WO2020069739A1

WO2020069739A1 - Network node and method performed therein for handling operations in a set of communication networks

Info

Publication number: WO2020069739A1
Application number: PCT/EP2018/076946
Authority: WO
Inventors: Maxim TESLENKO; Aneta VULGARAKIS FELJAN; Adam Bergkvist; Rafia Inam
Original assignee: Telefonaktiebolaget Lm Ericsson (Publ)
Priority date: 2018-10-04
Filing date: 2018-10-04
Publication date: 2020-04-09
Also published as: EP3861679A1

Abstract

Embodiments herein relate, in some examples, to a method performed by a network node (11) for handling operations in a set of communication networks (1), wherein the set of communication networks (1) at least comprises a first communication network (2) and a second communication network (3). The network node (11) is configured to obtain a computational model created using data associated with operations of the first communication network (2), wherein the computational model comprises a first set of input parameters and one or more outputs. The network node (11) is further configured to obtain data from the second communication network (3), wherein the data comprises data of a second set of input parameters, wherein the second set of input parameters has non empty intersection with the first set of input parameters. The network node (11) further is configured to obtain additional data from a distribution set of data, and to use, in the obtained computational model, the obtained data from the second communication network (3), and the obtained additional data from the distribution set. The network node (11) is further configured to send to the second communication network (3), output of the obtained computational model.

Description

NETWORK NODE AND METHOD PERFORMED THEREIN FOR HANDLING

OPERATIONS IN A SET OF COMMUNICATION NETWORKS

TECHNICAL FIELD

Embodiments herein relate to a network node and method performed therein for communication networks. Furthermore, a computer program product and a computer readable storage medium are also provided herein. In particular, embodiments herein relate to handling operations in a set of communication networks.

BACKGROUND

In a typical communication network, wireless devices, also known as wireless communication devices, mobile stations, stations (STA) and/or user equipments (UE), communicate via e.g. access points to servers or to one another. Typical challenges in communication network operations are to handle different types of incoming alarms, understand the alarm conditions and find a resolution to keep the communication network at the highest performance. Today, this part of the work is largely human effort driven, where a human employs rule-based solutions in resolving the issues related to alarms.

A multi-site wireless communication network is a very rich source of information in the form of logs, alarms and trouble tickets. However, it is very time consuming and expensive for the operator to process the data.

Insights or perceptions learned from fixing problems on one site may be reused on another site. The challenge here is that each site is configured differently based on geographical location, operator, country-based policies, technology domain, etc. The heterogeneity of data among sites makes it less than straight forward to reuse the insights from one site to another site, even within the same operator of the wireless

communication network.

Reusing heterogeneous data and insights from one site to another site is complex. The reuse of insights across different operators is even more challenging. In addition to the competitive aspect, operators do not want to disclose sensitive information about their network. And since currently there is no mechanism to differentiate insights from sensitive and non-sensitive part of data, the operators generally choose not to share anything.

But even this non-sharing policy is hard to guarantee in Managed Services Network Operation Centers, because when the same human employees are fixing similar problems across different operators they intentionally or unintentionally reuse their past experiences. In a way it is a beneficial feature of network operation center (NOC) operation since it makes resolution of problems better and faster. But when human employees are involved it is nearly impossible to draw firm lines on what insights people should reuse and what insights people should not reuse.

SUMMARY

An object of embodiments herein is to provide a mechanism for improving operations of a wireless communication network in an efficient manner.

According to an aspect the object is achieved by providing a method performed by a network node for handling operations in a set of communication networks. The set of communication networks at least comprises a first communication network and a second communication network. The network node obtains, e.g. trains, a computational model created using data associated with operations of the first communication network, wherein the computational model comprises a first set of input parameters and one or more outputs. The network node obtains, e.g. receives or samples, data from the second communication network, wherein the data comprises data of a second set of input parameters, wherein the second set of input parameters has non empty intersection with the first set of input parameters. The network node further obtains, e.g. receives or retrieves, additional data from a distribution set of data and then uses, in the obtained computational model, the obtained data from the second communication network, and the obtained additional data from the distribution set. The network node furthermore sends to the second communication network, output of the obtained computational model.

