WO2020088734A1

WO2020088734A1 - Method and recommendation system for providing an upgrade recommendation

Info

Publication number: WO2020088734A1
Application number: PCT/EP2018/079542
Authority: WO
Inventors: André Mendes Cavalcante; Maria Valéria MARQUEZINI; Pedro Henrique GOMES DA SILVA; Amadeu NASCIMENTO JUNIOR
Original assignee: Telefonaktiebolaget Lm Ericsson (Publ)
Priority date: 2018-10-29
Filing date: 2018-10-29
Publication date: 2020-05-07

Abstract

A method and a recommendation system (100) for providing a recommendation or decision for upgrading a first network element (104) of a first communication network (102). The recommendation system identifies a previously made upgrade of a second network element (106) having characteristics which are similar to characteristics of the first network element (104). At least one Key Performance Indicator, KPI, is then identified which has been improved by said previously made upgrade. The recommendation system finally provides a recommendation or decision for upgrading the first network element (104) based on the identified at least one improved KPI.

Description

METHOD AND RECOMMENDATION SYSTEM FOR PROVIDING AN UPGRADE

RECOMMENDATION

Technical field

The present disclosure relates generally to a method and a recommendation system, for providing a recommendation or decision for upgrading a first network element of a first communication network.

Background

In a typical wireless communication network, wireless devices, also known as wireless communication devices, mobile stations, stations (STA) and/or User Equipments (UE), communicate with one or more core networks (CN) via an Access Network such as a WiFi network or a Radio Access Network (RAN). The RAN covers a geographical area which is divided into service areas or cell areas, which may also be referred to as a beam or a beam group, with each service area or cell area being served by a radio network node such as a radio access node e.g., a Wi-Fi access point or a radio base station (RBS), which in some networks may also be denoted, for example, a NodeB, eNodeB (eNB), or gNB as denoted in 5G. A service area or cell area is a geographical area where radio coverage is provided by the radio network node. The radio network node communicates over an air interface operating on radio frequencies with the wireless device within range of the radio network node.

Specifications for the Evolved Packet System (EPS), also called a Fourth

Generation (4G) network, have been completed within the 3rd Generation

Partnership Project (3GPP) and this work continues in the coming 3GPP releases, for example to specify a Fifth Generation (5G) network also referred to as 5G New Radio (NR). The EPS comprises the Evolved Universal Terrestrial Radio Access Network (E-UTRAN), also known as the Long Term Evolution (LTE) radio access network, and the Evolved Packet Core (EPC), also known as System Architecture Evolution (SAE) core network. E-UTRAN/LTE is a variant of a 3GPP radio access network wherein the radio network nodes are directly connected to the EPC core network rather than to RNCs used in 3G networks. In general, in E-UTRAN/LTE the functions of a 3G RNC are distributed between the radio network nodes, e.g. eNodeBs in LTE, and the core network. As such, the RAN of an EPS has an essentially“flat” architecture comprising radio network nodes connected directly to one or more core networks, i.e. they are not connected to RNCs. To compensate for that, the E-UTRAN specification defines a direct interface between the radio network nodes, this interface being denoted the X2 interface.

In addition to faster peak Internet connection speeds, 5G planning aims at higher capacity than current 4G, allowing higher number of mobile broadband users per area unit, and allowing consumption of higher or unlimited data quantities in gigabyte per month and user. This would make it feasible for a large portion of the population to stream high-definition media many hours per day with their mobile devices, when out of reach of Wi-Fi hotspots. 5G research and development also aims at improved support of machine to machine communication, also known as the Internet of things, aiming at lower cost, lower battery consumption and lower latency than 4G equipment.

The new businesses that will be enabled by the 5G networks will demand that the network operators are agile and responsive to the ever-changing needs of the users of the network, e.g. businesses. This means that the telecom network infrastructure should be highly adaptable. Ideally, the operators should provide continuous and seamless adjustment of their infrastructure in response to new requirements from the network users. Unfortunately, this has historically not been the case. The interface between users and network infrastructure typically requires manual and somewhat ad-hoc efforts in order to transform the needs of the users into the network infrastructure changes that may be required.

Summary

It is an object of embodiments described herein to address at least some of the problems and issues outlined above. It is possible to achieve this object and others by using a method and a recommendation system as defined in the attached independent claims. According to one aspect, a method is performed by a recommendation system for providing a recommendation or decision for upgrading a first network element of a first communication network. In this method, the recommendation system identifies a previously made upgrade of a second network element having characteristics which are similar to characteristics of the first network element. The

recommendation system further identifies at least one Key Performance Indicator, KPI, which has been improved by said previously made upgrade. The

recommendation system then provides a recommendation or decision for upgrading the first network element based on the at least one identified improved KPI.

According to another aspect, a recommendation system is arranged to provide a recommendation or decision for upgrading a first network element of a first communication network. The recommendation system is configured to identify a previously made upgrade of a second network element having characteristics which are similar to characteristics of the first network element. The

recommendation system is further configured to identify at least one Key

Performance Indicator, KPI, which has been improved by said previously made upgrade and provide a recommendation or decision for upgrading the first network element based on the at least one identified improved KPI.

When using either of the above method and recommendation system, it is an advantage that a relevant and useful upgrade of a network element can be accomplished based on an automatically created recommendation or decision which does not require any investigation or analysis made by a human such as a specialist. Another advantage is that the upgrade may be performed proactively, that is before any degradation or deterioration related to the network element starts to occur. The above method and recommendation system may be configured and implemented according to different optional embodiments to accomplish further features and benefits, to be described below.

A computer program is also provided comprising instructions which, when executed on at least one processor in the above system, cause the at least one processor to carry out the method described above. A carrier is also provided which contains the above computer program, wherein the carrier is one of an electronic signal, an optical signal, a radio signal, or a computer readable storage medium.

Brief description of drawings

The solution will now be described in more detail by means of exemplary embodiments and with reference to the accompanying drawings, in which:

Fig. 1 is a communication scenario illustrating an example overview of how the solution may be employed, according to some example embodiments.

Fig. 2 is a flow chart illustrating a procedure which may be performed by a recommendation system, according to further example embodiments.

Fig. 3 is a flow chart illustrating an example of a more detailed procedure which may be performed by a recommendation system, e.g. in combination with the procedure of Fig. 2, according to further example embodiments.

