CN111148142B - Dormant cell detection method based on anomaly detection and integrated learning - Google Patents

Dormant cell detection method based on anomaly detection and integrated learning Download PDF

Info

Publication number
CN111148142B
CN111148142B CN201911408236.8A CN201911408236A CN111148142B CN 111148142 B CN111148142 B CN 111148142B CN 201911408236 A CN201911408236 A CN 201911408236A CN 111148142 B CN111148142 B CN 111148142B
Authority
CN
China
Prior art keywords
cell
abnormal
data
key performance
corr
Prior art date
Legal status (The legal status 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 status listed.)
Active
Application number
CN201911408236.8A
Other languages
Chinese (zh)
Other versions
CN111148142A (en
Inventor
李秀华
明钊
孙川
袁传奇
范琪琳
王悦阳
唐永川
毛玉星
李剑
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chongqing University
Original Assignee
Chongqing University
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 Chongqing University filed Critical Chongqing University
Priority to CN201911408236.8A priority Critical patent/CN111148142B/en
Publication of CN111148142A publication Critical patent/CN111148142A/en
Application granted granted Critical
Publication of CN111148142B publication Critical patent/CN111148142B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/04—Arrangements for maintaining operational condition
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/24323—Tree-organised classifiers
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/06—Management of faults, events, alarms or notifications
    • H04L41/0677—Localisation of faults
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/142—Network analysis or design using statistical or mathematical methods
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Signal Processing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Algebra (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Physics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

The invention discloses a dormant cell detection method based on anomaly detection and integrated learning, which comprises the following steps: 1) screening out all cells with normal key performance index KPI data sequences, and reserving abnormal cells; 2) by marking the correlation index corr of the cell1Correlation index corr2And correlation index corr3As training data, training a random forest model to obtain a cell classification model M; 3) the correlation index corr of the abnormal cell1Correlation index corr2And correlation index corr3Inputting the test data into the cell classification model M, and outputting the abnormal reason of each abnormal cell as pattern1、pattern2… and patternnA probability matrix A of (2); 4) and if the maximum element a of the probability matrix A is not more than epsilon, the abnormal cell is a dormant cell. The invention can better position the dormant cell in the network, can improve the detection accuracy according to continuous practice, and has strong flexibility and stable performance.

Description

Dormant cell detection method based on anomaly detection and integrated learning
Technical Field
The invention relates to a network optimization technology in a mobile communication network, in particular to a dormant cell detection method based on anomaly detection and integrated learning in the mobile communication network.
Background
In mobile communication networks, detection of dormant cells has been an important research direction. The detection application range of the dormant cell is very wide, and the method plays an important role in positioning an abnormal cell, improving the network quality, improving the user perception and the like.
In the long-term evolution of the network, key indicator (KPI) data of some cells greatly changes relative to long-term historical data, the change exceeds a certain range, and the cells are positioned as abnormal cells. In addition, among these abnormal cells, the cause of abnormality in some cells is determined, and abnormality in other cells is caused for some unknown cause. In communication network optimization work. These cells that have generated anomalies and are unable to locate their cause of the anomalies are often referred to as dormant cells.
Currently, the detection of dormant cells is divided into two categories: one is completely based on abnormal cell detection, a normal interval of each index of each cell is dynamically defined by using historical data of each cell, whether real-time data exceeds the interval is judged, and if the real-time data exceeds the interval, the real-time data is directly marked as a dormant cell; the other type adds some new contents on the basis of the first type. Since the cause of the abnormality of some cells can be detected by the detector in the network, these cells do not belong to the dormant cells. Similarly, some human-operated interference causes cell abnormalities, and abnormalities caused by these known interferences also do not belong to dormant cells. Therefore, the second method increases the content of some service screening, and can more accurately position the dormant cell.
In the existing dormant cell detection method, all abnormal cells are marked as dormant cells by a dynamic threshold detection method. In actual production there are various cells that are out of order for known reasons and should not be marked as sleeping cells, and this approach does not take this into account. The second detection method considers the possibility, so that on the basis of the first detection method, a service screening step is added, the detection method of the dormant cell is improved to a certain extent, and the accuracy is improved. It is worth pointing out that in long-term network optimization, some abnormal cell abnormal causes can be given based on the existing optimization experience. These causes of abnormalities are often difficult to accurately calculate and, instead, they are often derived from expert optimization experience. For the cells with the reasons being clearly marked, the problem can be solved by aiming at the specific reasons, and for the dormant cells with unknown abnormal reasons, the problem is solved by restarting the corresponding base station.