According to another aspect the object is achieved by providing a network node for handling operations in a set of communication networks, wherein the set of

communication networks at least comprises a first communication network and a second communication network. The network node is configured to obtain a computational model created using data associated with operations of the first communication network, wherein the computational model comprises a first set of input parameters and one or more outputs. The network node is further configured to obtain data from the second

communication network, wherein the data comprises data of a second set of input parameters, wherein the second set of input parameters has non empty intersection with the first set of input parameters, and to obtain additional data from a distribution set of data. The network node is in addition configured to use, in the obtained computational model, the obtained data from the second communication network, and the obtained additional data from the distribution set; and to send to the second communication network, output of the obtained computational model.

It is furthermore provided herein a computer program product comprising instructions, which, when executed on at least one processor, cause the at least one processor to carry out any of the methods above, as performed by the network node. It is additionally provided herein a computer-readable storage medium, having stored thereon a computer program product comprising instructions which, when executed on at least one processor, cause the at least one processor to carry out the method according to any of the methods above, as performed by the network node.

Embodiments herein use an artificial intelligence (Al) and thereby making sure that there is no un-authorized insight reuse in the network node. Embodiments herein use the data from the different communication networks and may use their profiles, and relations and correlation between profiles. Then based on e.g. policies that exist in the system, which network operators may have signed agreements to reuse or not to reuse insights on data between each-other, and the network node may apply rules according to the policy. Since embodiments herein provide an enhanced reuse of data and insights in the same operator and across different operators and that the network operations are not biased, as the decisions and insights may be reused according to the programmed policy of sharing of data, embodiments herein enable that operations of the wireless

communication network is improved in an efficient manner.

BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments will now be described in more detail in relation to the enclosed drawings, in which:

Fig. 1 is a schematic overview depicting a set of communication networks;

Fig. 2 is a flowchart depicting a method performed by a network node according to embodiments herein;

Fig. 3 is a combined flowchart and signaling scheme according to embodiments herein;

Fig. 4 is a combined flowchart and signaling scheme according to embodiments herein;

Fig. 5 is a schematic overview depicting a set of wireless communication networks according to embodiments herein; and

Fig. 6 is a block diagram depicting embodiments of a network node according to embodiments herein. DETAILED DESCRIPTION

Embodiments herein relate to communication networks in general. Fig. 1 is a schematic overview depicting a set of communication networks 1. The set of

communication networks 1 comprises at least a first communication network 2 and a second communication network 3. The set of communication networks 1 may be any kind of communication networks such as wireless communication networks, see Fig. 5. Each communication network may serve a number of wireless devices, such as user equipments 10. Embodiments herein relate to a network node 11 for handling operations, such as providing insights etc., in the set of communication networks 1.

The methods according to embodiments herein are performed by the network node 1 1. As an alternative, a Distributed Node (DN) and functionality, e.g. comprised in a cloud 19 as shown in Fig. 1 may be used for performing or partly performing the methods.

The method actions performed by the network node 1 1 for handling operations in the set of communication networks 1 , wherein the set of communication networks 1 at least comprises the first communication network 2 and the second communication network 3 according to embodiments herein will now be described with reference to a flowchart depicted in Fig. 2. The actions do not have to be taken in the order stated below, but may be taken in any suitable order. Actions performed in some embodiments are marked with dashed boxes.

Action 201. The network node 1 1 may obtain an indication indicating a sharing policy for sharing, with other communication networks, a subset of input parameters that is allowed to be used in a computational model e.g. policy for sharing insights from collected data. Input parameters may be e.g. temperature and humidity see example below where output is fan speed. The computational model is used to take a number of parameters into consideration to learn insights of a system in operation. Thus, policies are provided defining a subset of input parameters that is allowed to be used when the computational model is used. E.g. the network node 1 1 may develop a table of sharing policies defining allowable input parameters between operators. The computational model may be a neural network that is trained with weights and parameters.

Action 202. The network node 1 1 obtains the computational model created using data associated with operations of the first communication network 2, wherein the computational model comprises a first set of input parameters and one or more outputs e.g. fan speed as in the example below. E.g. the network node 1 1 may create, develop or receive the computational model. The computational model may be a trained model using a machine learning algorithm, e.g. a model with insights such as weights and/or parameters. Thus, the computational model may be developed by training the

computational model using data from the first communication network.

