF) 110 and a first user plane function (UPF) 111, and each instance 109 of the second core network slice 107 includes a second session management function (SMF) 112 and a second user plane function (UPF) 113. The SMFs 110, 112 are for processing PDU (Protocol Data Unit) sessions, that is, for generating, updating, and deleting PDU sessions and managing session contexts by the user plane function (UPF).

[0031] The RAN 103 and the core network 119 form the network side of the mobile wireless communication system, that is, the mobile wireless communication network. The mobile wireless communication network and the mobile terminals accessing the mobile wireless communication network together form the mobile wireless communication system.

[0032] The mobile wireless communication system 100 can further include an OAM (Operation, Administration and Maintenance) system 116 implemented by, for example, one or more OAM servers connected to the RAN 103 and the core network 119 (connections are not shown for simplicity). The OAM 116 can include an MDAS (Management Data Analytics Service). The MDAS can provide, for example, an analysis report regarding the load of a network slice instance. The load of a network slice instance can be affected by various factors such as the number of UEs accessing the network, the number of QoS flows, and the resource utilization rate of various NFs related to the network slice instance.

[0033] Furthermore, the core network 118 includes an NRF (Network Repository Function).

[0034] The core network 119 may further include a Network Data Analytics Function (NWDAF) 117. The NWDAF is responsible for providing network analysis and / or prediction information in response to requests from network functions. For example, a network function may request specific analysis information regarding the load level of a particular network slice instance. Alternatively, a network function can use a subscription service to ensure that it is notified by the NWDAF when the load level of a network slice instance changes or reaches a specific threshold. The NWDAF 117 can have interfaces to various network functions on the mobile communication network side, such as the AMF 101, SMF 110, 112, PCF 115, etc. For simplicity, only the interface between the NWDAF 117 and the AMF 101 is shown.

[0035] For example, in NWDAF analysis, the number of UEs registered in a network slice instance and their observed service experience can be monitored. Not only does the OAM system enforce SLA (service level agreement) guarantees, but the 5GC NF can also take actions based on NWDAF slice QoE analysis to prevent further degradation of service experience in the network slice instance.

[0036] The NSSF105 or AMF101 can determine the timing when load distribution decisions are needed to address issues identified, for example, by processing the analysis results (i.e., network analysis or prediction information) provided by the NWDAF117. For example, if it is detected or predicted that a network slice instance is experiencing a degradation in service experience, new UE registrations or PDU sessions may be prevented from being assigned to that network slice instance by triggering the load distribution mechanism for the network slice. For example, the NSSF105, AMF101, and / or the OAM system 116 can also subscribe to both slice service experience and slice load analysis from the NWDAF117 simultaneously. Subscriptions to one or more S-NSSAIs and NSIs are possible.

[0037] To generate network analysis and / or prediction information, the NWDAF117 collects the necessary input data (e.g., for deriving slice service experience analysis), i.e., information for analyzing the state of the network slice instance. The NWDAF117 can obtain such types of information by subscribing to the network functions that are notified accordingly.

[0038] Figure 2 shows the collection of input data by NWDAF 201 (corresponding to NWDAF 117 for example) from components such as one or more network functions 202 and one or more application functions 203 or OAM 211. NWDAF 201 can also access the data in the data repository 212. From the collected information, NWDAF 201 derives analysis data for further components. These components may also be one or more network functions 204 and one or more application functions 205 (and may be the same as or partially the same as the network functions 202 and application functions 203 that collected the input data).

[0039] To generate analysis data, NWDAF can store (and implement) one or more machine learning (ML) models 206. Further, NWDAF includes, for example, an operating system 207, a data management component 208, a general management component 209, and an analysis management component 210. As will be described in detail below, according to various embodiments, the OAM system 211 (corresponding to OAM system 116) is used not only as a data source for NWDAF 201, but can also control the operation of NWDAF 201, particularly the use of the ML model 206, the acquisition of a new ML model 206 by NWDAF 201, and the training of the ML model 206.

[0040] According to 3GPP Release 17 (Rel-17), NWDAF 117, 301 is decomposed into two functions.

[0041] Figure 3 shows NWDAF AnLF (Analytics Logical Function) 301 and NWDAF MTLF (Model Training Logical Function) 302, which are connected, for example, by Nnwdaf.