Fig. 4 is a block diagram schematically illustrating how functions can be organized in a recommendation system, according to further example embodiments.

Fig. 5 is a logical block diagram schematically illustrating an example of how a recommendation system may operate for identifying an improved KPI of a network element, according to further example embodiments.

Fig. 6 is a logical block diagram schematically illustrating another example of how a recommendation system may operate for identifying network elements with degraded KPI, according to further example embodiments.

Fig. 7 is a logical block diagram schematically illustrating another example of a procedure performed in a recommendation system for identifying upgraded and un-upgraded network elements within a cluster, according to further example embodiments.

Fig. 8 is a schematic diagram illustrating a number of network elements clustered according to their characteristics, according to further example embodiments. Fig. 9 is a flow chart illustrating another procedure which may be performed by a recommendation system, e.g. in combination with the procedure(s) of Fig. 2 and/or Fig. 3, according to further example embodiments.

Fig. 10 is a schematic block diagram illustrating how a recommendation system may be structured, according to further example embodiments.

Detailed description

General recommendation systems may be based on a continuous measurement of user preference for a given item. The item may e.g. be a product, a movie, or even a friend. These systems use historical information regarding the user, e.g. items previously bought or people previously followed, and decisions made by similar users, e.g. sharing similar characteristics with regards to age, gender, zip code etc., to provide the recommendations. These recommendation systems are e.g. utilized to recommend items such as movies, books, music, friends, goods etc.

An example of a recommendation system is a system which proposes hardware and software upgrade recommendations for a managed network of devices. In this system, recommendations are generated based on information collected from the environment of the network as well as information regarding the hardware configuration and hardware/software dependency.

Another known system recommends classes to students based on their social network profiles. This recommendation system identifies a group of members with profiles similar to the student using the system, and based on skills the student is missing compared to the members with similar profile, infers courses to be recommended for the student.

Briefly described, a solution is provided where a relevant and useful

recommendation or decision for a network infrastructure upgrade is achieved in order to enable proactive adaptation of a network or network part to user requirements or preferences. The recommendation or decision for upgrading a network part or element as described herein may thus be provided proactively, that is before any degradation occurs due to shortcomings in the network part or element.

Throughout this disclosure, the term“recommendation” should be understood such that a decision to actually execute the upgrade is made, which means that the recommendation or decision for upgrading a network part or element as described herein will result in execution of the proposed upgrade, and not just a description of the upgrade.

The recommendation or decision is based on historical data related to Key Performance Indicators, KPIs, determined for the same or similar networks or network parts. By comparing how the performance, as indicated by one or more KPIs, of a similar network part has changed or improved after an upgrade has been previously performed, a recommendation or decision for making a similar upgrade to the network part at hand may be provided. By analyzing a large amount of historical data, different upgrades may be compared and weighted providing a sound statistical basis for recommendations and upgrade decisions. Furthermore, by analyzing the KPIs of the network part that may potentially be upgraded, KPIs of that network part which are degrading, i.e. getting worse, may be identified. Information regarding these degraded KPIs may then be used in conjunction with information of KPIs that were improved by upgrades of similar network parts in order to arrive at an upgrade recommendation or decision for the network part which is likely to result in an improvement.

The procedures and actions described herein may be performed by a

recommendation system which may also be referred to as an upgrading system. The term recommendation system may thus be replaced by the term upgrading system throughout this disclosure.

The recommendation system described herein may perform an analysis and provide recommendations and decisions for upgrading an entire network or a subpart of the network. The unit on which the recommendation system operates will henceforth be called a“network element”. The network element as described herein may thus be related to a network node such as an eNodeB, a virtual or physical network function, a link, a network cell, a network core, a management system such as an Operations Support System (OSS) or a Business Support System (BSS), cloud infrastructure and/or platform, a radio system, and an antenna system. Thus, the aim of the solution is to provide a method for generating

recommendations and decisions for upgrades to be performed on a network element. The recommendation or decision may be based on correlated upgrades in different network elements that exhibit similarities to the network element. The recommendation or decision may also be based on upgrades on the equipment vendor product catalogues. The upgrade recommendations and decisions may be centered on the end-user Quality of Service (QoS) or Quality of Experience (QoE). The upgrade may e.g. be a telecom network infrastructure upgrade.

As a part of developing embodiments herein, the applicant has identified a problem which will first be discussed. Today, upgrades of both hardware and software components in communication networks are typically done based on the expertise of a specialist which requires manual efforts. Furthermore, the upgrades are mostly executed reactively to changes in the network element, such as e.g. network element degradations. Known art in this field do not correlate upgrades in different network elements exhibiting similar characteristics. Furthermore, conventional upgrade procedures are not centered on improving KPI, for enhancing end-user QoS or QoE. Thus, no prior art has directly addressed hardware and software upgrades in a

communication network infrastructure based on automated recommendation systems.

A network employing a recommendation system which analyses the history of the network and constituent parts of the network, taking into consideration previous upgrades performed on the network or similar networks in order to provide upgrade recommendations and decisions, would allow a proactive upgrade procedure. Furthermore, no solution has been proposed which correlates data from multiple network elements in order to provide the recommendations and decisions.

Providing such a solution would result in more agility for network operators and thus a higher responsiveness to user requirements.

The solution will now be described in terms of functionality in a recommendation system. Fig. 1 illustrates an example of a communications scenario where a recommendation system 100 obtains information regarding a first network element 104 and one or more second network elements 106. The first network element 104 is comprised in a first communication network 102. The second network element 106 may be comprised in the first communication network 102 or in a second communication network 103 different from the first communication network 102. The first 102 and second 103 communication networks may e.g. be telecommunication networks. The recommendation system 100 analyses the obtained information, e.g. in a logical unit 100A, and provides a recommendation or decision for upgrading the first network element 104 based on the analysis. The analysis will be explained in more detail below. The first 104 and second 106 network elements may be comprised in a group of network elements 802, 804, 806 as illustrated in Fig. 8 which will be described and explained later below.

