Method of providing data related to at least one device of a network user, method of obtaining data and entity implementing these methods

Abstract

The invention relates to a method for providing a data analysis facility (DAF) with data relating to at least one user equipment, the method comprising: - a step of collecting, for a plurality of events in the network affecting at least one state of the user equipment, data relating to said at least one affected state; and - a step of providing the analysis facility (DAF) with, for at least one such state of the user equipment, at least one statistical datum obtained for the state by aggregating the data relating to the state collected during the collecting step, and a portion of said data corresponding to an integer X greater than or equal to 1 events among the most recent events of said plurality of events.

Description

TECHNICAL FIELD

[0001] The present application relates to the general field of telecommunications. BACKGROUND

[0002] Modern telecommunication networks, such as the fifth generation or 5G networks defined by the 3GPP standards, face complex situations, in particular caused by the very large number of terminals to be managed, the diversity of the uses implemented by the network (and the requirements in terms of latency, throughput, final capacity) and the diversity of the behavior of the users of the network in time and space.

[0003] To address these complex situations, operators envisage implementing, within their network, one or more specialized entities to perform statistical analysis and predictions (also commonly referred to as "analytics" in the 3GPP documents) in terms of quality of service regarding requests and responses provided by the network. These predictions can be global predictions, i.e. predictions established on the network, the servers, the applications or even the regions. Examples of global predictions are the load rate of network resources, the average quality of service, the number of users connected to the network via user equipment (or simply "UE" in the following) or the number of active sessions. Individual predictions can also be established, i.e. predictions related to a user or a group of users, for example the future location of the UE of a user or the amount of future communication sessions established by a user via his UE. Such an entity is also called a network data analytics function (DAF). By way of illustration, in the core of a 5G mobile network, the network data analytics function or "NWDAF" plays such a role.

[0004] This prediction hypothesis is fulfilled by the DAF function thanks to the different entities forming the network that collect beforehand data representative of the network facts (e.g. connection status of a UE, cell in which it is located, etc., hereinafter called “raw data”) usually also called “network functions” (or NFs). These raw data can be global (overall) data per NF function or even related to each user. Once the predictions are established, expected corrective modifications can be made on the network’s parameters to optimize the network’s operation. The entities using these predictions are usually the NF functions, the DAF function’s client functions, which can be different from the NF functions collecting the raw data and providing them to the DAF function or identical. These client NF functions are then able to adjust their behavior according to the predictions received from the DAF function in order to optimize the network’s operation and the quality of the service delivered to each user on the UEs. The document 3GPP TR 23.791 v16.2.0 entitled “Technical Specification Group Services and System Aspects; Study of Enablers for Network Automation for 5G (Release 16)” published in June 2019 describes different use cases of such predictions in a 5G network.

[0005] The amount of global raw data collected by the DAF function from each NF function (e.g. load level of a NF function, its operational status, etc.) can be quite important but the number of NF functions requested by the DAF function is actually quite small, which limits the complexity generated by the collection of these global raw data.

[0006] However, this is not the case for individual raw data related to a user or a group of users, e.g. the location or activity of a user or a group of users, the corresponding changes of state of their UEs (registered, connected, standby connected, etc.). The collection of these individual raw data in order to establish predictions for each user of the network under consideration requires the feedback of all the network events related to the UE of this user to the DAF function, where the amount of raw data to be fed back can vary from one event to another. Since the number of users of the network to be monitored by the DAF function is huge, the amount of individual raw data to be collected by the DAF function can prove to be extremely large, which not only leads to a large amount of signaling to be exchanged between the NF functions feeding back these individual raw data and the DAF function but also to a large computational load of the DAF function.

[0007] Moreover, even if the statistical analysis or the prediction from the DAF function required by the customer NF function at a given moment is not directly targeted at a certain user, the DAF function must continuously collect information about network events related to the UE of this user in order to be able to perform such statistical analysis or such prediction at the moment the request will occur, if applicable.

[0008] There is therefore a need to reduce the amount of raw data collected (and incidentally processed) by the DAF function by means of the various NF functions.

[0009] In response to this need, according to the 3GPP standard, document S2-186464 (entitled "Solution to Key Issues 9", Ericsson, 2-6 July 2018) proposes to integrate a local DAF function into the NF functions in order to be able to perform the prediction calculations directly on the NF functions. The predictions thus locally performed by the NF functions can be used directly or fed back to the central DAF function in order to be used for more complex predictions. The central DAF function can also subscribe to the raw information fed back collected by the NF functions.

[0010] In the solution proposed in document S2-186464, the reduction of the amount of information exchanged between the NW functions and the DAF function is only allowed when the DAF function uses the predictions made by the NF functions. This leads to a potential loss of information that can be detrimental to the statistical analysis and the predictions performed by the DAF function, as they are constrained by the predictions performed by the NF functions.

[0011] Alternative solutions to the one proposed in document S2-186464 can be envisaged to reduce the signaling exchanged between the DAF function and the NF functions of the network.

[0012] The solution can therefore involve a continuous collection of individual raw data only for a sample of users considered to be representative of all users. However, such representativeness can only be established by comparison with the actual behavior of the targeted users. The implementation of such a solution therefore requires the use of a technique for a priori classification of the behavior of each user before being able to apply the relevant representativeness model. The use of such a solution can prove to be complex in some cases, in particular for users who are mobile, talkative or even occasional speakers, or for vehicles or drones.

[0013] Another solution can involve performing continuous or batch logging of raw data collected by the NF functions in a data warehouse already present in the network, also called a "data lake", for example a data lake containing metering tickets (or "charging data records", CDRs) containing statistics reflecting the calls and data sessions established on the network. This results in logs of raw data collected by the NF functions about the users they manage.

[0014] The use of such a data warehouse limits the signaling exchanged between the NF functions and the DAF functions. When a request about statistics and / or predictions for a user arrives at a DAF function, said function can check the data warehouse thus provisioned by the NF functions in order to adjust the statistics and predictions it keeps about this user. The DAF function can also have to consult the data warehouse at various appropriate times in order to determine the usual behavior of the user.

[0015] If this solution is based on an already existing data lake, like the one for metering tickets mentioned previously, this data lake can contain only a part of the raw data needed to optimize the operation of the network. Moreover, the problem of the quantity of raw data exchanged remains between the data warehouse and the NF functions.

[0016] Another solution involves introducing a dedicated mediation entity in the network in the vicinity of the NF functions. This dedicated mediation entity is configured to collect raw data from the NF functions and distribute these raw data to the DAF functions that need them. The drawback of this solution is the introduction of a new functional entity in the network, thus requiring new signaling exchanges. This can have a significant impact on the deployment, especially in terms of complexity, computing load or even operational management.

[0017] It should be noted that although the previous considerations are introduced with reference to a 5G network defined by the 3GPP standard and to NF and NWDAF entities, these considerations are also applicable to any type of telecommunication network in which an operator tries to collect data from the various functions of the network and make predictions based on these data in order to improve the operation of the network. SUMMARY

[0018] The present invention particularly addresses the above-mentioned drawbacks of the prior art by providing a method for providing data related to at least one user equipment to an entity of a network, called data analytics entity, wherein the method comprises:

[0019] - a step of collecting data related to at least one state of said user equipment, for a plurality of events in the network affecting said at least one state of said user equipment; and

[0020] - for at least one of said affected states of the user equipment, providing to the data analysis entity, the following: at least one statistical data acquired for this state by aggregating the data related to this state collected during the collecting step; and a portion of the data related to this state collected during the collecting step, this portion corresponding to an integer X greater than or equal to 1 of the most recent events of said plurality of events.

[0021] Correspondingly, the application also relates to an entity comprising a module activated for at least one user equipment, and comprising:

[0022] - a collecting module configured to collect, for a plurality of events affecting at least one state of this user equipment in a network, data related to said at least one state affected by said events;

[0023] - a module for acquiring statistical data configured to acquire, for at least one of said affected states of the user equipment, at least one statistical data by aggregating the data related to this state collected by the collecting module; and

[0024] - a transmission module configured to provide, for at least one of said affected states of the user equipment, to an entity of the network, called data analysis entity, the at least one statistical data acquired for this state by the module for acquiring statistical data and a portion of the data related to this state collected by the collecting module, this portion corresponding to an integer X greater than or equal to 1 of the most recent events of said plurality of events.

[0025] For simplicity, this entity is hereinafter referred to as data providing entity or providing entity.