[0042] The NWDAF 301 containing the Analytics logic function is denoted as NWDAF(AnLF) or NWDAF - AnLF or simply AnLF, can perform estimations, derive analysis information, and publish analysis services, namely Nnwdaf_AnalyticsSubscription or Nnwdaf_AnalyticsInfo.

[0043] The NWDAF 302 containing the model training logic function is denoted as NWDAF(MTLF) or NWDAF - MTLF or simply MTLF, can train a machine learning (ML) model, and publish a new training service (e.g., provide a trained model).

[0044] NWDAF(AnLF) 301 provides an analysis service. It may receive a connection from an NF303 (e.g., AMF or SMF) that functions as an (analysis) service consumer in order to provide analysis information (e.g., via the Nnf interface).

[0045] The input request parameters from NF303 to NWDAF(AnLF) 301 are, for example, an Analytic ID and an S - NSSAI. The output from NWDAF(AnLF) 301 is analysis information, such as statistics and / or predictions of the Analytic ID (also referred to here as network analysis information).

[0046] Examples of Analytic IDs are UE communication, UE mobility, UE behavior, user data congestion, and network performance, etc. Note that these Analytic IDs are not slice - specific.

[0047] NWDAF 301 can generate an analysis using a vendor - specific ML model embedded (stored or installed) therein. Specifically, NWDAF(MTLF) 302 trains one or more ML models to determine analysis information, and NWDAF(AnLF) 301 obtains one or more ML models from NWDAF(MTLF) 302 and generates analysis information for provision.

[0048] Figure 4 shows a flowchart 400 depicting the subscription / unsubscription of NWDAF (AnLF) as a service consumer 401 to NWDAF (MTLF) 402 for ML model provision in 403.

[0049] When NWDAF (AnLF) 401 is subscribed for model provision, for example, at 404, it is notified by NWDAF (MTLF) about the ML model (i.e., the values of trainable parameters such as the weights of the neural network).

[0050] Figure 5 shows a flowchart 500 depicting a request by NWDAF (AnLF) as a service consumer 501 to NWDAF (MTLF) 502 in 503 for information about an ML model trained by NWDAF (MTLF) 502 (e.g., the values of the trainable parameters of the ML model, e.g., the weights of the neural network).

[0051] In response to this request, NWDAF (MTLF) 502 can provide information about the ML model to NWDAF (AnLF) 501 at 504.

[0052] Two shareable NWDAFs (MTLF) (e.g., from different vendors) can share an ML model. For example, another NWDAF 304 can request NWDAF (MTLF) 302 for the ML model. The consumer NWDAF (here NWDAF 304) can also obtain the ML model stored by another NWDAF (MTLF) (e.g., here NWDAF (MTLF) 302) within the ADRF (Analytics Data Repository Function). NWDAF (MTLF) 302 can notify NWDAF 304 that it should obtain the ML model (requested by NWDAF 304) from the ADRF (when NWDAF 304 has determined that it should).

[0053] NWDAF 201 typically uses one or more ML models 206 installed (embedded) by a vendor in the NWDAF. However, for an operator of the wireless communication system 100, it may be desirable to use a custom model (e.g., a model obtained by the operator from a source other than the NWDAF vendor or a model designed (and perhaps pre-trained) by the operator).

[0054] For example, The NWDAF vendor provides ML model A to predict UE mobility patterns by the NWDAF. However, due to a special geographical configuration of the network (e.g., in the case of a special coverage area), the operator's simulations have shown that ML model B has better performance. Therefore, it may be desirable for the operator to use model B instead of model A in the NWDAF. The operator already has an AI engine built into a group of servers. The operator wants to connect the engine to the 5G communication system so that the AI engine can interact with the NWDAF. For example, to customize an ML model in the NWDAF, to provide a new ML model for a specific analysis of the NWDAF, to provide for distributed learning or federated learning, or to perform customized analysis by the NWDAF.

[0055] The vendor can also update and customize the NWDAF's ML models provided by the vendor.

[0056] In view of the above, an approach is provided that allows an operator to use a proprietary custom model to generate analysis information (e.g., prediction information) by a NWDAF in a 5G core network (5GC) according to various embodiments. For this purpose, procedures for providing, training, and using a custom model are provided, for example, procedures for providing a custom ML model in the 5GC, training a custom ML model in the 5GC, and using the custom ML model to generate analysis information. The operator can do this using an OAM system (e.g., OAM system 211).

