WO2022019728A1 - Procédé et système de détection de seuils dynamiques pour indicateurs de rendements clés (irc) en réseaux de communication - Google Patents

Procédé et système de détection de seuils dynamiques pour indicateurs de rendements clés (irc) en réseaux de communication Download PDF

Info

Publication number
WO2022019728A1
WO2022019728A1 PCT/KR2021/009653 KR2021009653W WO2022019728A1 WO 2022019728 A1 WO2022019728 A1 WO 2022019728A1 KR 2021009653 W KR2021009653 W KR 2021009653W WO 2022019728 A1 WO2022019728 A1 WO 2022019728A1
Authority
WO
WIPO (PCT)
Prior art keywords
kpi
data
module
anomaly
threshold
Prior art date
Application number
PCT/KR2021/009653
Other languages
English (en)
Inventor
Saravanan Balasubramanian
Seema Pareek
Varadarajan SEENIVASAN
Akash Roy
Manjunath Channappagoudar
Original Assignee
Samsung Electronics Co., Ltd.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Samsung Electronics Co., Ltd. filed Critical Samsung Electronics Co., Ltd.
Publication of WO2022019728A1 publication Critical patent/WO2022019728A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/04—Arrangements for maintaining operational condition

Definitions

  • the present invention relates to the wireless communication networks and more particularly relates to dynamic threshold detection for key performance indicators in communication networks.
  • the KPIs on the Network Elements (NE) and the backhaul that connect the KPIs and the NE to maintain the network in an operational state is necessarily measured and monitored.
  • the KPIs threshold values for maintaining the KPI metric are predetermined at the NE /Cell level and not at more granularity level like hourly or daily. This is because maintaining the performance KPI metric at more granularity level like hourly or daily the usually takes long time for an operator to determine whether the cause of the cell's performance deterioration is in an alarm. Further, the huge amount of time and cost will be spent considering the number of cells in the commercial network where there are thousands of devices connected to each other.
  • the intelligent thresholding exist which uses the time history of the parameters and adjusts the threshold, based on that history. There the threshold is learnt by looking at the past values of the measurement.
  • the thresholds utilized dynamically and automatically adjust based on trends and patterns in the KPI values.
  • the solutions provided above defines/predetermines the threshold value statically from past results and will not give accurate result.
  • the present invention relates to method and system for dynamic threshold detection for key performance indicators in communication networks.
  • a dynamic threshold detection method includes monitoring received, KPI data corresponding to a plurality of network nodes, wherein the KPI data includes a plurality of KPI values of each of the plurality of the network node.
  • the method further includes detecting, at least one KPI anomaly value in the KPI data based on a trained ML model.
  • the method includes analyzing a KPI metrics associated with the KPI data based on the detection of the at least one KPI anomaly value. Subsequently, the method includes determining, as a KPI threshold a specific anomaly value among the detected at least one KPI anomaly value based on the analysis of the KPI metrics.
  • the present invention may provide method and system for dynamic threshold detection for key performance indicators in communication networks.
  • Figure 1 a illustrates a block diagram of a system for dynamic threshold detection for key performance indicators in communication networks, according to an embodiment of the present subject matter
  • Figure 2 a illustrates a block diagram of a system for determining threshold values for network KPIs dynamically of 5G communication system, according to an embodiment of the present subject matter
  • Figure 3 a illustrates a flow diagram depicting an exemplary embodiment of a method for dynamic threshold detection for key performance indicators in communication networks according to an embodiment of the present subject matter
  • Figure 4 a illustrates a flow diagram depicting method for evaluating the anomaly, according to an embodiment of the present subject matter
  • Figure 5 a illustrates a flow diagram depicting an exemplary embodiment of a KPI to log / alarm message correlation, in accordance with an embodiment of the present subject matter
  • Figure 6 a illustrates a flow diagram depicting an exemplary embodiment of a method for a KPI to log / alarm message correlation, in accordance with an embodiment of the present subject matter
  • Figure 7 a illustrates tables for showing a comparison between conventional way of threshold setting and the threshold setting and calculation using the proposed solution, according to an embodiment of the present subject matter.
  • Figure 8 a illustrates tables for showing a comparison between conventional way of threshold setting and the threshold setting based on the granularity, according to an embodiment of the present subject matter.
  • any terms used herein such as but not limited to “includes,” “comprises,” “has,” “consists,” and grammatical variants thereof do NOT specify an exact limitation or restriction and certainly do NOT exclude the possible addition of one or more features or elements, unless otherwise stated, and furthermore must NOT be taken to exclude the possible removal of one or more of the listed features and elements, unless otherwise stated with the limiting language “MUST comprise” or “NEEDS TO include.”