[0026] In a particular embodiment, this entity is an entity of the network, and more particularly a NF (Network Function) entity hosting a network function, for example an AMF (Access and Mobility Management Function) entity or an SMF (Session Management Function) entity of a 5G network according to the 3GPP standard.

[0027] Although the technical problem that the application seeks to solve has been highlighted with reference to the exchange between a NF entity of a network and a DAF or NWDAF data analytics entity, as already mentioned above, the application is not limited to this case. Thus, in another embodiment, the data providing entity can also be a user equipment (i.e. UE) of a user, for example a terminal, a sensor, a Private Automatic Branch Exchange (PABX) of a company, a customer equipment of the CPE (Customer Premises Equipment) type, etc. The nature of the user equipment in relation with the application is not limited. For the sake of simplicity, this user equipment is referred to hereinafter by the equivalent use of the expressions "user equipment", "equipment of a user" or even "UE". Thus, the application allows an analytics entity to acquire data related to a user of a network, originating from various sources, which have access to various types or levels of data about the user.

[0028] The application also relates to a method of allowing an entity of a network, called data analytics entity, to acquire data related to at least one user equipment, said analytics entity being able to communicate with at least one other entity configured to collect, for a plurality of events in the network affecting at least one state of said user equipment, data related to said at least one affected state, said method comprising the step of receiving, for at least one of said affected states of said user equipment:

[0029] - at least one statistical data acquired by the other entity for this state by aggregating the data related to this state collected by the other entity for said plurality of events; and

[0030] - a portion of the data related to this state collected by the other entity for said plurality of events, said portion of data corresponding to an integer X greater than or equal to 1 of the most recent events of said plurality of events.

[0031] Correspondingly, the application also relates to an entity of a network, called data analytics entity, able to communicate with another entity configured to collect, for a plurality of events in the network affecting at least one state of at least one user equipment, data related to said at least one affected state, wherein said analytics entity comprises a receiving module activated for said at least one user equipment and configured to acquire, for at least one of said affected states of said user equipment, from the other entity:

[0032] - at least one statistical data acquired by the other entity for this state by aggregating the data related to this state collected by the other entity for said plurality of events; and

[0033] - a portion of the data collected by said other entity in relation with the status for said plurality of events, wherein said portion corresponds to an integer X greater than or equal to 1 of the most recent events of said plurality of events.

[0034] In a particular embodiment, the acquisition method comprises a step of formulating a prediction or a synthetic statistical analysis based on said at least one acquired statistical data and said data portion.

[0035] Correspondingly, the analysis entity further comprises an analysis module configured to formulate a prediction or a synthetic statistical analysis based on said at least one acquired statistical data and said data portion.

[0036] The analysis entity can be a dedicated and centralized data analysis entity of the network for this analysis function, which is hereinafter referred to as DAF entity for simplicity. Such an entity is for example the NWDAF entity of a 5G network.

[0037] However, the application can also apply to other entities. Thus, the analysis entity can be an entity of the network other than a dedicated and centralized DAF or NWDAF entity, provided that this entity is configured to analyze data of the network (i.e. data related to the network, to the operation of the network, etc.). For example, it can be envisaged that the analysis entity is a NF entity within the meaning of the application, which is appropriately configured to perform a statistical analysis or a prediction.

[0038] It should be noted that the nature of the entities involved in the application is not limited, i.e. the data providing entity and the analysis entity are not limited. Indeed, it can relate to a software entity or a function hosted by various physical devices of a network entity or a physical entity.

[0039] The nature of the status of the user equipment in the network that can be envisaged within the scope of the application is also not limited, wherein a user can designate a single user or a group of users, for example a company or even a machine (e.g. a sensor, an actuator, a robot, etc.) operated by a user or a group of users. The notion of user is associated in this text with the existence of a subscription of this user to the network operator. The status of the user equipment can also refer equally to a general status of this user equipment, for example "registered", "connected", "not connected" or more specific statuses such as "present in cell A", "in progress of voice communication", etc. within the meaning of the application.

[0040] Then, within the meaning of the application, an event is reflected as a transition between two statuses of the user equipment (for example, a transition from "not registered status" to "registered" status, or "user equipment arrival in cell A"). The affected status of the user equipment can be inferred from the events detected in the network for this user equipment (for example, start or end of a communication, entry into a cell, etc.).

[0041] The application advantageously proposes a data providing entity (e.g. a NF entity) that feeds back to an analytics entity (e.g. a DAF entity) two types of collected information based on determined events detected by the providing entity in the network for a given user, in particular to allow the analytics entity to build statistical analysis and / or predictions, wherein said two types of collected information are:

[0042] - on the one hand, statistical data acquired by aggregating raw data (i.e. network facts) collected during these events with respect to the state of the user equipment impacted by these events. It should be noted that within the meaning of the application, statistical data are built by merging (i.e. by aggregating) raw data collected over a determined duration (e.g. in a number of determined events or during a determined time window) by means of a given mathematical function (e.g. mean, variance, etc.), i.e. the statistical data are merely acquired from past raw data (as opposed to predictions providing future indications). To some extent, it involves a compressed version of the past raw data that the providing entity can acquire very easily; and

[0043] - on the other hand, raw data collected with respect to recent events that have impacted one or more states of the user equipment related to the user. These data reflect a recent past of the user equipment observed in the network. It should be noted that this recent past can have different durations depending on the considered user and the respective activity of these users in the network (for some users, the event X can be distributed over a longer time window than other more generic events in the network). In a particular embodiment, it can be envisaged to configure the number of events X as a function of the type of considered user (e.g. mobile user, fixed user, etc.).

[0044] These two types of information advantageously allow the exchange between the data providing entity and the analytics entity to be minimized while preserving the relevance of the feedback information. By feeding back raw data related only to the recent past of the considered user equipment, combined with statistical data representative of a longer period, the application offers to the analytics entity the possibility to combine two different levels of information and to complement its knowledge of the user behavior by effectively matching these two levels of information. This matching can be performed in any way, for example by an evolutionary prediction algorithm based on Markov chains. The raw data corresponding to the recent past of the user equipment actually provide contextual information (lost when aggregating data to form statistical data) but over a limited time window, which makes the amount of information fed back to the analytics entity reduced. The number X of events fed back can obviously be particularly configured as a function of the events of interest of the user.

[0045] Like the analyzing entity, the providing entity thus does not need to store exhaustively all the raw data collected about the events related to the users detected in the network. The providing entity also does not need to transmit all the data it collects to the analyzing entity. Moreover, the pre-aggregated statistical data provided by the providing entity makes the computing and protocol management load of the analyzing entity decrease.

[0046] By way of illustration, the providing entity is for example an AMF network function for managing the access and mobility of the 5G network. This AMF function can provide the DAF entity of the 5G network (analyzing entity within the meaning of the invention) with aggregated statistical data, the most frequent locations of the user equipment over different time ranges, and the last events related to the mobility and the access of this user equipment to the network, including its last location. Thus, the DAF entity can determine, via the list of the last feedback events, when the user with the user UE is close to one of its favorite locations (provided by the feedback statistical data). Moreover, with the two types of information fed back by the NF entity, the knowledge of the network topology combined with the model of the long-term behavior of the user established by the DAF entity makes it possible for the DAF entity to predict the next location of the user (in other words, of its UE).

[0047] Of course, this example is provided by way of illustration only and other NF functions (for example, a session management function (SMF)) and / or other events and / or other statistical data (for example, communication statistical data) can be envisaged.

[0048] In a particular embodiment, for each detected event related to a user equipment and affecting the state of the user equipment, at least one of said statistical data related to this state is updated incrementally.

[0049] In other words, for each new detected event E(N) affecting the considered state, with N representing an integer greater than 1, the relevant statistical data related to this state is taken from the statistical value G(N-1) taken from the detection of the preceding event affecting said state and of the single event E(N), denoted G(N).

[0050] In this way, the amount of data to be stored by the providing entity is further limited (the oldest events do not need to be kept, in addition to those belonging to the most recent past events and to the event E(N) fed back to the analyzing entity) and is thus constrained in terms of storage capacity of the providing entity. Moreover, the calculations performed by the latter to take the statistical data are simplified.

[0051] Thus, with reasonable efforts in terms of memory and computing of the providing entity, all the users managed by the providing entity can have access to information about the analyzing entity.

[0052] The providing entity can evaluate various types of statistical data.

[0053] Thus, for example, statistics can be evaluated over a sliding time window or a sliding number of events.

[0054] Statistics can also be evaluated over a periodic time window of period denoted T or over a time window of width equal to an integer multiple of the period T.