Action 203. The network node 1 1 obtains data from the second communication network 3, wherein the data comprises data of a second set of input parameters, wherein the second set of input parameters has non empty intersection with the first set of input parameters, i.e. the second set of input parameters overlaps at least a part of the first set of input parameters.

Action 204. The network node 1 1 obtains additional data from a distribution set of data, e.g. sampling data from distribution set of data. The second set of input parameters may differ at least in one input parameter, e.g. no temperature data is collected, compared to input parameters of the first set of input parameters and the obtained additional data from the distribution set of data may be data of the at least one input parameter. The network node 1 1 may take the sharing policy into account when obtaining the additional data. The obtained additional data may be data for one or more input parameters excluded to be shared according to the sharing policy or not available in the data of the second set of input parameters. The distribution set of data may be a histogram describing typical frequency distribution of data. E.g. natural distribution is one of the most common distributions in nature and may be represented by a Gaussian distribution. If e.g. temperature follows a normal distribution with mean 20°C means that a majority of outcomes of the temperature will be around 20°C and with a low probability that the temperature will be other than 20°C. Each communication network may create this distribution or the distributions may be shared. Sharing parameters of these

distributions among communication networks may be governed by polices, e.g. the sharing policies mentioned in action 201. The distribution may be obtained from a function, from a normal distribution set, from previous collected data or similar. The distribution may be a continuous function describing typical frequency distribution of data. The distribution set of data may be developed over time, i.e. the distribution is collected over a period of time and values are set as distribution values based on the previous collected data. The distribution set of data may be stored and maintained locally in a memory or retrieved from another network node or similar.

Action 205. The network node 1 1 uses, in the obtained computational model, the obtained data from the second communication network 3, and the obtained additional data from the distribution set. E.g. the network node 1 1 computes outputs from the computational model with trained insights such as weights or parameters using the data and the obtained additional data. If an input parameter is a variable i from the second communication network 3 and the input parameter is an input parameter not present among variables in the computational model e.g. wind speed, the data of the input parameter may be discarded. In addition, if the second set of input parameters lacks an input parameter of the computational model one may thus take X number of samples from the distribution set in e.g. the memory and use them in the computational model, wherein X depends on available computational resources and decision latency requirements,. E.g. weighted average may be computed when the distribution is used for the purpose of computing output of the computational model.

Action 206. The network node 1 1 send to the second communication network 3, output of the obtained computational model. E.g. alarms/warnings, resolutions, and/or actuation signals from the computational model based on the second set of data.

For example, a computational model that relates ambient temperature, T, and humidity, H, to an optimal speed of cooling fans in base station equipment. It is herein assumed for simplicity a linear model where Fan Speed(FS) is linearly proportional to temperature (T) and humidity (H), i.e. FS(T,H)=aT+3H

Assume that during operation of site 1 empirical data is collected on what is optimal fan speed for a given combination of temperature and humidity. So the data may look like:

Using regression analyses parameters a and b are estimated to be 2 and 0,1 respectively, so that the computational model become FS(T,H)=2T+0,1 H. So the model (the linear relationship) and its parameters are insights which were extracted from empirical data. The computational model has two input parameters T and H. This is an example of e.g. action 202.

If e.g. a policy of site 1 stated that it may share with site2 only insights of the computational model based on humidity, but not temperature, are used. Thus, only humidity from site 2 is used for evaluation of the computational model. When evaluating the computational model then the temperature is taken from a typical distribution of temperature in base station, e.g. in action 204. Assume that on average the temperature is 20 degrees in 80%, 19 degrees in 5% and 21 degrees in 15% of the time. Then fan speed for site 2 is computed, e.g. in action 205, as such:

Fan_speed_site2 (H)= 0,05^*FS(19,H)+0,8^* FS(20,H)+ 0,15 ^*FS(21 ,H)

For example for humidity H=40%:

Fan_speed_site2 (40)= 0,05^*FS(19,40)+0,8^* FS(20,40)+ 0,15 ^*FS(21 ,40)=

0,05^*42+0,8^*44+0,15 ^*46=44,2

Fig. 3 is a combined signalling scheme and flowchart according to some embodiments herein.

Action 301. Different operators of different communication networks provide a respective indication of sharing policies to the network node 1 1 .