  • Figure 1 illustrates the block diagram 100 a system for dynamic threshold detection for key performance indicators in communication networks that includes a plurality of network nodes103a, 103b,103n, a server 107, a data preprocessor module 111, an anomaly detection module 115, an anomaly prediction module 117, a threshold detection module 113, a recommendation engine 121 and a KPI correlation module 119, according to an embodiment of the present subject matter.
  • the NMS/EMS receives a plurality of KPIs values for at least one network node.
  • the KPIs can take anywhere from a few hours to a few days depending upon the circle/area under analysis.
  • the NMS/EMS receives the plurality of KPIs values at least one of circle wise or country wise. In an example, one circle covers 5-8 NMS/EMS.
  • the NMS determines the threshold values for Network KPIs dynamically.
  • the server maybe configured to consolidate one or more of country wise data, or region wise data or time wise data, or a circle wise data.
  • the server 107 include data the pre-processor module 111, the anomaly detection module 115, the anomaly prediction module 117, and the recommendation engine 121.
  • the received KPI values are inputted from the data pre-processor module 111 to the anomaly detection module 115 and anomaly prediction modules 117.
  • the KPI correlation module 119 may be configured to co-relates the received KPI values with at least one of an alarm, an observed error, current date and time, location of the cell to identify an anomaly.
  • the KPI correlation module 119 may be configured to correlates the alarms/events and analyzing the exception/error from the log whenever KPI anomaly is detected.
  • the anomaly detection module 115 may be configured to detect the anomaly in the live communication network.
  • the anomaly prediction module 117 may be configured to predict a possibility of a time at which at least one KPI value of the plurality of KPI values reaches a specific threshold, and the prediction of the possibility of the time is based on an analysis of a past trend of the KPI data.
  • the recommendation engine 121 may be configured to recommend at least one action based on the identification of anomaly in the live network communication system. The operation of system described below may be performed by the server 107 or a plurality of server devices.
  • FIG. 2 illustrates the block diagram 200 a system for determining threshold values for Network KPIs dynamically of 5G communication system 200.
  • the present invention can be applied in 5G system where there are various Radio Unit (RU), Central unit (CU), Access Unit (AU) and Distributed unit (DU), Access and Mobility Management Function (AMF)/ Session Management Function (SMF), User Plane Function (UPF) and other Network Functions (NFs) are indicated which are connected as a network through network slice.
  • RU Radio Unit
  • CU Central unit
  • AU Access Unit
  • DU Distributed unit
  • AMF Access and Mobility Management Function
  • SMF Session Management Function
  • UPF User Plane Function
  • NFs Network Functions
  • the gNBs are connected to an orchestrator (EMS/CMS& NSM).
  • EMS/CMS& NSM orchestrator
  • there are around 250 management systems are deployed and there are around 2000-10000 NFs are connected to each of the management system.
  • the management system receives a plurality of KPIs values for at least one network node.
  • the KPIs can take anywhere from a few hours to a few days depending upon the circle/area under analysis.
  • the management software system receives the plurality of KPIs values at least one of circle wise or country wise.
  • one circle covers 5-8 Management nodes.
  • the MDAS (Management Data Analytics Service) 209 determines the threshold values for Network KPIs dynamically with the added functionality defined in the Threshold detection module 213.
  • the threshold values are not predetermined and static values and are dynamically determined.
  • the dynamically threshold values are learnt based on cell geographical location, alarm/events occurred, and error/exception observed in the logs in real time. Further in the embodiment, dynamically determined threshold values are for each KPI in the network at a granular level (hourly, daily, monthly based on its location). Further, the threshold values of the KPIs are adjusted by the NMS/EMS to prevent the identified anomaly.
  • Figure 3 illustrates a flow diagram 300 depicting an exemplary embodiment of a method for dynamic threshold detection for key performance indicators in communication networks, according to an embodiment of the present subject matter.
  • the KPI data corresponding to a plurality of network nodes may be received by the server 107.
  • the method 300 may include monitoring at a step 301 by the threshold detection module 113, a KPI data corresponding to a plurality of network nodes, wherein the KPI data includes a plurality of KPI values of each of the plurality of the network node.
  • the method 300 may include detecting at a step 303, by the threshold detection module 113, at least one KPI anomaly value in the KPI data based on a trained ML model.
  • the threshold detection module 113 may be configured to identify the trained ML model to perform a gaussian analysis of the KPI data to detect the at least one KPI anomaly value.
  • the ML model may be trained on using a KPI data based on at least one of network characteristics associated with quality Class, connection setup, QoS Class Identifier (QCI) class and mobility class for a specific time, and wherein the KPI data is received from the plurality of network nodes.
  • QCI QoS Class Identifier
  • the method 300 include analyzing at a step 305, by the threshold detection module 113, a plurality of KPI metrics associated with the KPI data based on the detection of the at least one KPI anomaly value.