An example of how the solution may be employed in terms of actions which may be performed by a recommendation system, such as the recommendation system

100, is illustrated by the flow chart in Fig. 2, which will now be described. As noted above, the recommendation system described herein may alternatively be denoted an upgrading system. The actions in Fig. 2 could thus be performed by a recommendation or upgrading system 100 for providing a recommendation for upgrading a first network element 104 of a first communication network 102. The actions may be performed in any suitable order. Dashed boxes in Fig. 2 represent optional actions.

Action 201

According to the example embodiments described herein, the recommendation system 100 needs information regarding previous upgrades performed on the second network element 106 having similar characteristics as the first network element 104 in order to provide a relevant and useful upgrade recommendation or decision.

As a preliminary step the recommendation system 100 may bundle or classify similar network elements 802, 804, 806 into groups or clusters 800. The first network element 104 and the second network element 106 may thereafter be identified or selected from the network elements 802, 804, 806 in the cluster 800. Thus the recommendation system 100 may create or define a cluster 800 of network elements 802, 804, 806 having similar characteristics. In order to create the cluster 800 of network elements 802, 804, 806 having similar characteristics, the recommendation system 100 may previously have obtained data related to the characteristics of the network elements 802, 804, 806. The type and character of the data or information obtained will be discussed in greater detail below. Action 202

Having bundled, also referred to herein as clustered, network elements 802, 804, 806 having similar characteristics, the recommendation system 100 may limit the search for network elements 802 which have previously been upgraded to the cluster 800 of network elements sharing similar characteristics. This increases the efficiency of the search as well as allowing for statistical comparisons between a large amount of network elements 802.

Thus, if a cluster 800 of network elements 802, 804, 806 has been created as described in action 201 above, then the recommendation system may identify network elements 802 within the cluster 800 which have previously been upgraded.

Action 203

The recommendation system 100 identifies a previously made upgrade of a second network element 106 having characteristics which are similar to characteristics of the first network element 104. Examples of such similar characteristics are given below.

The recommendation system 100 may in an initial step have obtained information or data related to the characteristics of the network elements 104, 106 as described above in action 201.

Having similar characteristics as described herein means that the characteristics, such as e.g. the topology, hardware version, software version, surrounding environment etc. of the network elements 104, 106 could be similar or even identical. Characteristics being similar may e.g. herein mean that the

characteristics of the network elements 104, 106 are within an interval of each other when evaluated according to some kind of metric. For example, the physical environment may be evaluated as similar if the number of buildings associated with the network elements 104, 106 are of the same order of magnitude. Other examples may be the network elements 104, 106 being the same kind of cell type, e.g. pico-cells, or provide the same kind of services, having a software version close to each other, providing a coverage within the same magnitude etc. Some further examples may include: having similar number of connected users, having similar traffic profile such as e.g. call, video, web browsing, having a similar throughput, exhibiting similar delay, having a similar setup in terms of radio or antenna systems such as e.g. transmit power, antenna tilt, number of antenna elements, antenna height, etc.

The characteristics of the first 104 and second 106 network element may further be related to at least one of: performance of the first 104 and second 106 network elements, capabilities of the first 104 and second 106 network elements, UE usage and physical environment. These characteristics or categories will be explained in greater detail below.

The second network element 106 may be comprised in the first communications network 102. Alternatively, the second network element 106 may be comprised in a second communications network 103 which is different from the first communications network 102. Optionally, the second network element 106 may be the same as the first network element 104.

It was mentioned above that the recommendation system 100 may have created a cluster 800 of network elements 802, 804, 806 having similar characteristics and identified network elements 802 within the cluster 800 which have previously been upgraded, as described in actions 201-202 above. In that case, the

recommendation system may identify the previously made upgrade of the second network element 106 by selecting the second network element 106 from the identified network elements 802 within the cluster 800 which have previously been upgraded.

Action 204

In order to determine whether a given upgrade has had an effect on performance of network elements 802 similar to the first network element 104, the

recommendation system 100 need to evaluate the result of the upgrade. Various KPIs may be useful to provide a metric suitable for this evaluation.

Thus, the recommendation system 100 identifies at least one KPI which has been improved by the previously made upgrade.

In some illustrative but non-limiting examples, the at least one identified KPI may e.g. be cell availability, downlink throughput, upload and download traffic etc. For brevity, the at least one KPI may in some parts of this disclosure be referred to simply as a KPI. It should however be understood that there may be more than one KPI which are being identified and used in the analysis, regardless of whether the discussed KPI is an improved KPI or a degraded KPI.

Identifying the KPI may comprise analysing measurements of the KPI during a time interval starting before the previously made upgrade and ending after the previously made upgrade.

Action 205 There may be a large number of possible upgrades which may be performed in order to improve the performance of the first network element 104. In order to provide as optimal a recommendation or decision as possible, the

recommendation system 100 may identify KPIs of the first network element 104 which have been degraded. By performing upgrades on the network element 104 which improves the network element 104 such that the identified degraded KPIs are specifically targeted and improved, an efficient and relevant upgrade is achieved.

Thus, if the recommendation system 100 has created or defined a cluster 800 of network elements 802, 804, 806 having similar characteristics as described in action 201 above, the recommendation system 100 may then, as a preliminary step for identifying KPIs which have been degraded, identify network elements 806 within the cluster 800 which have degraded KPI.

Action 206 As mentioned above, the recommendation or decision for upgrading the first network element 104 may be provided by identifying at least one KPI which has been degraded, wherein the at least one degraded KPI is related to the first network element 104.

If the recommendation system 100 has identified network elements 806 within a cluster 800 which have degraded KPIs as in action 205 above, then the above identified KPI(s) which has/have been degraded may be identified from KPIs of the network elements 806 within the cluster 800. This will be described in more detail below.

Action 207 Having identified at least one improved KPI, the recommendation system is ready to provide a recommendation or decision for upgrading the first network element 104. Thus, the recommendation system 100 in this action provides a recommendation or decision for upgrading the first network element 104 based on the at least one identified improved KPI. In this action, the recommendation system 100 thus initiates the recommended upgrade in a suitable manner, e.g. by issuing an upgrade instruction.

Furthermore, if the recommendation system 100 has information which suggests that a specific successful upgrade of a second network element 106 has resulted in an improved KPI for the second network element 106, it may be assumed that such an upgrade on the first network element 104 will also improve the same KPI of the first network element 104 in a corresponding manner. Thus, the

recommendation or decision for upgrading the first network element 104 may correspond to the identified previously made upgrade of the second network element 106.