[0057] FIG. 6 shows a flow diagram 600 showing a custom ML model provided from an OAM system 601 (e.g., corresponding to OAM system 211) to a NWDAF 602 (e.g., corresponding to NWDAF 201 according to one embodiment (“direct approach”)).

[0058] At 603, the OAM system 601 constructs a custom ML model (e.g., by other tools of the operator other than the OAM system tools of the OAM system 601 and e.g., via external AI engine interaction).

[0059] At 604, the OAM system 601 provides an ML model provision message to the NWDAF 602. The ML model provision message includes: (1) ML model information, i.e., information about the model such as: Model ID Model version Model size Model address (e.g., URL) (e.g., of the storage location) Model date Training data information (input features, output) Accuracy information The framework used in the model etc. (2) (Model) Provisioning parameters, i.e., information regarding the validity of the model, and who can access it Temporal validity Spatial validity Access control (e.g., whether it is shared with other NWDAFs, storage,...) etc.

[0060] At 605, NWDAF 602 obtains an ML model file (i.e., a file containing a complete description of the model including the values of the trainable parameters of the machine learning model such as the weights of the neural network, and a description of the model architecture including, for example, the type of activation function (e.g., sigmoid)), and registers the model for use with one or more analysis IDs. NWDAF 602 can obtain the ML model file from a URL (uniform resource locator) provided by the OAM system 601 as part of the model information, for example.

[0061] At 606, NWDAF 602 responds to the ML model provision message, for example, by positively responding to the receipt and storage of the ML model.

[0062] Figure 7 shows a flowchart 700 for providing a custom ML model from an OAM system 701 (e.g., corresponding to the OAM system 211) to a NWDAF 702 (e.g., corresponding to the NWDAF 201 in another embodiment ("indirect approach")) involving the participation of the ADRF 703.

[0063] At 704, the OAM system 701 constructs a custom ML model (e.g., by other tools of an operator other than the OAM system tools of the OAM system 701, and, for example, via external AI engine interactions).

[0064] At 705, the OAM system 701 provides the ADRF 703 with an ML model storage message. The ML model storage message contains ML model information (see the above example).

[0065] At 706, the ADRF 701 retrieves the ML model file, stores the model, and at 707 sends an affirmative response for the storage of the ML model.

[0066] At 708, the OAM system 701 provides the NWDAF 702 with an ML model provision message. The ML model provision message contains model provision parameters (see the above example).

[0067] In response to the reception of the ML model provision message, at 709, the NWDAF 702 sends a model acquisition request to the ADRF 703. In response to this, at 710, the ADRF 703 sends a model acquisition response message containing the ML model information and the (complete) specification of the model (e.g., the ML model file) to the NWDAF 702.

[0068] At 711, the NWDAF 702 acquires and registers the model for use with one or more analysis IDs.

[0069] At 712, the NWDAF 702 responds to the ML model provision message, for example, by sending an affirmative response for the reception and storage of the ML model.

[0070] FIG. 8 shows a flowchart 800 illustrating a procedure by which an OAM system 801 controls an NWDAF 802 to train a custom ML model according to an embodiment.

[0071] At 803, the OAM system 801 determines which model training parameters to configure (e.g., by other tools of an operator other than the OAM system tools of the OAM system 801 and, for example, via external AI engine interactions).

[0072] At 804, the OAM system 801 provides a training negotiation request to the NWDAF 802.

[0073] The training negotiation request includes the following available training parameters (i.e., training parameters that the OAM system 801 can provide values for): Loss function Learning rate Number of epochs Batch size Number of iterations Optimization algorithm Dropout rate Etc. These are used for training.

[0074] At 805, the NWDAF 802 responds with a training negotiation response that includes the training parameters it supports (i.e., that can be set by the OAM system 801).

[0075] In this example, assume these are as follows: Loss function Learning rate Optimization algorithm At 806, the OAM system 801 determines the values of the training parameters supported (e.g., by other tools of an operator other than the OAM system tool of the OAM system 801 and also e.g., via external AI engine interactions). According to the above example of the training parameters supported by the NWDAF 802, for example, determine the following: Loss function: MSE (mean squared error) Learning rate: 0.05 Optimization algorithm: SGD

[0076] At 807, the OAM system 801 provides the determined values to the NWDAF 802, and the NWDAF gives a positive response at 808.

[0077] At 809, the NWDAF 802 trains a custom model using the training parameter values provided by the OAM system 801, and at 810, notifies the OAM system 801 about the training (e.g., about having completed the training).