  • the KPI metrics includes information associated with at least one of an occurrences of an alarm, an occurrence of an event, an observation of an exception, an observation of an error, or a geographical location of a cell within the live network communication system.
  • alarm is an asynchronous notification sent from the network node to management system to take action on abnormality. For example, CPU usage increased to 90% with resource alarm.
  • Event is an asynchronous notification sent from the network node to management system on network change which does not require any operator acknowledge/action like new neighbor is added for the cell.
  • Exception/Error is messages are seen in the SW block of network node or management node which says about the problem occurred due to software. Like DSP Restart happened continuously causing call drop/failure. Abnormality of Geolocation is based on cell location, interference will be high. So based on the KPIs related to measure interference, will recommend antenna tilt change, SSF change.
  • the method includes determining at a step 307, as a KPI threshold by the threshold detection module 113, a specific anomaly value among the detected at least one KPI anomaly value based on the analysis of the KPI metrics.
  • the method may includes identifying by the KPI correlation module 119, an anomaly in the live network communication system, based on a correlation between the KPI threshold value and the plurality of KPI values in the KPI data.
  • the method further includes determining, by the threshold detection module 113, the KPI threshold at each of a plurality of time intervals at which the presence of the information is determined by the KPI correlation module 119.
  • the method may further include predicting, by the anomaly prediction module 117, a possibility of a time at which at least one KPI value of the plurality of KPI values reaches a specific threshold, and the prediction of the possibility of the time is based on an analysis of a past trend of the KPI data.
  • the method further includes recommending, by the recommendation engine 121, at least one action based on the identification of anomaly in the live network communication system.
  • Figure 4 illustrates a flow diagram 400 method for evaluating the anomaly, according to an embodiment of the present subject matter.
  • the method includes at a step 401, loading, by data preprocessor module 111, the KPI data organized by cell and time. Further, a variation in a plurality of environmental parameters such as geo-location of cells no of users, climate conditions, interference, tropospheric ducting affecting the load on the network is monitored over a time. Further, a variation in the operating threshold of a plurality of KPIs in response to the variation of the environmental parameters is monitored over a time. The method further includes determining at a step 403, by the anomaly detection module 115, the anomaly.
  • the method includes detecting at step 405, anomaly using at least one of a trained machine learning model or an unsupervised learning based on at least one statistical model.
  • the method further includes correlating at a step 407, the detected anomaly with the KPI metrics.
  • the KPI metrics includes information associated with at least one of an occurrences of an alarm, an occurrence of an event, an observation of an exception, an observation of an error, or a geographical location of a cell within the live network communication system.
  • the variation of the operating threshold of the KPIs relative to the variation of the environmental parameters is correlated. Further, a change is detected in a first environmental parameter indicative of an anomaly in relation to the operating threshold of a first KPI.
  • the method further includes deciding at a step 409, by the threshold detection module 113, the anomaly value as KPI threshold if there is associated alarm or exception observed.
  • Air properties in cellular network has KPIs which change per cell due to geographical nature, physical interference, and number of users,
  • Air properties has KPIs where each cell each hour have a different behavior due to user activity variations
  • Air properties has KPIs where values are naturally GAUSSIAN in nature where most cell-hour KPI values are around average and slowly decreasing on either side i.e., Air metrics are naturally continuous, uncontrolled and GAUSSIAN.
  • a neural network such as XGBOOST is selected, which is best close to statistical z-score technique as statistical analysis is impossible at site due to time and space consideration.
  • the Z-score based supervised "Extreme Gradient Boosting" (XGBoost) ML algorithm is used which gives very high accuracy of 98-99% anomaly detection.
  • the operating threshold of the first KPI is varied as a result of the correlation, to suit the variation in the first environmental parameter, so as to minimize the load fluctuation on the network.
  • the performance KPI information collected from mobile communications network corresponds to the step 201.
  • the data is divided into two sets- training set and validation set. Training data is collected for one month corresponding to 100 cells belonging to a city.
  • the testing data constitute of 3 weeks' worth of KPI information corresponding to same city and cells as in training dataset.
  • the KPI data collected for training and testing are aggregated at a day level, to get a generalized KPI trend across a month. Since most of KPI are correlated,
  • the KPI correlation module 119 is configured to selected more than 60 important KPI that provides overall health of the network.