When a degraded KPI has been identified according to action 206 above, the recommendation or decision for upgrading the first network element 104, may then be further based on the identified degraded KPI. Additionally, in this case the recommendation or decision for upgrading the first network element 104 may further be based on an association or correlation between the improved KPI and the degraded KPI. An association between two KPIs means basically that they are somehow connected or related to one another so that one KPI may affect the other, without necessarily being similar.

When the recommendation system has created a cluster 800 and identified network elements 806 within the cluster 800 which have degraded KPI, then the recommendation or decision for upgrading the first network element 104 may be further based on the identified network elements 806 within the cluster 800 which have degraded KPI.

The procedure described above with reference to Fig. 2 will now be further explained and exemplified. In a telecommunication network, such as e.g. the first 102 and second 103 communication networks illustrated in Fig. 1 , a large amount of data may be available and collected from the network elements 104, 106. The data may also be referred to as information related to the network elements 104, 106. As has been described above, it is possible to evaluate the performance, parameters and configurations of the different network elements 104, 106 based on this data. The network elements 104, 106 may herein also be referred to as nodes and may include but are not limited to e.g. base stations such as an eNodeB or a gNodeB, network functions, links, etc. A method and a recommendation system 100 is herein described which use such data to provide software and hardware upgrade recommendations and decisions based on the behavior of the network elements 104, 106 as well as historical upgrades performed on the network element 106. The network element may or may not be restricted to the network of the operator performing the method. Furthermore, the network elements 104, 106 referred to herein may encompass various different granularity, such as e.g. cell, node, core etc. Data from any network elements 104, 106 may be used, e.g. as obtained from different operators, countries, cities, neighborhoods, etc.

The procedure described in conjunction with actions 201-207 above will now be described in more detail with reference to Fig. 3. When referencing to Fig. 3 the procedure will be described as including a number of tasks 301-310 being performed, although it will become apparent that these tasks 301-310 more or less correspond to the actions 201-207 described above.

The tasks in Fig. 3 may be performed by a recommendation system, such as e.g. the recommendation system 100 depicted in Fig. 1. The tasks include collecting data, identifying upgrades, matching KPI improvements to previous upgrades, clustering similar network elements, recommending upgrades based on clusters, identifying degraded KPIs and bottlenecks etc. These tasks may correspond to modules or units in the recommendation system, where a specific module performs the corresponding task. The tasks may therefore also be referred to as modules herein. Initially, the tasks will be described in a general manner, and later each individual task will be described in detail. Fig. 3 is thus a flowchart illustrating an exemplifying procedure performed by the recommendation system 100.

In task 301 , data related to at least the network elements 104, 106 are collected. The data may relate to characteristics of the network elements 104, 106, such as e.g. the topology, the hardware version, the software version, various KPIs, etc. The characteristics may also be referred to as properties of the network elements 104, 106. The data may also relate to the network elements 104, 106 indirectly, such as e.g. KPIs of UEs associated with the network elements 104, 106 etc. In relation to Fig. 2, task 301 can be performed before action 201 or before action 203 depending the implementation. The recommendation system 100 would thus, as a preparation operation, obtain data related to the network elements 104, 106.

Historical data related to characteristics of the network elements 104, 106 may in task 302 be used to identify previous network element upgrades. Task 302 correspond to action 203 described above. Information relating to the previous upgrades, such as e.g. type of upgrade, time or timestamp of when the upgrade was performed etc. may be recorded.

Having identified previous upgrades with associated timestamps, task 305 may be performed. In task 305, a series of KPIs which were collected in task 301 are analyzed. Based on the analysis it is identified, also referred to as detected, which of the KPIs have hopefully been improved after the upgrade. As has been described above, the KPIs may be related to or associated with a specific network element 104, 106. Task 305 correspond to action 204 above.

In task 308 the KPIs which have been improved by an upgrade may be used in the procedure together with KPIs which have been degraded. Thus, the improved KPIs may be associated with the respective degraded KPIs. For example, an improved KPI may be associated with a degraded KPI as follows. The number of degraded KPIs may be maximized by matching the degraded KPIs with a minimum number of different upgrades that have improved them. The information regarding the KPIs which have been degraded may be manually set in task 308, or automatically identified in tasks 303 and 306. When the degraded KPIs are automatically identified in tasks 303 and 306, the identification may be performed using the data collected in task 301. Identifying a degraded KPI according to tasks 303 and 306 corresponds to action 206 described above.

After the KPIs which have been improved by an upgrade, which were found in task 305, are correlated with degraded KPIs, which were found in tasks 303 and 306, an upgrade is recommended or decided in task 309. For example, a list may be created with network elements 104 exhibiting poor performance and a ratio of relevance for any upgrade performed in other network elements 106. The ratio of relevance for upgrades herein refers to how relevant a specific upgrade is deemed to be in order to improve the performance of the network element 104 exhibiting poor performance.

The tasks 308 and 309 correspond to the option in action 207 of correlating improved KPIs with degraded KPIs in order to provide a recommendation of an upgrade. Fig. 9 is a flow chart illustrating an example algorithm that the recommendation system 100 may employ in order to determine a recommendation or decision for upgrading the network element 104 based on a correlation between improved KPIs and degraded KPIs.

Clustering of network elements has been described above and an example of such clustering is illustrated in Fig. 8. Another alternative path for a

recommendation procedure moves to task 304 where similar network elements 802, 804, 806 are clustered using historical data, which may also be referred to as past time data, related to these network elements 802, 804, 806. Thus, the recommendation system 100 creates clusters 800 of the network elements 802, 804, 806. The historical data may e.g. relate to available product features, frequency of different alarms e.g. related to temperature, CPU usage, restarts, power outages, number of antennas, number of cells per eNodeB, cell range, number of connected UEs per cell, attachment failure rate, etc. With similar network elements is herein meant network elements having similar characteristics, such as e.g. being a cell, eNodeB, or any other granularity. Thus, task 304 corresponds to action 201 described above.

In task 307 the recommendation system 100 searches the clusters 800 created in task 304, for network elements 806 with degraded KPIs and network elements 802 that have been upgraded since the time when the data used for the clustering was collected. The search is thus performed using current up-to-date data. Task 307 correspond to actions 202 and 205 described above. The clustering will be further described in more detail below.