Action 302. The network node 1 1 creates the computational model based on data from e.g. Operator 1 site 1.

Action 303. The second communication network 2, e.g. operator 2 site 2, provides data, such as logs, alarms and trouble tickets.

Action 304. The network node 1 1 obtains additional data from the set of distribution also referred to as distribution set. The additional set may be for an input parameter lacking the received data or for an data type not allowed to be shared according to the policies.

Action 305. The network node 1 1 may then compute outputs from the

computational model with the trained insights such as weights or parameters using the data and the obtained additional data.

Action 306. The outputs from the computational model are then shared back to the second communication network 2. Thus, according to embodiments herein, data from different sites may belong to different operators, with different policies and rules. Some embodiments herein only use data which is allowed according to the policies as explained with reference to Fig. 4.

Action 401. Each communication network transmits its initial policy configuration or subsequent update of configuration comprising e.g. an insight sharing policy table. This may be bootstrapped on a data or dedicated signalling. For each site (and all operators) the insight sharing policy table, policy table for short, is maintained. See tables 1 , 2, 3 and 4 as examples. Each column may represent data policies towards specified site and operator, and each row corresponds to a data stream generated at the site, i.e. data to be used as input to the computational model.

Example of policy tables for Operators A, B and C, and different sites of these operators, wherein operators A and B have two sites and operator C may have one or more sites.

Table 1 Sharing policy table for Operator A site 1

Table 2 Sharing policy table for Operator A site 2

Table 3 Sharing policy table for Operator B site 1

able 4 Sharing policy table for Operator C site 1 From Table 1 it should be noted that OpA site 1 generates data of different data types (Dtype), e.g. Dtypel , Dtype3, Dtype4, and policies allow reusing insights from all data types on the other OpA sites but not allowing reusing insights from Dtypel on all sites of OpB and not allowing reusing insights from Dtype4 on all sites from OpC.

From Table 2 it should be noted that OpA site 2 generates data (d) of Dtypel , Dtype2 which is not exactly the same as data generated at site 1.

From Table 1 and Table 2 it can be observed that even though OpB site 1 and OpA site 1 generate exactly the same data the policies define that one cannot reuse insights on all data in either direction.

In either case if there is a mismatch in generated data sets or due to policy restriction, the problem of reuse of insights with heterogeneous data is the same.

Effectively there may be two possible types of problems. Either insight data depends on variables which are not available in the input data or the input data has variables not present in the insight data. Embodiments herein show how to deal with both of the problems.

Action 402. The network node generates the computational model for e.g. the first operator, e.g. for each site X the network node builds a model function prediction e.g. X_(d1 ,d2,...,dx) from data generated at this site (d1 ,d2,...,dx). In general, Machine Learning (ML) algorithms are trained with past experiences in form of relations between instances of input data such as alarms, tickets, etc. and corresponding outputs such as resolutions or observed values. Using the training data the ML makes an extrapolation model that maps the data type corresponding to the training data to an output, i.e.

(d e Dtype) ® Output(d).

Generally, the data may be multidimensional because there may be multiple sources of information. For example, one data source or input parameter is alarms from radio units being e.g. Dtypel , another one is weather status being e.g. Dtype2, and so forth ...DtypeN. Thus, formally speaking the learned computational model may look like:

(dl e Dtypel, d2 e Dtype2, ... dN e DtypeN) ® Output(dl, d2, ... dN).

In order to utilize the ML model, the input data should be of exactly the same type as in the training set i.e. the same input parameters. For example, if training data relates radio link failure with humidity and temperature, then in order to make a prediction of failure, the input data should include both humidity and temperature as input parameters. In case when the training data and input data for ML comes from the same source (the same site), typically the data is of the same dimension and type, so there is no problem of data matching.

As depicted in the policy tables, when data is coming from different sources or there are some policy restrictions then not all the data is available or can be used in order to utilize the ML model. Below it is described how these types of issues are dealt with.

Problem definition: Here is an example of a situation which may occur if insights learned on the OpA site 2 are tried to be used on the OpC site:

Mode\ Output(Dtypel, DtypeZ)

Data: dl e Dtypel, d3 e Dtype 3

The task is to evaluate (dl e Dtypel, d3 e Dtype3) ® Prediction(dl, d3) based on available model Output(Dtypel, Dtype2).