  • the server 107 may be configured to received data from at least 100 cells data across a cosmopolitan region for almost 60 selected important KPI's mainly in quality class like Call drop rate, throughput, Packet delay, Session time, connection setup, QCI Class and mobility class for every hour for 15 days from the live 4G LTE network. These KPI data are collected as PM files and transferred to the data preprocessor module.
  • the threshold detection module 113 identify XGBoost, which gives 98-99% accuracy results. Whenever anomaly is detected by Threshold detection model using XGBoost, then the KPI correlation engine 119 checks for any alarm/event occurred or any exception/error observed in the collected logs from the same management system where the KPI data is collected. If there is any matching pattern found from the log sequence, then that value decided as threshold value for that KPI. This model was tested with ⁇ 80000 cells of unseen data from one circle.
  • the model may be learns in 3 to 15 days to build a scalar model for the same and the same threshold detection system can be used for 5G various network components to find its threshold value dynamically.
  • Table 1 provide non limiting example of KPI data classification.
  • KPI classification Sr. no. Class KPI Category Description 1 Connection Connection setup success rate Success rate of overall connections to device and S1 interface Number of Connections Number of connections attempted 2 Quality Call drop rate Percentage of calls released abnormally Session time Avg. session time per UE Throughput Statistical information related to cell throughput Packet delay Avg. delay between packets Number of connected users Number of active users 3 QCI QCI1 data volume Data consumed by voice services QCI2 data volume Data consumed by video services QCI5 data volume Data consumed by IMS services QCI9 data volume Data consumed by streaming services 4 Mobility Handover success rate Percentage of handover between multiple interfaces (like X2, S1) Number of successful handovers Number of successful handovers between multiple interfaces (like X2, S1) Handover time Avg. time for handover between multiple interfaces (like X2, S1)
  • the experiment run on a VM with i7 processor, 16 core CPU and 16 GB DDR3 RAM. Evaluation model developed in python programming language. Machine learning algorithms like KNN, CBLOF, OCSVM, RF, XGBOOST and IF analyzed. After generating the labels using a Z-test as mentioned in the Normalization above, the business rules ae applied and the ground truth labels are refined. Table 2 depicts the comparison of different supervised models.
  • the present invention can be implemented on OAM (Operation Administration and Maintenance) - Virtualized radio access network (vRAN), Multi-access Edge Computing (MEC), Core as it is related to KPI behavior.
  • KPIs are applicable across all layers of the 3GPP RAN stack from Physical Layer (PHY), Medium Access Layer (MAC), Radio Link Control (RLC), Packet Data Convergence Control (PDCP), Radio resource management (RRM) and Non-Access Stratum (NAS) in both 4G and 5G as well as for CORE.
  • the KPI anomaly value threshold determination through correlation can be extended for any networks such as Fiber to the "x" (FTTX), Dynamic Circuit Network (DCN), Transport Networks, Software-Defined Networking (SDN), Cloud systems.
  • FTTX Fiber to the "x"
  • DCN Dynamic Circuit Network
  • SDN Software-Defined Networking
  • Figure 5 illustrates a flow diagram 500 depicting an exemplary embodiment of a KPI to log / alarm message correlation, in accordance with an embodiment of the present subject matter.
  • the behavior of each cell in mobile communications network is different from other cell.
  • KPI information obtained from different cells in the network will have different scales. For example, traffic across one cell can range up to few 80%, while another cell can have 2% traffic.
  • the data pre-processor module 111 may be configured to parse the log event, which may be extracted from each raw log message, to form a log sequence and to group log sequences into clusters.
  • the fm log file contains complete process, DB transaction related information and this extra information not required for parsing.
  • Raw Log will have complete details from FM process startup, request and response of each operation related to that process, debugging information, asynchronous notification (event) message and DB transaction related information.
  • the data processor module 111 may be configured to extract a representative for each cluster, which serves as the pattern of a group of similar log sequences.
  • the correlation module 119 may be further configured to identify clusters that highly correlate with KPI's changes by using supervised machine learning model such as multivariate linear regression (MLR) model to find the correlation between cluster sizes and KPI values over multiple time intervals. Further, each cluster represents one log sequence.
  • MLR multivariate linear regression
  • each cluster represents one log sequence.
  • the data processor module 111 may be configured to be implemented on alarms & events as well as their message structure follows the same pattern of log message i.e., timestamp & message body.
  • Figure 6 illustrates a flow diagram 600 depicting an exemplary embodiment of a method for a KPI to log / alarm message correlation, in accordance with an embodiment of the present subject matter.
  • method include extracting by the data pre-processor module 111, a plurality of log events from the KPI data.
  • Each log event of the plurality of log events corresponds to presence of information related to one of an alarm, an error, an exception, or an abnormality associated with a cell within the live network communication system.
  • the behavior of each cell in mobile communications network is different from other cell.