[0078] FIG. 9 shows a flowchart 900 illustrating the procedure by which an OAM system 901 controls the use of a custom ML model by an NWDAF 902 to provide analysis information to an NF 903 according to one embodiment (the “direct approach”).

[0079] At 904, the OAM system 901 sends a configuration message to the NF 902. The configuration message indicates an analysis ID and a model ID, and includes analysis customization parameters that specify the cases in which the model identified by the model ID should be used by the NWDAF 902 for the analysis specified by the analysis ID.

[0080] Target area Target period Target UE Target NF etc.

[0081] At 905, in response to the configuration message, the NWDAF 902 registers the specified model as the primary model for the analysis ID of the specified case, and at 906 sends an affirmative response to the configuration of the configuration message.

[0082] Thereafter, at 907, the NWDAF 902 can train the model (e.g., under the control of the OAM system 901 described with reference to FIG. 8).

[0083] At 908, the NF 903 subscribes to the analysis service. The NWDAF 902 (specifically, the NWDAF's NWDAF (AnLF)) sends an acknowledgement response to the subscription at 909.

[0084] In response to the subscription, NWDAF 902 generates analysis information using a model at 910 (assuming that the analysis subscribed to by NF903 meets the specified case), and at 911, notifies NF903 about the analysis information using one or more notification messages.

[0085] Figure 10 shows a flowchart 1000 illustrating the procedure by which an OAM system 1001 controls the use of a custom ML model by NWDAF 1002 to provide analysis information to NF1003 according to another embodiment (the "indirect approach").

[0086] At 1004, the OAM system 1001 sends a suitable model message to NF1003. Referring to the example described above with respect to Figure 9, the configuration message indicates an analysis ID and a model ID, and includes analysis customization parameters that specify the case in which the model identified by the model ID is to be used by NWDAF 1002 for the analysis specified by the analysis ID.

[0087] At 1005, in response to the suitable model message, NF1003 registers the specified model as the model for the analysis ID of the specified case, and at 1006, positively responds to the configuration of the suitable model.

[0088] At 1007, NF1003 subscribes to the analysis service of the analysis ID. In the subscription, NF1003 indicates the model (registered as the suitable model for the analysis ID), assuming that NF1003 requests an analysis of the specified case. NWDAF 1002 (specifically, the NWDAF (AnLF) of NWDAF) acknowledges the subscription at 1008.

[0089] In response to the subscription, NWDAF 1002 generates analysis information using the model indicated by NF1003 at 1009, and at 1010, notifies NF1003 about the analysis information using one or more notification messages.

[0090] In summary, according to various embodiments, a 5G communication system including a network data analysis function and an operation, administration, and maintenance system is provided (see, for example, FIG. 2). The operation, administration, and maintenance system is configured to specify a machine learning model used by the network data analysis function to generate analysis information. The network data analysis function is configured to receive the specification of the machine learning model and generate analysis information using the machine learning model.

[0091] The 5G communication system can be understood as a communication system according to any 3GPP release.

[0092] According to various embodiments, a method is provided as shown in FIG. 11.

[0093] FIG. 11 shows a flowchart 1100 illustrating a method of providing analysis information in a 5G communication system.

[0094] At 1101, the operation, administration, and maintenance system of the 5G communication system specifies a machine learning model for the network data analysis function of the 5G communication system to be used to generate analysis information.

[0095] At 1102, the network data analysis function receives the specification of the machine learning model.

[0096] At 1103, the network data analysis function uses the machine learning model to generate analysis information.

[0097] In other words, according to various embodiments, an operator of a 5G communication system can configure one or more machine learning models used by the NWDAF to provide analysis information using the operator's OAM tool.

[0098] Accordingly, according to various embodiments, an operator can use one or more custom models for analysis in a 5G communication system (particularly in the 5GC which is the core network of the 5G communication system). As described above, according to various embodiments, methods for providing a custom ML model in NWDAF, methods for training a custom ML model, and methods for using a custom model for analysis generation in 5GC are provided.

[0099] Accordingly, instead of using one or more ML models embedded in NWDAF by an NWNDAF vendor, an operator can use one or more other ML models to generate analysis information, that is, the operator can manage which ML models are used within the operator's communication system, particularly within the core network (without having to rely on the ML models provided by the vendor).