[0055] This allows to obtain accurate statistics evaluated over a fixed duration, which can be longer or shorter according to the envisaged variants. In particular, if a window of width equal to an integer multiple Q of the period T is envisaged for evaluating the statistics, this allows to obtain statistics over a quite long duration period Q.T representative of the long-term behavior of the user. Such statistics can be derived from the statistics obtained over the periodic time windows of period T without having to store additional data than the statistics evaluated periodically over each period.

[0056] The providing entity can thus evaluate several items of statistics related to the state of the user according to one and / or another of these techniques. Of course, other techniques can be envisaged as variants.

[0057] In a particular embodiment, the statistics can be evaluated using an exponential moving average (or EMA).

[0058] This embodiment (EMA) allows for example to obtain statistics (for example, mean and variance) on the estimation of the interval between two events affecting the same state.

[0059] As mentioned above, according to the application, the providing entity provides the analyzing entity with the statistics evaluated on the raw data it collects for the considered user equipment and on the raw data representative of the recent past of the users in the network. However, it can be envisaged that the providing entity provides the analyzing entity with other information than the above-mentioned information.

[0060] Thus, in a particular embodiment, the providing method further comprises providing, for each type of event among said plurality of events, at least one activity indicator of this type of event to said analyzing entity.

[0061] Correspondingly, the obtaining method comprises obtaining, for each type of event among said plurality of events, at least one activity indicator of this type of event.

[0062] For a given type of event (e.g. presence in a cell, loss of communication, etc.), this activity indicator comprises for example an estimate of the duration of the interval between two such types of events detected in the network for the user equipment (mean and variance). Of course, other types of activity indicators can be envisaged, for example an estimate of the regularity of the type of event indicating whether the type of event is fairly regularly spaced or irregularly spaced.

[0063] The activity indicators advantageously allow the establishment in the network of an activity profile of the user (e.g. frequency of movement of the user, quality of the radio environment in the vicinity of the user, etc.) in order to better characterize the user. This allows for example to distinguish a behavior involving a continuous parking in one cell then in another cell from a behavior involving frequent back-and-forth between these two cells for the same cumulative duration.

[0064] This knowledge of the activity indicators for each type of event also offers the possibility of choosing more relevant calculation models for the statistical analyses and predictions made by the analysis entity for this user.

[0065] In a particular embodiment, the providing method according to the application comprises a step of representative filtering of the collected data according to the data of the usual state of the user equipment.

[0066] This filtering only allows to retain the most significant values of the user behavior, thereby limiting the storage capacity necessary to implement the application.

[0067] As an alternative embodiment, it can be envisaged in other ways to retain the least significant values of the usual state of the user equipment and to eliminate some values more consistent with this usual state which are considered redundant with other data already fed back or do not provide additional information on the fed back statistical data.

[0068] Different criteria can be applied to filter the collected data. Thus, for example, the data can be filtered according to the time of collection of the data (e.g. deletion of the oldest data), the number of events associated with the data, the total duration of the events, etc.

[0069] In a particular embodiment, the providing step is determined by an explicit notification request from the analysis entity.

[0070] This explicit notification request can for example take the form of a subscription or membership of the analysis entity to the providing entity in order to be informed of the information items collected by the providing entity relating to a determined user or similarly determined user equipment. This allows to automatically inform the analysis entity.

[0071] As an alternative embodiment, it can involve an occasional or regular request (e.g. periodically) of the analysis entity.

[0072] In a particular embodiment, the providing step of the acquisition method according to the application is implemented periodically and / or in response to a request from the analysis entity.

[0073] As an alternative embodiment, the providing step can be triggered by detecting a particular criterion (e.g. exceeding a threshold).

[0074] In a particular embodiment of the application, the providing method and the acquisition method are implemented by a computer.

[0075] The application also relates to a computer program on a storage medium, in which the program is able to be implemented in a computer or more generally in a data providing entity according to the application, and comprises instructions adapted to implement the providing method as described above.

[0076] The application also relates to a computer program on a storage medium, in which the program is able to be implemented in a computer or more generally in an analysis entity according to the application, and comprises instructions adapted to implement the acquisition method as described above.

[0077] Each of these programs can use any programming language, and can be in the form of source code, object code, or intermediate code between source code and object code, such as in a partially compiled format, or in any other desired format.

[0078] The application also relates to a computer-readable information medium or storage medium comprising the instructions of a computer program as mentioned above.

[0079] This information or storage medium can be any entity or device capable of storing the program. For example, this medium can comprise a storage means, such as a ROM, for example a CD-ROM or a microelectronic circuit ROM, or even a magnetic recording means, for example a hard disk or a flash memory.

[0080] Furthermore, this information or storage medium can be a transmissible medium, such as an electrical or optical signal, which can be routed via a cable or an optical cable, by radio or by other means. The program according to the application can be particularly downloaded by means of a network of Internet type.

[0081] The program according to the application can be particularly downloaded via a network of Internet type.

[0082] Alternatively, this information or storage medium can be an integrated circuit incorporating the program, this circuit being adapted to execute or to be used in the execution of the providing method or acquisition method according to the application.

[0083] According to another aspect, the application relates to a telecommunication system comprising at least one data providing entity and an analysis entity according to the application.

[0084] The system according to the application benefits from the same advantages as the providing entity and the analyzing entity according to the application mentioned above.

[0085] In a particular embodiment, the data providing entity is an entity of a network managing a plurality of user equipment of the network, for example a NF entity hosting a network function.

[0086] In another embodiment, the data providing entity is the user equipment under consideration, i.e. for example a terminal of a user.

[0087] In other embodiments, it can also be envisaged that the providing method and the acquisition method according to the application, the providing entity and the analyzing entity and the system according to the application have a combination of all or some of the above-mentioned features. BRIEF DESCRIPTION OF DRAWINGS

[0088] Further features and advantages of the application will become apparent from the following description, given with reference to the attached drawings, which illustrate a non-limiting embodiment. In the drawings:

[0089] [ Figure 1 ] Figure 1 a telecommunication system according to the application in a particular embodiment is shown in its environment;

[0090] [ Figure 2 ] Figure 2 a hardware architecture of the providing entity and of the analyzing entity of the system of Figure 1 is schematically shown;

[0091] [ Figure 3 ] Figure 3 the main steps of the acquisition method implemented by the analyzing entity (DAF entity) of the system of Figure 1 are illustrated using a flowchart;

[0092] [ Figure 4 ] Figure 4 the main steps of the providing method implemented by the data providing entity (NF entity) of the system of Figure 1 are illustrated using a flowchart;

[0093] [Figure 5] Figure 5A and Figure 5B various ways of evaluating the statistical data are illustrated (periodic window and sliding window, respectively). DETAILED DESCRIPTION

[0094] Figure 1A telecommunication system 1 according to the present application in its environment in a particular embodiment is shown. The system 1 provides the possibility of collecting information about various events detected in a telecommunication network NW and related to users or groups of users of the network (more specifically, to the respective user equipment (UE) of the users), reducing the complexity in order to perform statistical analysis and / or predictions to optimize the operation of the network NW.

[0095] The term "event related to or about a user equipment" is understood herein to mean an event detected in the network that causes the user equipment to transition between two states (for example, the user equipment moves from one cell to another, the user equipment transitions from a "not registered" state to a "registered" state, etc.). The nature of the event considered is not limited, nor is the nature of the state of the user equipment considered. As mentioned above, within the meaning of the present application, the state of a user equipment can equally represent a general state of that user equipment (for example "registered", "connected", "not connected") as well as a more specific state such as "present in cell A", "in communication", etc.

[0096] In the example envisaged in Figure 1 , the telecommunication network NW is a 5G network defined in the 3GPP standards (except for the functions specific to the present application). However, this assumption is not limiting in itself and the present application is applicable to other types of networks (6G, proprietary telecommunication networks, etc.).

[0097] According to the present application, the system 1 comprises at least one data providing entity according to the present application. In the embodiments envisaged herein, this providing entity is an NF entity (or even an NF (Network Function) function) that manages a plurality of user equipment of the network and hosts at least one network function.

[0098] In a manner known per se, the NF entity is a functional block (software or physical) characterized by a behavior (in order to provide a service associated with said network function) and a defined external interface. The network function hosted by the NF entity is not limited: management of access to the network, management of mobility of users (i.e. user equipment), management of communication sessions made by users via their user equipment, storage of user profiles, gateway between networks, etc. The NF entity can be located in the control plane as in the user plane. By way of example, such NF entity is for example an AMF function or a session management function (SMF) for managing the mobility and access of users, etc.