In task 310, the result from task 307 is used by the recommendation system 100 to recommend an upgrade to a network element 806 having degraded KPI. This recommendation is based on what happened after an upgrade was performed on another network element 802 in the same cluster 800. For example, if the upgraded network element 802 exhibited an improved KPI after being upgraded, the recommendation system 100 may recommend a similar or identical upgrade to the network element 806 having degraded KPI. Task 310 corresponds to action 207. Further, the recommendation system may initiate the upgrade in either of tasks 309 and 310 in a suitable manner.

Fig. 4 is a block diagram illustrating in more detail the data collected by the recommendation system 100 in task 301. The data collected may constitute network data 400, which may also be referred to as network element

performance characteristics 400 data. This category of data corresponds to the characteristics of the network elements 104, 106 related to the performance of the first 104 and second 106 network elements described under action 203 above.

This network performance characteristics data 400 may include, but is not limited to, the number of cells in an eNodeB, the power level of a cell, the number of connected UEs, the number of successful handovers (HO), the number of failed HOs, the download and upload data traffic e.g. in bytes, the frequency, the number of dropped calls, the number of dropped packages, the coverage, different types of alarms etc. Other data that could be collected may be referred to as the capabilities of the network element 402 and may e.g. constitute data related to the equipment model of the current network element, the software version, configuration parameters, features available on the equipment of the network element and all the hierarchy between different equipment installed in a given location. This category of data corresponds to the characteristics of the network elements 104, 106 related to the capabilities of the first 104 and second 106 network elements under action 203 above.

It is also possible to use data of UEs associated to the network elements 104, 106, herein referred to as UE usage 404. The data related to UE usage 404 may e.g. comprise Received Signal Strength Indication (RSSI), download and upload traffic, device description, geographical coordinates, the number of completed calls, or any other data related to UE network element 104, 106 usage. This category of data corresponds to the characteristics of the network elements 104, 106 related to UE usage under action 203 above.

A fourth category of data that could be collected by the recommendation system 100 may be referred to as physical environment 406 and includes any data related to the physical environment surrounding the specific network element 104, 106 for which data is collected. This may e.g. include data related to geographical features, such as e.g. buildings density, house density, city zones, roads, traffic, terrain, etc. This category of data corresponds to the characteristics of the network elements 104, 106 related to physical environment under action 203 above.

As has been described previously, the data may be obtained at different network levels, e.g., cell, eNodeB, region, etc. The level depends on how much detail is required by the recommendation system 100. The data may be used directly or may be first extracted by a machine learning technique such as e.g. a

convolutional neural network (CNN), and used thereafter. All collected data may be associated with geographical information and timestamps, since this

information may be used to consolidate data from different sources. Independent of how the data is collected, it is thereafter consolidated in a data consolidation block 408 as schematically illustrated in Fig. 4. A product information block 409 may maintain data related to product catalogues of different vendors in order to keep track of catalogue updates and changes over time. In this way, the recommendation system 100 can be aware of the different upgrades offered by the vendors. This information may then be used by the recommendation system 100 to validate if a given upgrade corresponding to a recommendation is still available. This validation is performed by consulting the information from the different catalogues of the vendors, which information should be up-to-date. If the upgrade is not available, the recommendation system 100 may try to find a new product in the catalogue information which new product correspond to the desired upgrade.

As has been described above, in task 302 the recommendation system 100 may use information such as hardware model, software version, network topology, device features and configuration settings to identify upgrades performed in a network element 106. It is possible to identify upgrades in the network element 106 by tracking changes in e.g. hardware, software, model, version, network element topology and node features. Once an upgrade is detected, the

recommendation system 100 may analyze and compare the behavior of different network elements 104, 106. This analysis should result in an identification of the KPIs which have been improved by the upgrade. This identification is performed in task 305 which is illustrated in more detail in Fig. 5.

Fig. 5 thus schematically illustrates a procedure for identifying KPIs which have been improved after upgrades. The recommendation system 100 may isolate a time window around the upgrade, e.g. a hardware or software upgrade, and analyze KPIs in this window to identify improved KPIs. In Fig. 5 a detailed description of the process between tasks 301 , 302 and 305 is illustrated. The corresponding actions are 203 and 204. In order to perform task 302 the recommendation system 100 need to collect historical data to identify the upgrades. This is illustrated in block 500 in Fig. 5.

The historical data may e.g. be from a particular time frame, in a particular resolution and for specific network element characteristics or features. The time frame may e.g. be the past six months, the past year, etc. The resolution may e.g. be per hour, per day, etc. Examples of the network element characteristics may e.g. be software/hardware version, hardware model, specifications, number of equipment in the network element, etc. Having obtained this data, the

recommendation system 100 may in task 302 identify upgrades and their associated timestamps, as illustrated in block 502.

Using the timestamps, the recommendation system 100 may analyze e.g. a set of KPIs, alarms and network element behaviors before and after the upgrade was performed in order to identify the improved KPIs, as illustrated in block 504. The KPIs may e.g. comprise cell availability, downlink throughput, upload and download traffic, etc. The alarms may e.g. comprise sleeping cell alarms, temperature alarms, partial outages, cyclic restarts, frequency of crashes etc. The network element behavior may e.g. be related to high-level correlations or associations such as between external events and service degradations, traffic pattern, KPI degradation, number of connected UEs and frequency of alarms. External events may e.g. comprise weather or venue events, etc. The network element behavior may thus represent a higher-order behavior which is caused by external event.

The analysis in block 504 may e.g. be performed by means of trend analysis, such as e.g. extracting or determining the first derivative of KPIs and/or alarms. A further example of the analysis may e.g. be a time series analysis, such as e.g.

Autoregressive Moving Average (ARMA), Autoregressive Integrated Moving Average (ARIMA), Autoregressive Moving Average Model with exogenous inputs (ARMAX), etc. Furthermore, the analysis may consider special events in the region, such as e.g. football matches, concerts, etc., and analyze what happened in the network element 106 during those events and the effect of the upgrades on the network element 106 in relation to those events.