[0100] Analysis information generated by one or more ML models (the "custom" models, i.e., one or more models provided by the operator to NWDAF for NWDAF to use in generating analysis information) can be used for other NFs to make decisions. For example, PFC: Derive appropriate policies SMF: Select UPF AMF: Paging method NEF: Selection of member UEs for application functions

[0101] Thereby, the operator can obtain more accurate analysis information that can improve decision-making in the network (for example, the operator can use a more appropriate ML model for the operator's network, such as an ML model adapted to the characteristics of the operator's communication system and / or service area, etc.). Therefore, the operator can improve network operation, enhance the quality of service and the quality of experience, and reduce costs, etc.

[0102] Components of a communication system may be implemented, for example, by one or more circuits. A "circuit" may be understood as any kind of logic implementation entity that may be a dedicated circuit or a processor that executes software, firmware, or any combination thereof stored in a memory. Thus, a "circuit" may be a programmable processor, such as a hardwired logic circuit or a programmable logic circuit like a microprocessor. A "circuit" may also be a processor that executes software, such as any kind of computer program. Any other kind of implementation of each of the functions described above may also be understood as a "circuit".

[0103] Although particular embodiments have been described, those skilled in the art will understand that various changes in form and detail may be made without departing from the spirit and scope of the embodiments of the disclosure as defined in the appended claims. The scope is thus indicated by the appended claims and should include all modifications within the equivalent meaning and scope of the claims.

Claims

1. A 5G communication system, comprising a network data analysis function and an operation, management and maintenance system, wherein the operation, management and maintenance system specifies a machine learning model used by the network data analysis function to generate analysis information, and transmits to the network data analysis function the specifications of one or more cases in which the network data analysis function uses the machine learning model to generate analysis information, and is configured as such, wherein the network data analysis function receives the specifications of the machine learning model, and uses the machine learning model to generate analysis information in the one or more specified cases, and is configured as such, a 5G communication system.

2. The 5G communication system according to claim 1, wherein the operation, management and maintenance system is configured to transmit the specifications of the machine learning model to the network data analysis function.

3. The 5G communication system according to claim 1 or 2, wherein the operation, management and maintenance system is configured to transmit to the network data analysis function information regarding the temporal validity, spatial validity and / or access rights of the machine learning model, and the network data analysis function is configured to use the machine learning model to generate analysis information and / or provide access to the machine learning model in accordance with the temporal validity, spatial validity and / or access rights of the machine learning model.

4. The 5G communication system according to claim 1, wherein the operation, management and maintenance system is configured to transmit to the network data analysis function information regarding the machine learning model including the storage location information of the machine learning model.

5. The 5G communication system according to claim 1, wherein the specifications of the machine learning model are model identification, and the network data analysis function is configured to obtain the values of the trainable parameters of the machine learning model from the storage location of the machine learning model.

6. further comprising an analysis data repository function, wherein the operation, management and maintenance system is configured to specify the machine learning model to the analysis data repository function, The analysis data repository function is configured to obtain information regarding values of trainable parameters of the machine learning model from a storage location, store the information regarding the values of the trainable parameters of the machine learning model, and provide the network data analysis function with the information regarding the values of the parameters of the machine learning model in response to a request from the network data analysis function. The 5G communication system according to claim 1.

7. The operation, management, and maintenance system is configured to determine values of training parameters for training the machine learning model and notify the network data analysis function of the values. The network data analysis function is configured to train the machine learning model to generate analysis information according to the values of the training parameters. The 5G communication system according to claim 1.

8. The operation, management, and maintenance system is configured to receive information regarding training parameters supported by the network data analysis function, determine values of training parameters for training a machine learning model supported by the network data analysis function, and notify the network data analysis function of the values. The 5G communication system according to claim 7.

9. A method for providing analysis information in a 5G communication system, comprising: the operation, management, and maintenance system of the 5G communication system designating a machine learning model used by the network data analysis function of the 5G communication system to generate analysis information, and transmitting to the network data analysis function specifications of one or more cases in which the network data analysis function uses the machine learning model to generate analysis information; the network data analysis function receiving the specifications of the machine learning model and generating analysis information using the machine learning model in the one or more designated cases. A method including the above steps.

10. The method according to claim 9, further comprising determining information regarding use of the 5G communication system by an end user and selecting the machine learning model according to the determined information regarding use.

11. A computer program and a computer-readable medium including instructions that, when executed by a computer, cause the computer to execute the method according to claim 9 or 10.

Images