The problem definition has both possible problems related to data mismatch, i.e. input parameters do not match. It is herein shown a computational model that depends on data not present in input data (Dtype2) and the other way around (d3 e Dtype3). Thus, the computational model is captured with insight such as rules or configuration parameters and/or weights.

Action 403. The second communication network transmits data for one of its sites to the network node 1 1 , e.g. Data_A(d1 ,d2,...,dn).

Action 404. The network node 1 1 stores data to use it later for model training Prediction A (d1 ,d2,...,dn).

Action 405. The network node 1 1 may then compute, based on data, prediction values. For example, based on Data_A(d1 ,d2,....,dn) the network node 1 1 may compute prediction values P_A=Prediction A(d1 ,d2,...dn) and P_X=Prediction_X(d1 ,d2,....,dn), where X are all sites which share insights with siteA and Prediction X is corresponding model of each site X. Note the model Prediction X might use different data types then we evaluate the model according to method described herein. The problem of having a variable, in the input data, which is not present among variables in the model, is dealt with by discarding this data of the input parameter not present in the computational model. In the example, since the model function prediction does not depend on Dtype3, data d3 is discarded.

According to the example data d1 is provided from the second communication network 3 but data d2 is not in the input data but the insights of the computational model depend on it. According to some embodiments the problem is dealt with in the following way:

Take X number of samples, wherein X depends on available computational resources and decision latency requirements, from the set of distribution e.g.

Dtype2distribution, i.e. S={s1 ,s2,... sX} where si is a sample of a random variable with Dtype2distribution distribution, this is an example of action 204.

Evaluate prediction using the formula:

Prediction vvhere si e S

Thus, embodiments herein effectively marginalize or average out unavailable data input by using a typical value distribution for that unavailable data.

For example, if temperature information is lacking, but the computational model depends on it, externally gathered information may define that on an average the temperature is 20 degrees in 80%, 19 degrees in 5% and 21 degrees in 15% of the time. Then the prediction may be a weighted average: Prediction(dl, d3) = 0,05 *

Output(dl,19) + 0,8 * Output(dl,20) + 0,15 * Output(dl,21). This may thus be used in the computational model such as the prediction model.

Action 406. The network node then transmits the output of the prediction model to the second communication network 3. For example, control data based on

Data_A(d1 ,d2,...,dn) P A and all P_X.

Embodiments herein relate to communication networks e.g. wireless

communication networks. Fig. 5 is a schematic overview depicting such an example wherein the set of communication networks is a set of wireless communication networks 100. The set of wireless communication networks 100 comprises one or more RANs e.g. a first RAN (RAN1 ), connected to one or more CNs. The set of wireless communication networks 100 may use one or a number of different technologies, such as Wi-Fi, Long Term Evolution (LTE), LTE-Advanced, New Radio (NR) of 5G, Wideband Code Division Multiple Access (WCDMA), Global System for Mobile communications/Enhanced Data rate for GSM Evolution (GSM/EDGE), Worldwide Interoperability for Microwave Access (WiMax), or Ultra Mobile Broadband (UMB), just to mention a few possible

implementations. Embodiments are thus applicable to 5G and also in further development of the existing communication systems such as e.g. 3G and LTE.

In the set of wireless communication networks 100, wireless devices e.g. wireless devices 10 such as mobile stations, non-access point (non-AP) STAs, STAs, user equipment and/or wireless terminals, are connected via the one or more RANs, to the one or more CNs. It should be understood by those skilled in the art that“wireless device” is a non-limiting term which means any terminal, wireless communication terminal,

communication equipment, Machine Type Communication (MTC) device, Device to Device (D2D) terminal, or user equipment e.g. smart phone, laptop, mobile phone, sensor, relay, mobile tablets or any device communicating within a cell or service area.