In task 303 bottlenecks in the network elements 104, 106 may be detected or found. With bottlenecks is herein meant network elements 104, 106 experiencing significant degradation, e.g. in terms of data throughput and/or latency, so that certain performance requirements are not met. If these network elements 104, 106 are part of a larger network unit they may act as bottlenecks for the larger unit. If the network elements 104, 106 experiencing problems already are the largest units they may be considered bottlenecks for their own functioning. The

bottlenecks may be determined by specialists within the specific domain or they may be automatically identified, which is shown in Fig. 6. As illustrated in Fig. 6, the recommendation system 100 may in task 303 analyze thresholds or the frequency of specific events related to KPIs to identify network elements 104 likely to present some problems. To automatically identify network elements 104, 106 experiencing problems, a set of KPIs may be obtained, e.g. collected as illustrated in block 600, and then analyzed. The obtained KPIs may correspond to the KPIs collected when determining a KPI which have been improved by an upgrade above. The KPIs may thus e.g. comprise cell availability, mobility success rate, downlink throughput, uplink throughput and node restarts. KPIs related to UEs associated with the network element 104, 106 may also be used, such as e.g. RSSI, packet error rate, download traffic and traffic. Having obtained the data the recommendation system may analyze the data in order to identify network elements with problems, as illustrated by block 602. From the network elements experiencing problems, network elements 104 having degraded KPI may be identified. This is performed in task 306.

The analysis in task 306 may e.g. be performed by comparing the obtained KPIs to pre-defined thresholds. Other ways of identifying network elements 104 having problems or issues may be to perform time series analysis of the data, or by analyzing the frequency of specific events. A further approach may be to use a standard behavior of a network element as a baseline to be compared with the present network 104. Additionally, supervised or unsupervised learning techniques may be applied to identify network elements 104 experiencing problems. The learning techniques may e.g. be machine learning techniques such as e.g. neural networks, classification trees, K-means, k-Nearest Neighbor algorithms (KNN), Density-Based Spatial Clustering of Applications with Noise (DBSCAN) etc. Any of these approaches may use one or more sets of KPIs. Tasks 303 and 306 correspond to action 206 described above. Figs. 7 and 8 schematically illustrate the clustering or grouping of network elements 802, 804, 806 having similar characteristics which may be performed by the recommendation system in tasks 304 and 307. The recommendation system 100 may initially obtain data related to the characteristics of the network elements 802, 804, 806, see block 700 in Fig. 7 as well as task 301 in Fig. 3. The obtained data may then be used by the recommendation system 100 to identify network elements 802, 804, 806 that have similar characteristics or behavior. These network elements 802, 804, 806 may then be clustered by the recommendation system 100, i.e. the recommendation system 100 creates a cluster 800 of network elements 802, 804, 806 having similar characteristics as illustrated in block 702 of Fig. 7 and task 304 in Fig 3.

The clusters 800 may be found or identified by e.g. using a machine learning or statistical technique where characteristics of the network elements 802, 804, 806 are analyzed. Examples of such characteristics may comprise number of cells, power level, number of connected UEs, number of HO failures, the ratio of successful HO, ratio of dropped calls, UE throughput, downloaded and uploaded data, dropped packages, latency and CPU usage. Further, data such as e.g.

density of buildings, density of houses, landmarks and terrain information may also be used. To obtain the clusters 800, unsupervised learning techniques such as e.g., K-Means, DBSCAN, K-NN and X-Means may be applied to the set of KPIs and other data as mentioned. Thus, task 304 where the cluster 800 is created corresponds to action 201 described above.

When having obtained the clusters 800 of network elements 802, 804, 806, the recommendation system 100 may analyze the network elements 802, 804, 806 within the cluster 800 further. This further analysis may entail grouping the network elements 802, 804, 806 within the cluster 800 into different groups depending on whether they have been upgraded or experiences problems. Thus, in task 707 the recommendation system 100 may identify network elements 802 that have been previously upgraded. The improved KPIs may then be identified or selected from this group of network elements 802. The remaining network elements 804, 806 in the cluster 800 may then be analyzed in order to identify network elements 806 having degraded KPIs, as is performed in task 306 and action 205. In this way two groups of network elements are obtained, upgraded network elements 802 and un-upgraded network elements having degraded KPIs 806. This is illustrated in blocks 704 and 706 of Fig. 7 as well as in Fig. 8. Finally, for the cluster 800 a recommendation or decision may be provided for upgrading a network element 806 within the cluster 800, corresponding to task 310 illustrated in Fig. 3. In turn, task 310 corresponds to a special case of action 207. The recommendation or decision may be provided based on a correlation of network elements 802 having KPIs which have been improved by an upgrade and network elements 806 having degraded KPIs. In task 310, the similarity of the network elements 802 and 806 has thus been taken into consideration through the clustering procedure.

The recommendation system 100 may use the same technique as in task 308 to correlate the network elements 802 having improved KPIs with the network elements 806 having degraded KPIs in order to recommend an upgrade. For example, by directly matching KPIs improved by a previous upgrade of one network element 802 with degraded KPIs of another network element 806, i.e. matching the same KPIs but associated with different network elements, the recommendation system 100 may suggest the same upgrade to the network elements 806 having degraded KPIs. However, as upgrades often improve more than one KPI, weights for different KPIs may be inferred or determined from the plurality of upgrades which have been performed.

The weights for different upgrades may be assigned based on the similarity between network elements, e.g. cosine distance, person correlation, Euclidean distance or fractional distances. Further, a score may be calculated using the number of matching degraded and improved KPIs weighted by the similarity between the network elements. In an optional implementation, a match between degraded and improved KPIs may be weighted by the inverse frequency of that given KPI on the upgrade candidates of the network element. Finally, the upgrade with the greatest score can be assigned with the highest priority. The weights may thereafter be used to provide the best match between degraded KPIs and upgrades, i.e. the best upgrade to perform in order to improve a specific KPI which have been degraded. The weighting may e.g. be performed by the

recommendation system 100 in conjunction with tasks 309 and 310 in Fig. 3.

The outcome or end-result of the recommendation system 100 performing either of tasks 309 and 310 and action 207 may include a list with network elements 104, 806 and a measure of the aptness or fitness of each upgrade for the respective network element 104, 806. The recommended and decided upgrade may e.g. be performed on an equipment or a set of equipment installed in a network element 104, 806. The upgrade may also involve a software upgrade. The type of upgrade achieved depends on the level of abstraction desired. Therefore, the above list may be used as a basis for some upgrades to one specific network element 104, 806 or an upgrade to a set of network elements 104, 806.