Disclosure of Invention
The present invention is directed to solving the problems of the prior art.
The technical scheme adopted for realizing the aim of the invention is that the dormant cell detection method based on abnormal detection and integrated learning in the mobile communication network comprises the following steps:
1) acquiring key performance indicator KPI data sequences of all cells of a mobile communication network, and cleaning the key performance indicator KPI data sequences of each cell, wherein the steps are as follows:
1.1) based on the set granularity g, screening out cells with non-compliant granularity.
And 1.2) screening out cells with unconventional time length based on the set data length S which is T/g. And T is a dormant cell detection period.
1.3) fill the remaining cell loss value with 0.
1.4) based on the service alarm information, screening out the service alarm cell.
2) And establishing a historical data sequence and a test data sequence based on the cleaned key performance indicator KPI data sequence.
Further, the history data sequence is t0-t1A key performance indicator, KPI, data sequence for a time segment. The test data sequence is t2-t3Key performance indicator KPI data sequences for a time segment。t0The starting time of the key performance indicator KPI data sequence is shown. t is t1∈(t0,t3) The end time for the historical data sequence. t is t2∈(t0,t3) Test data sequence start time. t is t3The end time of the key performance index KPI data sequence is obtained.
3) And screening all cells with normal key performance indicator KPI data sequences based on the historical data sequences and the test data sequences, and reserving abnormal cells.
The steps for determining the abnormal cell are as follows:
and 3.1) differentiating the historical data sequence and the test data sequence by using a seasonal difference method. And inputting the differentiated historical data sequence and the differentiated test data sequence into an addition model to obtain a historical trend data set and a test trend data set.
Further, the addition model Y [ T ] + S [ T ] + e [ T ]. Wherein, T [ T ], ST, e [ T ] are key performance index KPI data after difference. Y [ t ] represents historical trend data or test trend data output by the additive model.
And 3.2) respectively calculating the mean value mu and the variance sigma of different key performance index KPI data sequences in different cells based on the historical trend data set, and determining the normal interval I of each key performance index KPI data.
Further, the normal interval I of the key performance indicator KPI data is (μ - ω × σ, μ + ω × σ). ω is a constant.
3.3) judging whether each element in the test trend data set is positioned in a normal interval I of the corresponding key performance indicator KPI data sequence, if not, marking the corresponding key performance indicator KPI data as an abnormal indicator, and marking a cell in which the abnormal indicator is positioned as an abnormal cell.
4) Acquiring abnormal reasons of the abnormal cells within h time, classifying the abnormal reasons and recording the classified reasons as patterns1,pattern2,…,patternn。
Further, the abnormal reasons include uplink interference, adjacent cell abnormality and line abnormality.
5) And respectively labeling the cell data corresponding to each type of abnormal reason to obtain labeled cell data and a corresponding labeled cell.
6) The average received signal strength AVG _ RSSI _ PUCCH (RSSI1) of the physical uplink control channel and the average received signal strength AVG _ RSSI _ PUSCH (RSSI2) of the physical uplink shared channel of the marked cell are selected, and the IFP index is calculated. IFP is max (AVG _ RSSI _ PUCCH (RSSI1), AVG _ RSSI _ PUSCH (RSSI 2)).
7) Calculating correlation indexes corr of IFP indexes of the abnormal cell and the marked cell and the average received signal strength AVG _ RSSI _ PUCCH (RSSI1) of the physical uplink control channel respectively1Correlation index corr between IFP index and average received signal strength AVG _ RSSI _ PUSCH (RSSI2) of physical uplink shared channel2Correlation index corr of average received signal strength AVG _ RSSI _ PUCCH (RSSI1) of physical uplink control channel and average received signal strength AVG _ RSSI _ PUSCH (RSSI2) of physical uplink shared channel3。
Correlation index corr1As follows:
Figure GDA0003668872130000031
correlation index corr2As follows:
Figure GDA0003668872130000032
correlation index corr3As follows:
Figure GDA0003668872130000033
in the formula, cov () is covariance and var () is variance.
8) By marking the correlation index corr of the cell1Correlation index corr2And correlation index corr3And as training data, training a random forest model to obtain a cell classification model M.
9) The correlation index corr of the abnormal cell1Correlation index corr2And correlation index corr3Inputting the test data into the cell classification model M, and outputting the abnormal reason of each abnormal cell as pattern1、pattern2… and patternnThe probability matrix a of (a).
10) And if the maximum element a of the probability matrix A is not more than epsilon, the abnormal cell is a dormant cell. Epsilon is a preset probability threshold.