[0099] Each NF entity of the system 1 is configured herein to collect "raw" data (i.e. facts or information about facts that occur in the network) about a determined set of events it discloses and related to the user equipment it manages, these events being related to the network function(s) it provides. The type of events disclosed by a NF entity has a fixed number, typically there are about ten different types of events for a 5G 3GPP network. Thus, for example, an AMF entity is able to collect data about mobility events related to the user equipment it manages, like user equipment moving from one cell to another, etc. Moreover, according to the present application and as further detailed below, each NF entity of the system 1 is configured to evaluate statistical data based on the raw data it collects. The events disclosed by a NF entity are defined for each NF entity and are known by other entities of the network. For example, for an AMF entity, these events numbered 16 are described in section 6.2.6.3.3 of the 3GPP TS 29.518 document v16.4.0 entitled "Technical Specification Group Core Network and Terminals; 5G System; Access and Mobility Management Services; Stage 3 (Release 16)" published in June 2020. For a SMF entity, there are 10 events among these events and they are described in section 5.6.3.3 of the 3GPP TS 29.508 document v16.4.0 entitled "Technical Specification Group Core Network and Terminals; 5G System; Session Management Event Exposure Service; Stage 3 (Release 16)" published in June 2020.

[0100] According to the application, the system 1 also comprises an entity called DAF analysis entity. In the embodiments described herein, this analysis entity is a centralized NWDAF (Network Data Analytics Function) entity in charge of data analytics of the network NW and configured to perform various statistical analyses and / or predictions from a set of information items collected from one or more NF entities. These statistical analyses and / or predictions can in particular be performed by the NWDAF entity upon request of a client NF entity, which can be the same or different from the NF entities collecting raw data about events related to users of the network NW. For the purpose of the statistical analyses and predictions that must be performed, the NWDAF entity can also collect information items from entities other than NF entities, for example network management entities, also called OAM (Operation, Administration and Maintenance) entities, or Application Functions (AFs) or even network user equipment.

[0101] For the purpose of collecting these information items, for a specific user or group of users and for all or some types of events disclosed by the NF entities, the NWDAF entity can subscribe to be informed by the NF entities of information items related to these types of events when these types of events are detected by the NF entities in the network and affect the status of the equipment (UE) of this user. For example, this notification can be performed periodically or upon detection of any other criterion (for example, a parameter observed by the NF entity exceeding a given threshold). In alternative embodiments, the NWDAF entity can also actively, periodically or at determined moments request the NF entities to provide it with information items about the types of events related to the user or determined group of users disclosed by this NF entity. In the embodiments described herein, this subscription benefits the NWDAF entity with the information items collected according to the application after the occurrence of an event of the type of events specified in the subscription of the NWDAF entity and having affected the status of the user equipment (UE) of a given user. In the embodiments described herein, as described in further detail below, these information items include statistical data gathered by the NF entity about various states of the user equipment detected during these events, raw data collected related to the integer X last events detected for the user (X being able to be defined by the NWDAF entity during its subscription or determined by the NF2 entity or even a default setting), and for each type of event specified in the subscription of the NWDAF entity, an activity indicator of this type of event of the user equipment. It should be noted that as an alternative embodiment, the NWDAF entity can request to receive only a part of these information items (for example, only the most recent statistical data and events).

[0102] The NF entities and the DAF entity of the system 1 are herein respectively configured to implement the method for providing data and the method for acquiring data according to the application. To this end, in the embodiments described herein, these entities rely on the same data model as described above in relation to the method for providing data and the method for acquiring data according to the application. Figure 2Hardware architecture of a computer, schematically shown. This hardware architecture can be that of the entity under consideration, or that of the physical device carrying it used by the entity when it is implemented in the form of a software function.

[0103] This architecture notably comprises a processor 2, a random access memory 3, a read-only memory 4, a non-volatile memory 5 and communication means 6, notably comprising various physical and protocol interfaces allowing the NF entity and the DAF entity to communicate with each other and with other entities of the network. Such interfaces are described, for example, in the 3GPP TS 29.500 document v16.4.0 entitled “Technical Specification Group Core Network and Terminals; 5G System; Technical Realization of Service Based Architecture; Stage 3 (Release 16)” published in June 2020, and more particularly for the AMF entity and the SMF entity in the previously cited 3GPP TS 29.518 and TS 29.508 documents.

[0104] The read-only memory 4 constitutes a storage medium according to the application, which can be read by the processor 2 and comprises the computer programs according to the application, i.e. the PROG-NF program for the NF entity and the PROG-DAF program for the DAF entity.

[0105] The PROG-NF program defines the functional modules of the NF entity according to the application, which are based on or control the hardware elements 2 to 6 described above. In this case, these modules are activated for at least one user equipment (UE) of a user managed by the NF entity and of the determined type of event disclosed by it, and in the embodiment described herein, as illustrated, comprise: Figure 1

[0106] - a collection module 7 configured to collect data (referred to herein as “raw” data) of a plurality of events related to this user equipment detected in the network NW (the device detecting these events is not limited). These data are related to the state of the user equipment affected by the detected events. For example, these data relate to the features that the NF entity must use to disclose the events detected by it, but other data can also be collected;

[0107] ​- a statistical data evaluation module 8 (module for obtaining statistical data within the meaning of the application), configured to evaluate one or more statistical data by aggregating the raw data related to at least one state of the user equipment collected by the collection module 7, for this state;

[0108] - a module 9 for estimating statistical data, called profile statistical data and outlining events of the user and the network, module 9 being in this case configured to evaluate, for each type of event disclosed by the NF entity and for the user equipment considered, an activity indicator of this type of event. This activity indicator of a given type of event is for example an estimated value of the duration of the interval between two events of this type for the user equipment (for example, the mean and the variance), or an estimated value reflecting the regularity of the occurrence of this type of event, whether this type of event is fairly regularly spaced or not for the user equipment; and

[0109] - a transmission module 10 configured to provide, via the interface provided for this purpose, the NWDAF entity with different items of information about the user equipment, namely one or more statistical data evaluated by the module 8 for evaluating statistical data about the states of the user equipment, raw data about the last X events detected in the network for the user equipment and affecting these states, collected by the collection module 7 (where X represents an integer greater than or equal to 1), and the activity indicator of each type of event about the user equipment estimated by the profile statistical data estimation module 9, disclosed by the NF entity.

[0110] The functions provided by these modules 7 to 10 are further described in detail below with reference to the steps of the provision method according to the application.

[0111] Similarly, the PROG-DAF program defines the functional modules of the NWDAF entity according to the application, which are based on or control the hardware elements 2 to 6 described above. These modules are activated for at least one user equipment and, in the embodiment described herein, as shown in Figure 1 illustrated, comprise:

[0112] - a transmission / reception module 11 configured to obtain, via the interface provided for this purpose, from each NF entity requested by the NWDAF entity, different items of information about the user equipment, and more particularly the information items described above provided by the transmission module 10 of the NF entity; and

[0113] - an analysis and prediction module 12 configured to perform statistical analyses and / or predictions based on the information items acquired by the transmission / reception module 11 for one or more users. As mentioned above, one or more customer NF entities of the NW network can request the statistical analyses and predictions made, and the nature of these statistical analyses and these predictions depends on the NF entity that initiates the request. It can in particular concern global predictions (related to the network, a region, etc., for example, the load rate of network resources, the average quality of service offered, the amount of connected user equipment or active sessions, etc.) or individual predictions (related to a specific user or group of users, for example, "the user equipment (UE) of user U1 will reach cell A in 10 minutes" or "80% of the user equipment (UE) of the group G of users will reach cell B in 15 minutes", etc.). Such predictions can be performed using models and prediction algorithms known to the person skilled in the art and not described herein, for example machine learning techniques, and more particularly for predictions related to the location of user equipment, i.e. LSTM ("Long Short Term Memory") techniques. It should be noted that the analysis and prediction module 12 can also be configured to perform a targeted monitoring of a specific user over a short time window, and for this purpose is able to activate the transmission / reception module 11 to request information items from the NF entities of the system 1 regarding the equipment (UE) of this specific user for a given period of time.

[0114] The functions provided by the modules 11 and 12 of the NWDAF entity are described in further detail below with reference to the steps of the acquisition method according to the application.

[0115] The main steps of the acquisition and provision methods according to the application implemented by the NWDAF entity of the system 1 and each NF entity of the system 1 requested by the NWDAF entity in a particular embodiment will now be described with reference to Figure 3 and Figure 4 respectively.