The set of wireless communication networks 100 comprises a first wireless communication network with a first radio network node 12. The first radio network node 12 is exemplified herein as a RAN node providing radio coverage over a geographical area, a first service area or first cell 13, of a radio access technology (RAT), such as NR, LTE, UMTS, Wi-Fi or similar. The first radio network node 12 may be a radio access network node such as radio network controller or an access point such as a wireless local area network (WLAN) access point or an Access Point Station (AP STA), an access controller, a base station, e.g. a radio base station such as a NodeB, an evolved Node B (eNB, eNodeB), a gNodeB, a base transceiver station, Access Point Base Station, base station router, a transmission arrangement of a radio base station, a stand-alone access point or any other network unit capable of serving a wireless device within the service area served by the first radio network node 12 depending e.g. on the radio access technology and terminology used and may be denoted as a serving node, serving radio network node or primary serving radio network node providing a serving cell for the wireless device 10.

The set of wireless communication networks 100 further comprises a second wireless communication network with a second radio network node 14, the radio network node 14 for short. The second radio network node 14 is exemplified herein as a RAN node providing radio coverage over a geographical area, a second service area or cell 15, of a radio access technology (RAT), such as NR, LTE, UMTS, Wi-Fi or similar.

The second radio network node 14 may be denoted as a second node, second serving radio network node providing a second serving cell for a wireless device. It should be noted that a service area may be denoted as cell, beam, and beam group or similar to define an area of radio coverage.

It should be understood that the first and second service area may also be operated by different operators of different wireless communication networks, thus a first operator collects data of the first wireless communication network and the second operator collects data of the second wireless communication network. Furthermore, each radio network node may comprise one or more sites e.g. three sites, each covering one service area. According to embodiments herein the network node 1 1 may collect data from e.g. the first radio network node 12 comprising alarms and incidents, and train the

computational model with these data of input parameters. The network node 1 1 may then receive data and a request from the second wireless communication network e.g. from the second radio network node 14. According to embodiments herein the network node 1 1 obtains e.g. internally from a memory, additional data from the distribution set of data. The additional data may be data associated with insights not allowed to be shared or be data not in the data from the second wireless communication network. The network node 1 1 then uses, in the obtained computational model, the obtained data from the second wireless communication network and the obtained additional data from the distribution set. The results, i.e. the output of the computational model is then transmitted to the second wireless communication network for processing the result or performing some actions based on the result.

Fig. 6 is a block diagram depicting the network node in two embodiments for handling, e.g. analyzing or providing insights, operations of the set of communication networks 1. The set of communication networks 1 comprises at least the first

communication network 2 and the second communication network 3. It is herein disclosed an artificial intelligence network node that is responsible for e.g. predicting data values, and sharing the data according to operator sharing insight or knowledge policies is a logical function. This may be running somewhere in a cloud and a NOC may be automatically manage the communication networks using data, insight and/or knowledge from this network node.

The network node 1 1 may comprise processing circuitry 601 e.g. one or more processors, configured to perform the methods herein.

The network node 1 1 may comprise an obtaining unit 602, e.g. a receiver, transceiver or retrieving module. The network node 1 1 , the processing circuitry 601 , and/or the obtaining unit 602 is configured to obtain the computational model created using data associated with operations of the first communication network 2, wherein the computational model comprises the first set of input parameters and the one or more outputs. The network node 1 1 , the processing circuitry 601 , and/or the obtaining unit 602 is configured to obtain the data from the second communication network 3, wherein the data comprises data of the second set of input parameters, wherein the second set of input parameters has non empty intersection with the first set of input parameters, i.e. the input parameters of the first and second set are at least in part overlapping. The network node 1 1 , the processing circuitry 601 , and/or the obtaining unit 602 is configured to obtain additional data from the distribution set of data. The distribution set of data may be a histogram describing typical frequency distribution of data. The distribution set of data may be a continuous function describing typical frequency distribution of data. The obtained additional data may be data for one or more input parameters excluded to be shared according to the sharing policy or not available in the data of second set of input parameters. The second set of input parameters may differ at least in one input parameter and the obtained additional data from the distribution set of data may be data of the at least one input parameter. The network node 1 1 , the processing circuitry 601 , and/or the obtaining unit 602 may further be configured to obtain the indication indicating the sharing policy for sharing, with other communication networks, a subset of input parameters that is allowed to be used in the computational model, and further configured to take the sharing policy into account when obtaining the additional data.

The network node 1 1 may comprise a using unit 603, e.g. a prediction module. The network node 1 1 , the processing circuitry 601 , and/or the using unit 603 is configured to use, in the obtained computational model, the obtained data from the second communication network 3, and the obtained additional data from the distribution set.