Since the solution provides recommendations and decisions for network element 104 upgrades based on a recommendation system 100 and the recommendation system 100 uses previous information from the network element 104 being upgraded or other similar network elements 106 in different countries, cities, and across different operators, the recommendation system 100 enables upgrades to be performed in a proactive manner. These proactive upgrades have higher probability of improving, or avoiding reduction of, the performance of the network element 104 since the recommended upgrade is based on previous similar situations where improvements due to such similar upgrades were detected.

Furthermore, the upgrade recommendations are flexible from the point of view of the parameters, i.e. the KPIs, to be used for the upgrade recommendation analysis. For instance, the recommendation system 100 may use indexes related to end-user QoE or other business-related KPIs, which makes the upgrade recommendations flexible to the goal of the operator.

The block diagram in Fig. 10 illustrates a detailed but non-limiting example of how a recommendation system 1000 may be structured to bring about the above- described solution and embodiments thereof. In this figure, the recommendation system 1000 may be configured to operate according to any of the examples and embodiments of employing the solution as described herein, where appropriate. The recommendation system 1000 is shown to comprise a processor“P”, a memory“M” and a communication circuit“C” with suitable equipment for transmitting and receiving information and messages in the manner described herein.

The communication circuit C in the recommendation system 1000 thus comprises equipment configured for communication using a suitable protocol for the communication depending on the implementation. The solution is however not limited to any specific types of messages or protocols. The recommendation system 1000 is, e.g. by means of units, modules or the like, configured or arranged to perform at least some of the actions of the flow chart in Fig. 2 as follows.

The recommendation system 1000 is arranged to provide a recommendation or decision for upgrading a first network element 104 of a first communication network 102. The recommendation system 1000 is configured to identify a previously made upgrade of a second network element 106 having characteristics which are similar to characteristics of the first network element 104. This operation may be performed by an identifying module 1000B in the recommendation system 1000, as illustrated in action 203. The identifying module 1000B could alternatively be named an analysing module, a determining module or a selecting module. The second network element 106 may be comprised in the first communications network 102. The second network element 106 may alternatively be comprised in a second communications network 103 which is different than the first communications network 102. Optionally, the second network element 106 may be the same as the first network element 104. The characteristics of the first

104 and second 106 network element may be related to at least one of:

performance of the first 104 and second 106 network elements, capabilities of the first 104 and second 106 networks elements, UE usage and physical environment.

The recommendation system 1000 is further configured to identify at last one KPI which has been improved by said previously made upgrade. This operation may be performed by the identifying module 1000B in the recommendation system 1000, as illustrated in action 204. The recommendation system 1000 may be configured to analyse measurements of the at last one improved KPI during a time interval starting before the previously made upgrade and ending after the previously made upgrade to identify the at last one KPI.

The recommendation system 1000 is further configured to provide a

recommendation or decision for upgrading the first network element 104 based on the identified at last one improved KPI. This operation may be performed by a providing module 1000C in the recommendation system 1000, as illustrated in action 207. The providing module 1000C could alternatively be named a recommendation or decision module. The recommendation or decision for upgrading the first network element 104 may correspond to the identified previously made upgrade of the second network element 106.

The recommendation system 1000 may further be configured to identify at last one KPI which has been degraded, wherein the at last one degraded KPI is related to the first network element 104. This operation may be performed by the identifying module 1000B in the recommendation system 1000, as illustrated in action 206. In this case the recommendation system 1000 may be configured to provide a recommendation or decision for upgrading the first network element 104 where the recommendation or decision is further based on the identified at last one degraded

KPI. The recommendation or decision for upgrading the first network element 104 may in this case further be based on an association between the at last one improved KPI and the at last one degraded KPI. This operation may be performed by the providing module 1000C in the recommendation system. The recommendation system 1000 may further be configured to create a cluster

800 of network elements 802, 804, 806 having similar characteristics in order to identify the previously made upgrade of the second network element 106. This operation may be performed by a creating module 1000A in the recommendation system 1000, as illustrated in action 201 . The creating module 1000A could alternatively be named a clustering module or grouping module. The

recommendation system 1000 is in this case further configured to identify network elements 802 within the cluster 800 which have previously been upgraded. The recommendation system 1000 is in this case further configured to select the second network element 106 from the identified network elements 802 within the cluster 800 which have previously been upgraded. This operation may be performed by the identifying module 1000B in the recommendation system 1000, as illustrated in action 202.

When the recommendation system 1000 has created a cluster 800 of network elements 802, 804, 806 having similar characteristics, it may be further configured to identify network elements 806 within the cluster 800 which have degraded KPI. This operation may be performed by the identifying module 1000B in the recommendation system 1000, as illustrated in action 205. In this case, the recommendation system 1000 may be configured to provide a recommendation or decision for upgrading the first network element 104 where the recommendation or decision is further based on the identified network elements 806 within the cluster 800 which have degraded KPI. This operation may be performed by the providing module 1000C in the recommendation system 1000, as illustrated in action 207.

It should be noted that Fig. 10 illustrates various functional modules in the recommendation system 1000 and the skilled person is able to implement these functional modules in practice using suitable software and hardware equipment. Thus, the solution is generally not limited to the shown structure of the

recommendation system 1000, and the functional modules therein may be configured to operate according to any of the features, examples and

embodiments described in this disclosure, where appropriate.

The functional modules 1000A-C described above may be implemented in the recommendation system 1000 by means of program modules of a computer program comprising code means which, when run by the processor P causes the recommendation system 1000 to perform the above-described actions and procedures. The processor P may comprise a single Central Processing Unit (CPU), or could comprise two or more processing units. For example, the processor P may include a general purpose microprocessor, an instruction set processor and/or related chips sets and/or a special purpose microprocessor such as an Application Specific Integrated Circuit (ASIC). The processor P may also comprise a storage for caching purposes.