[0116] With reference to Figure 3 , it is previously assumed that in this case a NF entity of the network NW (for example, the customer entity NF1 identified in Figure 1 requests the NWDAF entity to perform a statistical analysis or a given prediction (step E10). By way of illustration, it is assumed in this case that the prediction requested by the entity NF1 aims to determine the next three cells that the user equipment (UE) of the user U, referred to throughout the remainder of this description as UE(U) and equally denoted as user equipment UE(U), or even as equipment UE(U) of the user U, or even simply as equipment UE(U), will visit. As mentioned above, the nature of the equipment UE(U) of the user U is not limited; for example, it is a terminal in this case.

[0117] Of course, this example is provided by way of illustration only and other predictions or statistical analyses can be envisaged in relation to a single user, multiple users or even a determined group of users and other user equipment (e.g. CPE, PABX, sensors, etc.).

[0118] Following this request, in the embodiments described herein, the NWDAF entity subscribes to the relevant NF entities of the system 1 (i.e. those that can provide information to it for performing the statistical analysis or the prediction requested by the customer entity NF1) to notifications about the device UE(U) of the user U and all or some of the events disclosed by these entities, i.e. the information items about the device UE(U) of the user U collected by these entities during these events (step E20). This subscription constitutes an explicit request for notifications from the NWDAF entity. It is assumed herein that the NWDAF entity subscribes to periodic notifications (period Tnotif) of the information items disclosed by the NF entities.

[0119] The NWDAF entity selects the events to which the device UE(U) of the user U is subscribed according to the statistical analysis or the prediction it has to perform.

[0120] The NWDAF entity can also record, independently of this subscription and asynchronously with the NF entities and on its own initiative, via a request provided for this purpose, the information items collected by the NF entities during events detected by the NF entities and related to the device UE(U) of the user U, for example, as mentioned above, in order to perform a targeted monitoring of the user U (i.e. of the user device UE(U)) over a short period of time or to improve its prediction model about the user U.

[0121] In the example envisaged herein, it is assumed for simplicity that the NWDAF entity sends a subscription request to a single NF entity (i.e. the NF2 entity identified in the middle of the figure) and subscribes to notifications of all the information items collected by the NF2 entity during events corresponding to all the types of events E1,..., EZ disclosed by the NF entity, where Z denotes an integer greater than or equal to 1 (statistics of the state of the device UE(U), the last X events, where X is an integer greater than or equal to 1, each event being defined by a certain number of characteristics depending on the type of event and an activity indicator for each type of event E1,..., EZ). Figure 1

[0122] As an alternative embodiment, it can subscribe to this type of notification from multiple NF entities and select only a subset of the events disclosed by each NF entity. It should be noted that the operating mode described below for the NF2 entity applies to any NF entity of the system 1 according to the application.

[0123] ​By way of illustration, when the NF2 entity is an AMF function as mentioned above, the AMF function can expose 16 types of events on its API Namf_EventExposure, as indicated in the previously cited 3GPP TS 29.518 document. These events include, for example: the location of the considered user equipment (“Location Report”), the presence of the user equipment in a zone of interest (“AOI Presence Report”), the time zone of the user equipment (“Time Zone Report”), the type of access network of the user equipment (“Access Type Report”), the registration status of the user equipment (“Registration Status Report”), the connection status of the user equipment (“Connection Status Report”), the accessibility status of the user equipment (“Reachability Report”), the communication failure of the user equipment (“Communication Failure Report”), etc.

[0124] As mentioned above, each event is modeled by a plurality of characteristics. The data collected by the NF entity (and by the NF2 entity) is partly conditioned by these characteristics. In the previous illustrative example of an AMF function, the “Location Report” event is modeled, for example, by the identifier of the user equipment and the location of the user equipment (in the form of a TAI (Tracking Area Identity) identifier, a cell identifier (or Cell ID), a wireline identifier (or Global Line ID), etc.). The “AOI Presence Report” event is modeled by the identifier of the user equipment, the identifier of the zone and the presence status (“in”, “not in” or “unknown”).

[0125] With reference to Figure 4 , the NF2 entity records a subscription of the NWDAF entity related to these types of events E1,..., EZ and the associated notification period Tnotif for the device UE of the user U (step F10).

[0126] While the NF2 entity receives the subscription, the entity continuously maintains a context CNT for each user equipment of the network NW that it needs to manage (i.e. that is involved in its use of the network NW), and therefore in particular for the device UE of the user U, in which it records the items of information related to the types of events E1,..., EZ detected in the network for the user equipment (UE) of these users. The context CNT is stored, for example, in the non-volatile memory 5 of the NF2 entity. Throughout the remainder of this description, the context of the device UE of the user U, denoted CNT(U) (also referred to hereinafter for simplicity as the context of the user U), is of particular interest, but all the contexts of the users maintained by the NF2 entity are thereby managed in the same way as the context CNT(U).

[0127] More specifically, in this case, the context CNT(U) of user U includes a table TAB(U) that lists the states of the device UE(U) of user U affected by events detected in the network for that device. Each different state presented by the device UE(U) during such an event is stored in table TAB(U) along with the characteristics defining that state. As mentioned above, table TAB(U) can list general states of the device UE(U), such as "connected" or "registered," and can also list more specific states of the device UE(U), such as "present in cell A," "present in cell B," etc. It should be noted that an event can affect several states of the device UE(U) in table TAB(U).

[0128] For each state of the device UE (U) of user U, table TAB (U) also includes statistics related to that state evaluated by the NF2 entity, as described in further detail below. As indicated above, it should be noted that, within the meaning of this invention, the statistics evaluated by the NF2 entity are obtained by aggregating (i.e., clustering) raw data related to the state of the device UE (U) during the detected event using elementary mathematical functions (e.g., averaging functions, variance functions, etc.), which is collected by the NF2 entity up to a given moment (e.g., the moment when statistics are requested or when statistics are evaluated): therefore, the statistics are obtained only from past raw data collected by the NF2 entity (as opposed to predictions that provide future indications). To some extent, it involves representing a compressed version of past raw data, which the NF2 entity can obtain very easily. Such statistics are typically the average duration and variance of the state of the device UE (U), the average interval and variance between two occurrences of that state, the number of occurrences of that state, etc. The ways in which these statistics can be evaluated are described in further detail below.

[0129] Furthermore, this paper also assumes that for each state of the device UE (U) of user U, in other words, for each entry of table TAB (U), the table includes the start and end timestamps of the last occurrence of that state.

[0130] In the embodiments described herein, in addition to table TAB(U), the context CNT(U) of user U also includes:

[0131] - Profile statistics related to each possible type of event for the device UE (U). In this case, these profile statistics consist of activity indicators for the possible types of events for the device UE (U). These activity indicators are also described in further detail below; and

[0132] - a table PAST(U) intended to include data collected by the NF2 entity regarding the last X events related to the device UE(U) impacting the state of this user equipment. In this case, X denotes an integer that can be chosen by default, fixed or by the NF2 entity or even by the NWDAF entity. This number X can be chosen so or derived from a fixed maximum time window so as to define the most recent past of events to be considered.

[0133] The NF2 entity then provides each user context CNT it holds as follows.

[0134] When the NF2 entity detects a new event of a user equipment (YES response in test step F20), for example for the device UE(U) of the user U, the NF2 entity processes the event in a manner known per se (in the example of a network NW according to the 3GPP standard envisaged herein, as defined by this standard) (step F30), then updates the context CNT(U) of the user U in its non-volatile memory 5 (step F40).

[0135] To perform this update, the NF2 entity collects, via its collection module 7, data characterizing one or more states of the device UE(U) impacted by the detected event (for example, for a change of location, data related to its new location, in particular the identifier of the new cell, the timestamp of the end of presence in the previous location, etc.), and stores these data in the table PAST(U) (step F42). If the table PAST(U) already includes X events, the oldest event is deleted so as to be able to store the data related to the newly detected event. It should be noted that the collection module 7 can determine these data directly or receive them from other entities of the network.

[0136] The data related to the state of the device UE(U) impacted by the newly detected event are recorded by the collection module 7 in an existing entry of the table TAB(U) of the context CNT(U) of the user U corresponding to the same state of the user equipment UE(U) (for example, presence in the same cell), or, in the absence of an existing entry corresponding to the same state (for example, new cell, new communication, etc.), cause the collection module 7 to create a new entry in the table TAB(U) (step F44).