  • KPI information obtained from different cells in the network will have different scales. For example, traffic across one cell can range up to few 80%, while another cell can have 2% traffic.
  • There are multiple factors like number of users, climate conditions, geo-location, etc. which are directly related to range of a KPI.
  • There are multiple factors like number of users, climate conditions, geo-location, etc. which are directly related to range of a KPI.
  • the method include generating, by the data preprocessor module 111, a plurality of log sequences based on parsing and linking of a set of log events among the extracted plurality of log events that shares one of a same task or a process ID.
  • the method includes, clustering, by the data processor module 111, plurality of log sequences to generate a plurality of clusters each having a different type of sequence.
  • the data processor module 111 may be configured to extract a representative for each cluster, which serves as the pattern of a group of similar log sequences.
  • the method includes extracting, by the data processor module 111 information for each cluster of the plurality of clusters, wherein the representative information indicates a pattern of a group of similar log sequences.
  • the method includes, correlating, by the KPI correlation module 119, cluster sizes of the plurality of clusters with the plurality of KPI values over multiple time intervals based on a supervised machine learning mechanism such as multivariate linear regression (MLR) to find the correlation between cluster sizes and KPI values over multiple time intervals.
  • a supervised machine learning mechanism such as multivariate linear regression (MLR) to find the correlation between cluster sizes and KPI values over multiple time intervals.
  • MLR multivariate linear regression
  • the data processor module 111 may be configured to be implemented on alarms & events as well as their message structure follows the same pattern of log message i.e., timestamp & message body.
  • the method includes identifying, by the KPI correlation module based on the correlation, a set of clusters among the plurality of clusters that is highly correlated with KPI changes.
  • the method further include determining, in the extracted plurality of log events, by the KPI correlation module 119, presence of information associated with at least one of an occurrence of an alarm, an occurrence of an event, an observation of an exception, an observation of an error, or an abnormality of a geographical location of a cell within the live network communication system to decide the KPI threshold value and the presence of the information in the extracted plurality of log events is determined in the real time.
  • the KPI threshold at each of a plurality of time intervals at which the presence of the information is determined by the KPI correlation module 119.
  • Figure 7 illustrates tables 700 for showing a comparison between conventional way of threshold setting and the threshold setting and calculation using the proposed solution, according to an embodiment of the present subject matter.
  • the table 700a indicates that threshold value is predetermined and statically defined irrespective the NE/Cell behaviour as vary in nature.
  • the table 700b indicates the threshold values which are determined dynamically based on NE/Cell abnormality observed by Hardware alarms raised or Exception/error in the log observed.
  • the table 700b clearly shows that the threshold values are defined dynamically for each Cell's KPI.
  • FIG. 8 illustrates tables 800 for showing a comparison between conventional way of threshold setting and the threshold setting based on the granularity, according to an embodiment of the present subject matter.
  • the table 800a indicates that threshold value is predetermined and statically defined irrespective of the granularity.
  • the table 800a indicates the threshold values which are determined dynamically based on NE/Cell abnormality observed by H/W alarms raised or Exception/error in the log observed.
  • the table 800b clearly shows that the threshold values are defined dynamically for each cell's KPI.
  • the threshold values are determined dynamically based on its granularity (hourly level).
  • the 24 threshold values are defined on each KPI of the Cell.
  • the threshold value can be defined in more granularity as the NE/Cell behavior vary in nature with respect to time, its geo location etc.
  • the advantages of the present invention include that the operator does not need to manually set and change the thresholds based on network conditions as the NE/Cell behavior vary in nature. Thus, it will lead to a more accurate threshold as well as reduction in manual work. Further, the threshold values are learnt and dynamically decided based on the NE/Cell behavior and network abnormal scenarios such as H/W alarms, Exception/error in the log messages, granularity of the KPI traffic trend (hourly, daily, monthly) and the geographical location of the NE/Cell.
  • threshold values No predetermined, Static values taken
  • correlating and do causal analysis on the metrics results in drastic accuracy increase than the current manual and statistical ways
  • the present method and system correlates and do causal analysis on the metrics based on cell geographical location, alarm/events occurred and error/exception observed in the logs in real time. Further, the present disclosure catches the cell deterioration and take corrective actives before it is too late as deviation captured upfront due to dynamic assessment.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Signal Processing (AREA)
  • Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