It is worth to be noted that, in order to exert the advantages of the dynamic threshold method and the method based on dynamic threshold improvement and make up the disadvantages of the dynamic threshold method and the method, the invention trains the model by using abnormal cell data in long-term evolution and an ensemble learning method and outputs an available random forest classifier. Meanwhile, the invention calculates the dynamic threshold value of the index for the historical data of each KPI of each cell in the network, finds out abnormal cells based on the threshold value and screens out the cells with service alarm. Then, for the rest abnormal cells, the invention uses a trained random forest classifier to output the probability matrix of each cell belonging to the labeled abnormal class, and labels the abnormal cells which do not obviously belong to any one class as dormant cells.
The technical effect of the invention is undoubted. The invention solves the problem that the existing dynamic threshold detection method and the improvement method thereof can not accurately position the abnormal cell with unknown reason, greatly improves the detection accuracy of the dormant cell, can better position the dormant cell in the network, can improve the detection accuracy according to continuous practice, and has strong flexibility and stable performance.
Drawings
Fig. 1 is a schematic diagram illustrating a process of detecting a dormant cell in a network.
Fig. 2 is a specific data flow diagram of the dormant cell detection system.
Detailed Description
The present invention is further illustrated by the following examples, but it should not be construed that the scope of the above-described subject matter is limited to the following examples. Various substitutions and alterations can be made without departing from the technical idea of the invention and the scope of the invention is covered by the present invention according to the common technical knowledge and the conventional means in the field.
Example 1:
referring to fig. 1 to 2, a dormant cell detection method based on anomaly detection and integrated learning in a mobile communication network includes the following steps:
1) acquiring key performance indicator KPI data sequences of all cells of a mobile communication network, and cleaning the key performance indicator KPI data sequences of each cell, wherein the steps are as follows:
1.1) based on the set granularity g, screening out the cells with the unqualified granularity.
And 1.2) screening out cells with unconventional time length based on the set data length S which is T/g. And T is a dormant cell detection period.
1.3) fill the remaining cell absence value with 0.
1.4) based on the service alarm information, screening out the service alarm cell.
2) And establishing a historical data sequence and a test data sequence based on the cleaned key performance index KPI data sequence.
The history data sequence is t0-t1A key performance indicator, KPI, data sequence for a time segment. The test data sequence is t2-t3A key performance indicator, KPI, data sequence for a time segment. t is t0The starting time of the key performance indicator KPI data sequence is shown. t is t1∈(t0,t3) The end time for the historical data sequence. t is t2∈(t0,t3) Is the test data sequence start time. t is t3The end time of the key performance index KPI data sequence is obtained.
3) And screening all cells with normal key performance indicator KPI data sequences based on the historical data sequences and the test data sequences, and reserving abnormal cells.
The steps for determining the abnormal cell are as follows:
and 3.1) differentiating the historical data sequence and the test data sequence by using a seasonal difference method. And inputting the differentiated historical data sequence and the differentiated test data sequence into an addition model to obtain a historical trend data set and a test trend data set.
Further, the addition model Y [ T ] + S [ T ] + e [ T ]. Wherein, T [ T ], ST, e [ T ] are key performance index KPI data after difference. Y [ t ] represents historical trend data or test trend data output by the additive model.
And 3.2) respectively calculating the mean value mu and the variance sigma of different key performance index KPI data sequences in different cells based on the historical trend data set, and determining the normal interval I of each key performance index KPI data.
The normal interval I of the key performance indicator KPI data is (μ - ω × σ, μ + ω × σ). Omega is a constant and can be specifically set according to parameters such as geographic environment, base station setting and the like.
3.3) judging whether each element in the test trend data set is positioned in a normal interval I of the corresponding key performance indicator KPI data sequence, if not, marking the corresponding key performance indicator KPI data as an abnormal indicator, and marking a cell in which the abnormal indicator is positioned as an abnormal cell.
4) In the long term evolution, the cells which are abnormal in the network and can locate the abnormal reasons are collected, the abnormal reasons of the abnormal cells in the h time are obtained, the abnormal reasons are classified and recorded as pattern1,pattern2,…,patternn。
The abnormal reasons comprise uplink interference, adjacent cell abnormity and line abnormity.
5) And respectively labeling the cell data corresponding to each type of abnormal reason to obtain labeled cell data and a corresponding labeled cell.
6) The average received signal strength AVG _ RSSI _ PUCCH (RSSI1) of the physical uplink control channel and the average received signal strength AVG _ RSSI _ PUSCH (RSSI2) of the physical uplink shared channel of the marked cell are selected, and the IFP index is calculated. IFP is max (AVG _ RSSI _ PUCCH (RSSI1), AVG _ RSSI _ PUSCH (RSSI 2)).