[0137] As an alternative embodiment, the updating of the context CNT(U) with the data related to the detected event can be performed by a module of the NF2 entity other than the data collection module 7.

[0138] Then, the NF2 entity evaluates the statistics associated with each entry just created or input into the table TAB(U) (step F46). In the embodiments described herein, in order to preserve the storage resources of the NF2 entity, in particular the space occupied by the user context CNT in the non-volatile memory 5, the statistics are in fact updated incrementally. In other words, each time (indicated by the integer n) the statistic G of the table TAB(U) is evaluated for the state denoted ST of the user equipment under consideration, its value G(n) is calculated by considering only the value G(n-1) of the statistic acquired during the previous evaluation of the quantity G stored in the table TAB(U) and the current value of the state denoted ST(n) associated with the newly detected event. In this way, it is sufficient to store in the memory only the evaluated statistics and the current number of considered events used to evaluate these statistics.

[0139] Various types of statistics can be evaluated by the evaluation module 8 of the NF2 entity and stored in the table TAB(U) of the user U. In the example envisaged herein, the following are considered:

[0140] - statistics STAT1 evaluated over a determined number of events Ne;

[0141] - statistics STAT2 evaluated over a periodic time window having a width and a period T (for example, T = 1 hour), as Figure 5A illustrated (in the figure, ST1, ST2,..., ST5 denote the states of the user equipment UE(U)); and

[0142] - statistics STAT3 evaluated over longer time windows of width Tlong equal to integer multiples Q1, Q2, etc. of the period T (for example, Tlong= 6 hours, 24 hours, etc.), as Figure 5A also illustrated.

[0143] Of course, this example is provided by way of illustration only and other configurations of statistics can be considered for each state of the table TAB(U) (for example, statistics for each type of statistics mentioned above, or statistics evaluated only on a determined number of events, or periodic statistics and statistics over time windows equal to multiples of the period, etc.).

[0144] In addition, other types of statistics can be considered within the scope of the present application, for example, statistics evaluated by means of a continuous moving window (i.e. sliding). As Figure 5B illustrated, the use of a continuous sliding window involves continuously sliding a calculation window of time width T mob over the relevant collected data to evaluate the statistics: thus, at an evaluation instant t, the statistics acquired are those relating to the events detected in the time interval comprised between the instant t-Tmob The data collected in the time range between t is related to. In order to be able to get exact statistical data with a continuous sliding window, in addition to the statistical data, it is necessary to store the data related to the state and the events detected on this time window and all the timestamps of the state (start and duration). The statistical data can be evaluated in real time even if the arrival of the data is asynchronous. This can represent a considerable complexity if one wants to maintain several tens of statistical data for each user. This technique is preferably used for fixed elements (for example, events), i.e. elements that are not likely to change during the sliding window.

[0145] As indicated above, in the embodiments described herein, the evaluation of the statistical data STAT1, STAT2, STAT3 is considered with an incremental approach.

[0146] More particularly, the evaluation module 8 uses in this case the exponential moving average or EMA technique to get the statistical data STAT1 and STAT3. In this case, for a given state of the device UE(U) of a user U, these statistical data are related to the duration of the occurrence of this state (mean and variance) and the interval between two occurrences of this state (mean and variance). The number of occurrences of a state simply increases with each new occurrence.

[0147] In a manner known per se, using the preceding notation, according to the EMA technique, when the statistical quantity G to be evaluated is the mean, its value G(n) is derived during a step indicated by the integer n according to the value calculated in the preceding step n-1 as follows:

[0148] G(n) = (1 - A).G(n-1) + A.E(n),

[0149] where A represents a smoothing constant ranging between 0 and 1. To estimate the statistical quantity G(n), the value of the smoothing constant A determines whether to give significant importance to the past. For example, one can choose A = 2 / (1 + P), where P is the number of moving average samples to average. Thus, for example, A = 0.25 corresponds to a moving average estimated for 9 samples.

[0150] In the embodiments described herein, for the statistical data STAT1, P represents the number of events Ne to average in order to get these statistical data. For the statistical data STAT3, P represents the multiple Q1, Q2, etc. of the period T to consider when evaluating these statistical data based on the statistical data STAT2.

[0151] By means of this EMA technique, the evaluation module 8 of the NF2 entity can thus easily estimate the average duration spent by the device UE(U) in the considered state or the average elapsed time between two occurrences of this state.

[0152] The evaluation module 8 of the entity 2 can also use the EMA technique to estimate the variance. Indeed, by definition, the variance corresponds to the average of the square of the deviations with respect to the average. It is thus possible to apply an exponential moving average technique that takes into account the average of the square of the deviations of the data with respect to the estimated value of the average.

[0153] It should be noted that the EMA technique does not in itself provide information on whether the statistical data thus evaluated correspond to the same width of time window for different users. The time window considered actually depends on the activity rate of the users, which can vary from one user to another: depending on the level of activity of the users, these users can generate more or less states within the same period of time. Considering a periodic time window is preferable if the statistical data acquired are intended to be connected in real time. The use of the EMA technique can also be advantageous for data generated periodically, or even if, as in the case herein, the average value sought is evaluated on a determined number of data without linking this average value to a time concept.

[0154] As mentioned above, in order to acquire the statistical data STAT2, the evaluation module 8 considers a periodic time window of width and period T. This technique involves moving the calculation window in successive jumps of duration T. It thus acquires statistical data evaluated over several successive ranges of width T. Thus, at an evaluation instant t = kT, the statistical data acquired relate to the collected data related to the states detected in the current time range [(k-1)T, kT], with k representing an integer.

[0155] The statistical data STAT2 evaluated over a periodic time window of width T advantageously allow to acquire periodic statistical data on the elementary data, which are not and do not require data storage in addition to the current number of events and timestamps of the first and last states detected in the window. In this case, for each state of the device of each user managed by the NF2 entity, these statistical data STAT2 are reset to zero at the beginning of each period of width T and are evaluated over the considered period by performing an incremental accumulation operation on each new detected event of a state related to the user device of a given user.

[0156] In the embodiment envisaged herein, the incremental accumulation operation implemented is a simple recursion. More specifically, if the statistical quantity G(n) that the evaluation module 8 seeks to evaluate is the average value (for example, the average duration of the user device in a given state or the average interval between two occurrences of this state) of N occurrences of a state since the beginning of a period of duration T, the evaluation module 8 acquires the value G(N) via the following simple recursion:

[0157] G(n) = (G(n - 1). (n - 1) + E(n)) / n

[0158] If the statistical quantity G(n) is a variance, then it is sufficient to consider the square of the average minus the average of the squares, where each average is calculated using the recursive formula provided above.

[0159] Of course, an incremental method other than the simple recursion can also be envisaged as an alternative embodiment.

[0160] As indicated above, the evaluation module 8 evaluates the statistical data STAT3 on time windows of width equal to a multiple integer Q1, Q2, etc. of the period T (e.g. Tlong= Q1.T, Q2.T, etc.).

[0161] To this end, at the end of each period of width T, it is only necessary to update the statistical data STAT3 on Tlong. In the embodiments described herein, each statistical data is updated by selecting a smoothing constant A as a function of the multiple Q1, Q2, etc., using the EMA technique described above with respect to the value of the statistical data STAT2, considering the selection of the time window Tlong on which the statistical data STAT3 is evaluated. For example, if Tlong= Q1.T, the evaluation module 8 selects A = 2 / (1 + Q1)) as a smoothing constant according to the formula of the prior art introduced above. This latter formula can be applied because the element considered in the EMA calculation (statistical data STAT2) is a periodic element of period T.

[0162] Of course, as an alternative embodiment, another incremental calculation technique can be used, for example the simple recursive formula introduced above for calculating the statistical data STAT2.

[0163] In other words, for example, if the initial statistical data STAT2 is evaluated via a periodic window technique over a period T of a few minutes, it is possible to periodically evaluate the statistical data STAT3 over an hour, over 24 hours, over the previous days, etc. with minimal calculation effort via an EMA technique such as a simple recursive technique or another incremental technique, without the need to store elements other than the statistical data themselves.

[0164] As mentioned above, in the embodiments described herein, the NF2 entity employs an incremental method to calculate the statistical data STAT1, STAT2, STAT3: this method allows to reduce the amount of data stored in memory for the implementation of the present application. However, in another embodiment, it is possible to envisage storing the data necessary for the evaluation of each of the considered time periods for evaluating the statistical data STAT1, STAT2 and STAT3, and evaluating the statistical data at the end of the time period based on the stored data.