La présente invention concerne un procédé et un système de détection de seuils dynamiques destinés à des indicateurs de rendements clés (IRC) en réseaux de communication. Un procédé de détection de seuils dynamiques consiste à surveiller des données reçues d'IRC correspondant à une pluralité de nœuds de réseau, les données d'IRC comprenant une pluralité de valeurs d'IRC de chaque nœud de la pluralité du nœuds de réseau. Le procédé consiste en outre à détecter au moins une valeur d'anomalie d'IRC parmi les données d'IRC, selon un modèle instruit d'apprentissage automatique (ML). Le procédé consiste ensuite à analyser une métrique d'IRC associée aux données d'IRC, selon la détection de la ou des valeurs d'anomalie d'IRC. Enfin, le procédé consiste à déterminer, sous forme de seuil d'IRC, une valeur spécifique d'anomalie parmi lesdites valeurs d'anomalie d'IRC détectées, selon l'analyse des métriques d'IRC.
PCT/KR2021/009653 2020-07-24 2021-07-26 Procédé et système de détection de seuils dynamiques pour indicateurs de rendements clés (irc) en réseaux de communication WO2022019728A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
IN202041031669 2020-07-24
IN202041031669 2021-07-19

Publications (1)

Publication Number Publication Date
WO2022019728A1 true WO2022019728A1 (fr) 2022-01-27

Family

ID=79730019

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/KR2021/009653 WO2022019728A1 (fr) 2020-07-24 2021-07-26 Procédé et système de détection de seuils dynamiques pour indicateurs de rendements clés (irc) en réseaux de communication

Country Status (1)

Country Link
WO (1) WO2022019728A1 (fr)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114661947A (zh) * 2022-03-15 2022-06-24 北京达佳互联信息技术有限公司 偏好差异检测方法、装置、电子设备及计算机存储介质
WO2023165685A1 (fr) * 2022-03-01 2023-09-07 Telefonaktiebolaget Lm Ericsson (Publ) Détection d'anomalie et classification d'anomalies avec cause racine
WO2024098331A1 (fr) * 2022-11-10 2024-05-16 Telefonaktiebolaget Lm Ericsson (Publ) Procédé et appareil d'analyse de données de performance de service