The network node 1 1 may comprise a sending unit 604, e.g. a transmitter, transceiver or providing module. The network node 1 1 , the processing circuitry 601 , and/or the sending unit 604 is configured to send to the second communication network 3, output of the obtained computational model.

The network node 1 1 further comprises a memory 605. The memory comprises one or more units to be used to store data on, such as models, input parameters, output parameters, insights, data, processes to process the data, set of distributions, applications to perform the methods disclosed herein when being executed, and similar.

The methods according to the embodiments described herein for the network node are respectively implemented by means of e.g. a computer program product 606 or a computer program, comprising instructions, i.e., software code portions, which, when executed on at least one processor, cause the at least one processor to carry out the actions described herein, as performed by the network node 1 1 . The computer program 606 may be stored on a computer-readable storage medium 607, e.g. a disc or similar. The computer-readable storage medium 607, having stored thereon the computer program product, may comprise the instructions which, when executed on at least one processor, cause the at least one processor to carry out the actions described herein, as performed by the network node 1 1. In some embodiments, the computer-readable storage medium may be a non-transitory computer-readable storage medium. The network node 1 1 may comprise a communication interface comprising a transceiver, a receiver, a transmitter, and/or one or more antennas.

As will be readily understood by those familiar with communications design, that functions means, units, or modules may be implemented using digital logic and/or one or more microcontrollers, microprocessors, or other digital hardware. In some embodiments, several or all of the various functions may be implemented together, such as in a single application-specific integrated circuit (ASIC), or in two or more separate devices with appropriate hardware and/or software interfaces between them. Several of the functions may be implemented on a processor shared with other functional components of a radio network node, for example.

Alternatively, several of the functional elements of the processing circuitry discussed may be provided through the use of dedicated hardware, while others are provided with hardware for executing software, in association with the appropriate software or firmware. Thus, the term“processor” or“controller” as used herein does not exclusively refer to hardware capable of executing software and may implicitly include, without limitation, digital signal processor (DSP) hardware, read-only memory (ROM) for storing software, random-access memory for storing software and/or program or application data, and non-volatile memory. Other hardware, conventional and/or custom, may also be included. Designers of radio network nodes will appreciate the cost, performance, and maintenance trade-offs inherent in these design choices.

In some embodiments a non-limiting term“wireless device” is used. The wireless device herein may be any type of UE capable of communicating with network node or another UE over radio signals. The wireless device may also be radio communication device, target device, device to device (D2D) UE, machine type UE or UE capable of machine to machine communication (M2M), a sensor equipped with UE, iPad, Tablet, mobile terminals, smart phone, laptop embedded equipped (LEE), laptop mounted equipment (LME), USB dongles, Customer Premises Equipment (CPE) etc.

Also in some embodiments generic terminology“network node”, is used. It may be any kind of network node which may comprise of a core network node, e.g., NOC node, Mobility Managing Entity (MME), Operation and Maintenance (O&M) node, Self- Organizing Network (SON) node, a coordinating node, controlling node, Minimizing Drive Test (MDT) node, etc.), or an external node (e.g., 3^rd party node, a node external to the current network), or even a radio network node such as base station, radio base station, base transceiver station, base station controller, network controller, evolved Node B (eNB), Node B, multi-RAT base station, Multi-cell/multicast Coordination Entity (MCE), relay node, access point, radio access point, Remote Radio Unit (RRU)

Remote Radio Head (RRH), etc.

The term“radio node” used herein may be used to denote the wireless device or the radio network node.

The term“signaling” used herein may comprise any of: high-layer signaling, e.g., via Radio Resource Control (RRC), lower-layer signaling, e.g., via a physical control channel or a broadcast channel, or a combination thereof. The signaling may be implicit or explicit. The signaling may further be unicast, multicast or broadcast. The signaling may also be directly to another node or via a third node.

The embodiments described herein may apply to any RAT or their evolution, e.g., LTE Frequency Duplex Division (FDD), LTE Time Duplex Division (TDD), LTE with frame structure 3 or unlicensed operation, UTRA, GSM, WiFi, short-range communication RAT, narrow band RAT, RAT for 5G, etc.