[0165] In the embodiments described herein, when updating the context CNT(U), the NF2 entity also estimates, through its estimation module 9, statistics of the user U, called profile statistics, associated with the type of event detected, in addition to updating the table TAB(U) (step F48). In this case, this profile statistics consists of an indicator estimating the activity of this type of event of the device UE(U) over a long moving period, for example equal to several times the period T previously considered for the evaluation of the statistics STAT2 or corresponding to a determined number Ne of events. In the embodiments described herein, the activity indicator of a certain type of event comprises an estimate of the mean and variance of the time interval between two events of this type, which is in this case incrementally estimated by the EMA technique described above by the estimation module 9.

[0166] As an alternative embodiment, when updating the context CNT(U) of the user U, other activity indicators of a certain type of event can be estimated, for example an estimate of the regularity of the occurrence of this type of event of the device UE(U) of the user U.

[0167] The context CNT of the users of the network managed by the NF2 entity is updated upon the detection in the network of each new event related to the user devices UE of these users. The steps F20 to F40 just described are then repeated in order to process this new event.

[0168] It should be noted that, in the embodiments described herein, in order to limit the number of entries in the table TAB(U) of each user U (for obvious reasons of complexity and required storage space), the NF2 entity implements a mechanism for filtering the entries of the table TAB(U). Indeed, as previously indicated, during the step F42, when a new event related to the device UE(U) is detected, each state of the device UE(U) affected by this event is recorded in an existing entry or in a new entry in the table TAB(U). Thus, in the case where the states of the device UE(U) are very numerous, the table TAB(U) can have to include a large number of entries. In order to overcome this drawback, a maximum number K of entries in the table TAB(U) is set in order to keep a reasonable size of the table TAB(U) for each user U. The value chosen for K can vary from one NF entity to another (depending on the nature of the network functions hosted by the NF entity) and / or as a function of other parameters, for example the period of observation, the use implemented by the network, the environment in which the application is implemented, etc.

[0169] The NF2 entity then implements a mechanism for filtering the entries (in other words the data collected and stored in the table TAB(U)) as a function of the representativeness of this data of the states of the device UE(U) of the user U. More particularly:

[0170] - if an entry has to be added, the NF2 entity creates an entry in the temporary table TAB TEMP(U) via its collection module 7 (limited to an integer number M of entries, where M can be different from K), where the collected data and the timestamp information of the collected data allow the statistics to be initialized;

[0171] - if an entry is updated, the NF2 entity updates one of the already existing entries from the K entries of the table TAB(U) or the M temporary entries of the temporary table TAB TEMP(U) via its collection module 7. In this case, the entries of the temporary table TAB TEMP(U) are candidates to become permanent and to be stored in the table TAB(U) of the user U. If the table TAB(U) already comprises K entries, the entries of the table TAB(U) are deleted by the collection module 7 of the NF2 entity. To this end, the collection module 7 identifies the entries of the table TAB(U) that are less important with respect to a determined criterion, for example the time the device UE(U) remains in the state associated with this entry, the interval between two occurrences of the state, the timestamp of the last occurrence of the state, etc. Combinations of several criteria can also be envisaged: for example, the collection module 7 keeps the entries corresponding to the longest time spent in the state associated with these entries, then considers the entries with the smallest interval between two occurrences of the state associated therewith, then considers the entries associated with the state with the most recent timestamp. Of course, other criteria can be envisaged as alternative embodiments.

[0172] If the temporary table TAB TEMP(U) is saturated, the collection module 7 also filters the entries contained in this temporary table, for example based on a timestamp criterion of the logged data.

[0173] As an alternative embodiment, or in addition to the preceding filtering mechanisms, the collection module 7 can also filter a priori (i.e. delete) the data corresponding to a short duration of the device UE(U) remaining in a certain state, i.e. below a predetermined threshold duration, even before any insertion in the temporary table TAB(U).

[0174] The filtering mechanisms just described advantageously allow only the most important data (i.e. the most representative) of the usual states of the device UE(U) of the user U to be kept in the table TAB(U). In the embodiments described herein, this is supplemented by filtering (in other words, deleting) very old inactive user contexts kept by the NF2 entity, in order to keep only a limited number of active and inactive user contexts in its non-volatile memory 5.

[0175] Based on the timestamp of the last event detected by the user equipment for a given user, the NF2 entity can estimate the time during which the relevant user equipment has been inactive, if applicable, and delete the context associated with this user if this estimate exceeds a given maximum duration for keeping an inactive context. This filtering can be implemented periodically or when a given maximum number of inactive contexts is reached, stored in the non-volatile memory 5 of the NF2 entity.

[0176] Of course, the NF2 entity can implement other filtering mechanisms and other criteria in order to preserve its memory resources. For example, instead of filtering data that are the least representative of the usual state of the device UE(U), it can be decided in other ways to keep these data that can provide the NWDAF entity with context information of interest and to delete in other ways certain stored data that are more in line with the expected or usual state of the device UE(U), for example because these data are redundant with other data or according to statistical data evaluated by the NF2 entity (i.e. these data do not provide additional information).

[0177] As mentioned above, in the embodiments described herein, the NWDAF entity has subscribed to the NF2 entity in order to be notified periodically, for example during all time periods of duration Tnotif, of the items of information collected by the NF2 entity in relation with the events detected thereby for the device UE(U). Thus, if the time period Tnotif for notifying the NWDAF entity is reached (‘yes’ response in test step F50), the NF2 entity notifies the NWDAF entity, via its transmission module 10, of all the items of information it has collected for the device UE(U) of user U during the time period Tnotif (step F60).

[0178] More particularly, it provides the device UE(U) with the following items of information:

[0179] - the last update of the statistical data that have been evaluated for the state of the device UE(U) that has been modified during the time period Tnotif, in other words, in this case, the statistical data STAT1, STAT2, STAT3 evaluated for the state of the device UE(U) until the time period Tnotif has elapsed;

[0180] - the data collected by the collection module 7 of the NF2 entity regarding only the last X events detected for the device UE(U) and stored in the table PAST(U) (without providing other collected data than this “most recent”). In alternative embodiments, the NF2 entity only feeds back the data of the last X events collected during the time period Tnotif that have impacted the modified state; and

[0181] - profile statistics estimated for each type of event detected during the notification period Tnotif.

[0182] It should be noted that, as an alternative embodiment, the NF2 entity can inform the NWDAF entity of the above information items on the basis of criteria other than periodic criteria. Moreover, in alternative embodiments, the NWDAF entity can periodically or occasionally request the NF2 entity to inform it of the information items it collects during previous periods, instead of subscribing to the notification of these information items by the NF2 entity.

[0183] Reference is made to Figure 3 When receiving the information items informed by the NF2 entity via its reception module 11 (step E30), the NWDAF entity uses these information items (as well as the information items it has received in the past) to perform the statistical analysis or the prediction requested by the NF1 customer entity (step E40). To this end, this is done in a manner known per se by using its analysis and prediction module 12. To this end, by providing raw collected data related to the last X events detected for the device UE (U), the NWDAF entity advantageously has not only the statistics previously established by the NF2 entity, but also contextual information that will complement these statistics, where this contextual information can prove useful for making the statistical analysis and / or the prediction requested by the NF1 entity.

[0184] According to the application, the number of information items exchanged between the NF2 entity and the NWDAF entity, as well as the amount of information stored by the NF2 entity, is limited, without sacrificing the representativeness of these information items and their usefulness to the NWDAF entity.

[0185] In the remainder of the specification, illustrative examples of information items that can be collected / evaluated by an AMF-type NF entity and provided to the NWDAF entity are provided. For each user U it manages, the AMF entity can keep:

[0186] - statistics STAT2 of the state of the device UE (U) evaluated over a previous period of duration T. For example, the AMF entity counts the most frequent locations of the device UE (U) over a period of duration T = 1 hour;

[0187] - statistics STAT3 of the state of the device UE (U) evaluated over a previous long period, i.e. an average of the statistics STAT2 evaluated over a period of duration T. For example, the AMF entity can count the most frequent locations of the device UE (U) over the past 8 hours (Q1 = 8), 24 hours (Q2 = 24) or the previous week or before (Q3 = 168);

[0188] - a list PAST(U) of the last events detected by the device UE (U) over a maximum time window (e.g. X = 30 last events or events detected in the last 120 minutes) and the data collected during these events. It should be noted that the choice of the number of events X instead of a duration eliminates the case of a highly variable arrival rate of users; and

[0189] - the profile statistics of the average delay and variance between two events of the same type of the device UE (U) for all events that can be disclosed by the AMF entity for the user equipment UE (U) (e.g. loss of connectivity, accessibility of the user terminal, location reporting, etc.).