The computer program may be carried by a computer program product in the recommendation system 1000 in the form of a memory having a computer readable medium and being connected to the processor P. The computer program product or memory M in the recommendation system 1000 thus comprises a computer readable medium on which the computer program is stored e.g. in the form of computer program modules or the like. For example, the memory M may be a flash memory, a Random-Access Memory (RAM), a Read-Only Memory (ROM) or an Electrically Erasable Programmable ROM (EEPROM), and the program modules could in alternative embodiments be distributed on different computer program products in the form of memories within the recommendation system 1000.

The solution described herein may be implemented in the recommendation system 1000 by a computer program comprising instructions which, when executed on at least one processor, cause the at least one processor to carry out the actions according to any of the above embodiments and examples, where appropriate.

The solution may also be implemented at the recommendation system 1000 in a carrier containing the above computer program, wherein the carrier is one of an electronic signal, optical signal, radio signal, or computer readable storage medium.

While the solution has been described with reference to specific exemplifying embodiments, the description is generally only intended to illustrate the inventive concept and should not be taken as limiting the scope of the solution. For example, the terms“recommendation system”,“network element”,“Key

Performance Indicator, KPI” and“cluster” have been used throughout this disclosure, although any other corresponding entities, functions, and/or parameters could also be used having the features and characteristics described here. The solution is defined by the appended claims.

Claims

1. A method for providing a recommendation or decision for upgrading a first network element (104) of a first communication network (102), the method comprising: - identifying (203) a previously made upgrade of a second network element (106) having characteristics which are similar to characteristics of the first network element (104),

- identifying (204) at least one Key Performance Indicator, KPI, which has been improved by said previously made upgrade, and - providing (207) a recommendation or decision for upgrading the first network element (104) based on the at least one identified improved KPI.

2. A method according to claim 1 , wherein said recommendation or decision for upgrading the first network element (104) corresponds to the identified previously made upgrade of the second network element (106).

3. A method according to claim 1 or 2, wherein identifying the at least one

KPI comprises analysing measurements of the at least one KPI during a time interval starting before the previously made upgrade and ending after the previously made upgrade.

4. A method according to any of claims 1-3, wherein said characteristics of the first (104) and second (106) network element are related to at least one of:

- performance of the first (104) and second (106) network elements,

- capabilities of the first (104) and second (106) network elements,

- User Equipment, UE, usage, and

- physical environment.

5. A method according to any of claims 1-4, wherein the recommendation or decision for upgrading the first network element (104) is provided by identifying (206) at least one KPI which has been degraded, wherein the at least one degraded KPI is related to the first network element (104), and wherein the recommendation for upgrading the first network element (104) is further based on the identified at least one degraded KPI.

6. A method according to claim 5, wherein the recommendation or decision for upgrading the first network element (104) is further based on an association between the at least one improved KPI and the at least one degraded KPI.

7. A method according to any of claims 1-6, wherein the second network element (106) is comprised in the first communications network (102).

8. A method according to any of claims 1-6, wherein the second network element (106) is comprised in a second communications network (103) which is different than from first communications network (102).

9. A method according to any of claims 1-8, wherein the previously made upgrade of the second network element (106) is identified by creating (201 ) a cluster (800) of network elements (802, 804, 806) having similar characteristics, identifying (202) network elements (802) within the cluster (800) which have previously been upgraded, and selecting the second network element (106) from the identified network elements (802) within the cluster (800) which have previously been upgraded.

10. A method according to claims 5 and 9, wherein the degraded KPI is identified by identifying (205) network elements (806) within the cluster (800) which have degraded KPI and wherein the recommendation for upgrading the first network element (104) is further based on the identified network elements (806) within the cluster (800) which have degraded KPI.

1 1. A method according to any of claims 1-6, wherein the second network element (106) is the same as the first network element (104).

12. A recommendation system (1000) arranged to provide a

recommendation or decision for upgrading a first network element of a first communication network, the recommendation system being configured to:

- identify a previously made upgrade of a second network element having characteristics which are similar to characteristics of the first network element,

- identify at least one Key Performance Indicator, KPI, which has been improved by said previously made upgrade, and

- provide a recommendation or decision for upgrading the first network element based on the identified at least one improved KPI.

13. A recommendation system (1000) according to claim 12, wherein said recommendation or decision for upgrading the first network element corresponds to the identified previously made upgrade of the second network element.

14. A recommendation system (1000) according to claim 12 or 13, wherein the recommendation system is configured to identify the at least one KPI by analysing measurements of the at least one KPI during a time interval starting before the previously made upgrade and ending after the previously made upgrade.

15. A recommendation system (1000) according to any of claims 12-14, wherein said characteristics of the first and second network element are related to at least one of:

- performance of the first and second network elements,

- capabilities of the first and second network elements,

- User Equipment, UE, usage, and

- physical environment.

16. A recommendation system (1000) according to any of claims 12-15, wherein the recommendation system is configured to provide the recommendation or decision for upgrading the first network element by identifying at least one KPI which has been degraded, wherein the at least one degraded KPI is related to the first network element, and wherein the recommendation for upgrading the first network element is further based on the identified at least one degraded KPI.

17. A recommendation system (1000) according to claim 16, wherein the recommendation or decision for upgrading the first network element is further based on an association between the at least one improved KPI and the at least one degraded KPI.

18. A recommendation system (1000) according to any of claims 12-17, wherein the second network element is comprised in the first communications network.

19. A recommendation system (1000) according to any of claims 12-17, wherein the second network element is comprised in a second communications network which is different from the first communications network.

20. A recommendation system (1000) according to any of claims 12-19, wherein the recommendation system is configured to identify the previously made upgrade of the second network element by creating a cluster of network elements having similar characteristics, identifying network elements within the cluster which have previously been upgraded, and selecting the second network element from the identified network elements within the cluster which have previously been upgraded.

21. A recommendation system (1000) according to claims 16 and 20, wherein the recommendation system is configured to identify the degraded KPI by identifying network elements within the cluster which have degraded KPI and wherein the recommendation for upgrading the first network element is further based on the identified network elements within the cluster which have degraded KPI.

22. A recommendation system (1000) according to any of claims 12-17, wherein the second network element is the same as the first network element.

23. A computer program comprising instructions which, when executed on at least one processor, cause the at least one processor to carry out the method according to any one of claims 1 -1 1.

24. A carrier containing the computer program of claim 23, wherein the carrier is one of an electronic signal, optical signal, radio signal, or computer readable storage medium.