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040039968A1 (en) * 2000-09-11 2004-02-26 Kimmo Hatonen System, device and method for automatic anomally detection
WO2014183782A1 (fr) * 2013-05-14 2014-11-20 Nokia Solutions And Networks Oy Procédé et dispositif de réseau destinés à la détection des anomalies de cellules
US20190036789A1 (en) * 2017-07-31 2019-01-31 Accenture Global Solutions Limited Using machine learning to make network management decisions
US20190215230A1 (en) * 2018-01-08 2019-07-11 Cisco Technology, Inc. Analyzing common traits in a network assurance system
US20200092159A1 (en) * 2018-09-14 2020-03-19 Cisco Technology, Inc. Threshold selection for kpi candidacy in root cause analysis of network issues

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040039968A1 (en) * 2000-09-11 2004-02-26 Kimmo Hatonen System, device and method for automatic anomally detection
WO2014183782A1 (fr) * 2013-05-14 2014-11-20 Nokia Solutions And Networks Oy Procédé et dispositif de réseau destinés à la détection des anomalies de cellules
US20190036789A1 (en) * 2017-07-31 2019-01-31 Accenture Global Solutions Limited Using machine learning to make network management decisions
US20190215230A1 (en) * 2018-01-08 2019-07-11 Cisco Technology, Inc. Analyzing common traits in a network assurance system
US20200092159A1 (en) * 2018-09-14 2020-03-19 Cisco Technology, Inc. Threshold selection for kpi candidacy in root cause analysis of network issues

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023165685A1 (fr) * 2022-03-01 2023-09-07 Telefonaktiebolaget Lm Ericsson (Publ) Détection d'anomalie et classification d'anomalies avec cause racine
CN114661947A (zh) * 2022-03-15 2022-06-24 北京达佳互联信息技术有限公司 偏好差异检测方法、装置、电子设备及计算机存储介质
WO2024098331A1 (fr) * 2022-11-10 2024-05-16 Telefonaktiebolaget Lm Ericsson (Publ) Procédé et appareil d'analyse de données de performance de service

Similar Documents

Publication Publication Date Title
WO2022019728A1 (fr) Procédé et système de détection de seuils dynamiques pour indicateurs de rendements clés (irc) en réseaux de communication
US11811588B2 (en) Configuration management and analytics in cellular networks
US11018958B2 (en) Communication network quality of experience extrapolation and diagnosis
US11388040B2 (en) Automatic root cause diagnosis in networks
US9204329B2 (en) Distributed RAN information collection, consolidation and RAN-analytics
US10326640B2 (en) Knowledge base radio and core network prescriptive root cause analysis
Padmanabha Iyer et al. Mitigating the latency-accuracy trade-off in mobile data analytics systems
KR20140147872A (ko) 모바일 네트워크 성능 문제점들의 근본 원인 분석을 위한 시스템 및 방법
CN110062393B (zh) 一种网络差小区智能分析方法
CN114586420A (zh) 用于在通信网络中管理受损通信装置的方法和设备
Rizwan et al. A zero-touch network service management approach using AI-enabled CDR analysis
Chernogorov et al. N-gram analysis for sleeping cell detection in LTE networks
Gurbani et al. Detecting and predicting outages in mobile networks with log data
Slimen et al. Anomaly prevision in radio access networks using functional data analysis
CN108696376A (zh) 信令切换失败的故障排查方法和装置
Qian et al. Characterization of 3g data-plane traffic and application towards centralized control and management for software defined networking
US20230292156A1 (en) Trajectory based performance monitoring in a wireless communication network
Liu et al. KQis-driven QoE anomaly detection and root cause analysis in cellular networks
Muñoz et al. A method for identifying faulty cells using a classification tree-based UE diagnosis in LTE
CN116643937A (zh) 数据日志的图像分析
Shi et al. Towards automatic troubleshooting for user-level performance degradation in cellular services
Santos et al. An unsupervised learning approach for performance and configuration optimization of 4G networks
WO2022149149A1 (fr) Intelligence artificielle présentant un modèle causal dynamique pour analyse de défaillance dans un réseau de communication mobile
EP2989754B1 (fr) Système et procédé de prise de décision opérationnelle dynamique dans des réseaux d'accès radio sans fil de multiples types/couches sur la base d'une détection de flux de données progressif
Wu et al. A small cell outage prediction method based on RNN model

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21846101

Country of ref document: EP

Kind code of ref document: A1