[0190] In the embodiment just described, mechanisms are provided for limiting the amount of information stored by the NF entities and the amount of information fed back to the NWDAF entity by the latter. In addition to these mechanisms, careful selection of the deployment parameters of the invention (size of the tables maintained by the NF entities, width of the time window considered for building the statistics, etc.) can allow to further limit the implementation cost of the invention and allow a more efficient implementation of the invention.

[0191] The deployment parameters can be sized in particular according to the average activity of the users in the network NW. The deployment parameters can also vary according to the nature of the network, in particular its intended application (e.g. flexible workshop, telephone network company, public network, gas meter network, vehicle network).

[0192] In the embodiments that have been described, it can be necessary to set the following parameters according to the implementation adopted for evaluating the statistics:

[0193] - the number of events Z observed and disclosed by each NF entity, for example, according to the document TS 29.518 (version 16.4.0), Z = 16 for the AMF entity;

[0194] - the number of entries K of the table TAB(U), for example, for the NF entity, there are 10 to 15 favorite locations. However, this number can vary according to the user U;

[0195] - the width T of the periodic window considered for evaluating the statistics STAT2, for example, for the public network NW, T = 30 minutes;

[0196] - the number N1 of long evaluation period statistics STAT3, for example, N1 = 3;

[0197] - multiples Q1, Q2,..., QN1 of a period T, which are used to define long periods considered for the calculation of the statistics STAT3, for example, for N1 = 3 and T = 30 minutes, Q1 = 4 (which yields a period of 2 hours), Q2 = 16 (which yields a period of 8 hours) and Q3 = 48 (which yields a period of 24 hours);

[0198] - the number Ne of considered events in the sliding window for the calculation of the profile statistics (activity indicator of each type of event), for example, Ne = 40;

[0199] - the number X of recent events stored in the table PAST(U) and in the memory, for example, X = 10.

[0200] The number of information items provided periodically by the NF entity to the NWDAF entity is provided in table 1 (for simplicity, the size of each information item provided is omitted in this case, where the size depends on the amount of data included in each element and on the size of these data). In this case, it is assumed that the NF entity provides the statistics STAT1, STAT2, STAT3 of each state of the device UE(U) of the user U, the last X events in the recent past, and the profile indicator of each type of event disclosed by the NF entity.

[0201] [table 1]

[0202]

[0203] Thus, in view of the above, in order to set the deployment parameters, a compromise can be sought, for example, between:

[0204] - a reasonable size of the information to be kept and stored on each NF entity;

[0205] - an excessive flow of notifications (in particular periodic notifications) when feeding back the statistics; and

[0206] - a good observability of the UE state per user.

[0207] By way of illustration, table 2 provides two examples of statistics used to set the parameters that can be adjusted in order to implement this compromise for the AMF entity:

[0208] [table 2]

[0209]

[0210]

[0211] Of course, these examples are provided by way of illustration only and other criteria for setting the deployment parameters of the application can be envisaged.

[0212] In embodiments, and in the illustrative examples described herein, the data providing entity is considered to be a NF entity hosting a network function, and the data analyzing entity is a centralized DAF entity of the network, NW, or a NWDAF entity. Nonetheless, the application is applicable to other cases, as mentioned above. Thus, for example, the data providing entity can be a device of a user (UE), such as a terminal of a user, and the data collected during an event affecting at least one state of this user device, as detected by this user device or the network, is the data of interest. The analyzing entity can be a NF entity or another entity of the network capable of analyzing the data provided to it and, for example, capable of performing a prediction or statistical analysis based on these data.

Claims

1. A method for providing data related to at least one user equipment to an entity of a network, called data analytics entity (DAF), the method comprising: - a step (F42) of collecting raw data related to at least one state of said user equipment, which is affected by a plurality of events detected in the network, each event of said plurality of events being an event detected in the network causing the user equipment to transition between two states; and - a step (F60) of providing to the data analytics entity (DAF), for at least one of said affected states of the user equipment: • at least one statistical data acquired for this state by aggregating the raw data related to this state collected during the collecting step; and • a portion of the raw data related to this state collected during the collecting step, this portion corresponding to an integer X greater than or equal to 1 of the most recent events of said plurality of events.

2. The providing method of claim 1, wherein, The collecting step (F42) and the providing step (F60) are implemented by an entity (NF2) managing the network of a plurality of user equipments.

3. The providing method of claim 1, wherein, The collecting step and the providing step are implemented by said user equipment.

4. The providing method of claim 1, further comprising providing (F60) to the analytics entity (DAF), for each type of event of said plurality of events, at least one activity indicator of this type of event.

5. The providing method of claim 4, wherein, One of said activity indicators of said type of event comprises an estimate of the duration of the interval between two such type of events detected for said user equipment.

6. The providing method of claim 1, wherein, At least one of said statistical data related to said state of said user equipment is incrementally updated (F44) for each detected event related to said user equipment and affecting one of said states.

7. The providing method of claim 1, wherein, At least one of said statistical data is evaluated over a sliding time window, or over a periodic time window having a period denoted T, or over a time window having a width equal to an integer multiple of said period T.

8. The providing method of claim 1, wherein, At least one of said statistical data and / or activity indicators is evaluated using an exponential moving average.

9. The providing method of claim 1, comprising a step of representative filtering of the collected data according to the data of the usual state of this user equipment.

10. The providing method according to any one of claims 1 to 9, wherein, The providing step is implemented periodically and / or in response to a request from the analytics entity.

11. A method of allowing an entity of a network (NW), called a data analytics entity (DAF), to acquire data related to at least one user equipment, said analytics entity (DAF) being able to communicate with at least one other entity configured to collect raw data related to at least one state impacted by a plurality of events in the network affecting said user equipment, wherein, Each event of said plurality of events being an event detected in the network causing the user equipment to transition between two states, the method comprises a step (E30) of receiving, for at least one of said affected states of the user equipment: - at least one statistical data acquired for this state by the other entity by aggregating the raw data related to this state collected by said other entity for said plurality of events; and - a part of the raw data collected by the other entity in relation with the state for the plurality of events, wherein said part of data corresponds to an integer number X greater than or equal to 1 of the most recent events among the plurality of events.

12. A computer program product (PROG-NF, PROG-DAF) comprising instructions for implementing the providing method according to any one of claims 1 to 10 or the acquiring method according to claim 11.

13. A computer-readable storage medium (4) comprising the instructions in the computer program product according to claim 12.

14. An entity (NF) comprising a module activated for at least one user equipment and comprising: - a collecting module (7) configured to collect, for a plurality of events in a network affecting at least one state of the user equipment, raw data in relation with the at least one state affected by the events, wherein each event of the plurality of events is an event detected in the network causing a transition of the user equipment between two states; - a module for acquiring statistics (8) configured to acquire, for at least one of the affected states of the user equipment, at least one statistical data by aggregating the raw data in relation with the state collected by the collecting module; and - a transmission module (10) configured to provide, for the at least one of the affected states of the user equipment, to an entity of the network, called data analysis entity, the at least one statistical data acquired by the acquiring module for the state and a part of the raw data in relation with the state collected by the collecting module, said part corresponding to an integer number X greater than or equal to 1 of the most recent events among the plurality of events.

15. An entity (DAF) of a network (NW), referred to as a data analytics function, the entity being capable of communicating with another entity (NF) configured to collect raw data related to at least one impacted status of at least one user equipment for a plurality of events impacting said at least one impacted status in the network, wherein, each event of the plurality of events being an event detected in the network causing a transition of the user equipment between two states, the analysis entity comprising a reception module (11) activated for the at least one user equipment and configured to acquire, for at least one of the affected states of the user equipment, from the other entity: - at least one statistical data acquired by the other entity for the affected state by aggregating the raw data in relation with the state collected by the other entity for the plurality of events; and - a part of the raw data collected by the other entity in relation with the state for the plurality of events, wherein said part of data corresponds to an integer number X greater than or equal to 1 of the most recent events among the plurality of events.

16. A telecommunication system (1) comprising at least one entity (NF) according to claim 14 and an analysis entity (DAF) according to claim 15.

17. The system (1) as claimed in claim 16, wherein the entity according to claim 14 is an entity (NF2) managing a network of a plurality of user equipment.

18. The system of claim 16, wherein, the entity according to claim 14 is the user equipment.

Images