It will be appreciated that the foregoing description and the accompanying drawings represent non-limiting examples of the methods and apparatus taught herein. As such, the apparatus and techniques taught herein are not limited by the foregoing description and accompanying drawings. Instead, the embodiments herein are limited only by the following claims and their legal equivalents.

Claims

1. A method performed by a network node (1 1 ) for handling operations in a set of communication networks (1 ), wherein the set of communication networks (1 ) at least comprises a first communication network (2) and a second

communication network (3), the method comprising:

obtaining (202) a computational model created using data associated with operations of the first communication network, wherein the computational model comprises a first set of input parameters and one or more outputs; obtaining (203) data from the second communication network (3), wherein the data comprises data of a second set of input parameters, wherein the second set of input parameters has non empty intersection with the first set of input parameters;

obtaining (204) additional data from a distribution set of data;

using (205), in the obtained computational model, the obtained data from the second communication network (3), and the obtained additional data from the distribution set; and

- sending (206) to the second communication network (3), output of the

obtained computational model.

2. The method according to claim 1 , further comprising

obtaining (201 ) an indication indicating a sharing policy for sharing, with other communication networks, a subset of input parameters that is allowed to be used in a computational model, and wherein obtaining the additional data is taking the sharing policy into account.

3. The method according to the claim 2, wherein the obtained additional data are data for one or more input parameters excluded to be shared according to the sharing policy or not available in the data of the second set of input parameters.

4. The method according to any of the claims 1 -3, wherein the second set of

input parameters differs at least in one input parameter and the obtained additional data from the distribution set of data are data of the at least one input parameter.

5. The method according to any of the claims 1 -4, wherein the distribution set of data is a histogram describing typical frequency distribution of data.

6. The method according to any of the claims 1 -4, wherein the distribution set of data is a continuous function describing typical frequency distribution of data.

7. The method according to any of the claims 1 -6, wherein the computational model is a trained model using a machine learning algorithm.

8. The method according to any of the claims 1 -7, wherein the computational model is a neural network that is trained with weights and parameters.

9. A computer program product comprising instructions, which, when executed on at least one processor, cause the at least one processor to carry out a method according to any of the claims 1 -8, as performed by the network node.

10. A computer-readable storage medium, having stored thereon a computer program product comprising instructions which, when executed on at least one processor, cause the at least one processor to carry out a method according to any of the claims 1 -8, as performed by the network node.

1 1. A network node (1 1 ) for handling operations in a set of communication

networks (1 ), wherein the set of communication networks (1 ) at least comprises a first communication network (2) and a second communication network (3), wherein the network node (1 1 ) is configured to:

obtain a computational model created using data associated with operations of the first communication network (2), wherein the computational model comprises a first set of input parameters and one or more outputs; obtain data from the second communication network (3), wherein the data comprises data of a second set of input parameters, wherein the second set of input parameters has non empty intersection with the first set of input parameters;

obtain additional data from a distribution set of data;

use, in the obtained computational model, the obtained data from the second communication network (3), and the obtained additional data from the distribution set; and send to the second communication network (3), output of the obtained computational model.

12. The network node (1 1 ) according to claim 1 1 , wherein the network node is further configured to

obtain an indication indicating a sharing policy for sharing, with other communication networks, a subset of input parameters that is allowed to be used in a computational model, and further configured to take the sharing policy into account when obtaining the additional data.

13. The network node (1 1 ) according to the claim 12, wherein the obtained

additional data are data for one or more input parameters excluded to be shared according to the sharing policy or not available in the data of second set of input parameters.

14. The network node (1 1 ) according to any of the claims 1 1 -13, wherein the

second set of input parameters differs at least in one input parameter and the obtained additional data from the distribution set of data are data of the at least one input parameter.

15. The network node (1 1 ) according to any of the claims 1 1 -14, wherein the distribution set of data is a histogram describing typical frequency distribution of data.

16. The network node (1 1 ) according to any of the claims 1 1 -14, wherein the

distribution set of data is a continuous function describing typical frequency distribution of data.

17. The network node (1 1 ) according to any of the claims 1 1 -16, wherein the

computational model is a trained model using a machine learning algorithm.

18. The network node (1 1 ) according to any of the claims 1 1 -17, wherein the

computational model is a neural network that is trained with weights and parameters.