7) Calculating correlation of IFP indexes of abnormal cell and marked cell with average received signal strength AVG _ RSSI _ PUCCH (RSSI1) of physical uplink control channelIndex of Property corr1Correlation index corr between the IFP index and the average received signal strength AVG _ RSSI _ PUSCH (RSSI2) of the physical uplink shared channel2Correlation index corr of AVG _ RSSI _ PUCCH (RSSI1) of physical uplink control channel and AVG _ RSSI _ PUSCH (RSSI2) of physical uplink shared channel3。
Correlation index corr1As follows:
Figure GDA0003668872130000061
correlation index corr2As follows:
Figure GDA0003668872130000062
correlation index corr3As follows:
Figure GDA0003668872130000063
in the equation, cov (IFP, AVG _ RSSI _ PUCCH (RSSI1)) is a covariance of IFP and AVG _ RSSI _ PUCCH (RSSI1), and var (IFP) is an IFP variance.
8) By marking the correlation index corr of the cell1Correlation index corr2And correlation index corr3And as training data, training a random forest model to obtain a cell classification model M.
The method for establishing the cell classification model M comprises the following steps: a random forest model is obtained by establishing a plurality of decision trees for autonomous sampling, namely the random forest model is a cell classification model.
The steps of establishing the decision tree of the autonomous sampling are as follows:
I) through a self-service sampling method, a plurality of sample cell correlation index data are selected from all sample cells contained in a sample whole network cell data set in a mode of putting back random selection, and a decision tree is generated by utilizing the selected cell correlation index data for training.
II) when each node of the decision tree needs to be split, randomly selecting N correlation indexes from the N correlation indexes contained in the sample whole-network cell data set, wherein N is less than N, then using the selected N variables as subsets to be assigned to each sub-node, and selecting the one-dimensional characteristics with the best classification effect from the N variables as the classification attributes of the nodes.
9) The correlation index corr of the abnormal cell1Correlation index corr2And correlation index corr3Inputting the test data into the cell classification model M, and outputting the abnormal reason of each abnormal cell as pattern1、pattern2… and patternnThe probability matrix a.
Classifying the corresponding indexes of the test data by using a cell classification model M so as to obtain a probability matrix A, wherein the probability matrix A comprises the following steps:
9.1) inputting the data of the correlation indexes corresponding to the cell to be tested into a cell classification model M for processing, thereby obtaining the similarity matrix of the corresponding cell everywhere;
and 9.2) performing dimension reduction treatment, and performing dimension reduction on the similarity matrix of the corresponding cell by using a multi-dimensional scale analysis method, wherein the matrix obtained by dimension reduction is a probability matrix A.
10) And if the maximum element a of the probability matrix A is not more than epsilon, the abnormal cell is a dormant cell. Epsilon is a preset probability threshold.
Example 2:
the dormant cell detection method based on abnormal detection and integrated learning in the mobile communication network comprises the following steps:
1) acquiring key performance indicator KPI data sequences of all cells of a mobile communication network, and cleaning the key performance indicator KPI data sequences of each cell;
2) establishing a historical data sequence and a test data sequence based on the cleaned key performance index KPI data sequence;
3) based on the historical data sequence and the test data sequence, screening out all cells with normal key performance index KPI data sequences, and reserving abnormal cells;
4) acquiring abnormal reasons of the abnormal cells within h time, classifying the abnormal reasons and recording the classified reasons as patterns1,pattern2,…,patternn；
5) Respectively labeling the cell data corresponding to each type of abnormal reason to obtain labeled cell data and a corresponding labeled cell;
6) selecting the average received signal strength AVG _ RSSI _ PUCCH (RSSI1) of the physical uplink control channel of the marked cell and the average received signal strength AVG _ RSSI _ PUSCH (RSSI2) of the physical uplink shared channel, and calculating an IFP index; IFP ═ max (AVG _ RSSI _ PUCCH (RSSI1), AVG _ RSSI _ PUSCH (RSSI 2));
7) calculating correlation indexes corr of IFP indexes of the abnormal cell and the marked cell and the average received signal strength AVG _ RSSI _ PUCCH (RSSI1) of the physical uplink control channel respectively1Correlation index corr between IFP index and average received signal strength AVG _ RSSI _ PUSCH (RSSI2) of physical uplink shared channel2Correlation index corr of average received signal strength AVG _ RSSI _ PUCCH (RSSI1) of physical uplink control channel and average received signal strength AVG _ RSSI _ PUSCH (RSSI2) of physical uplink shared channel3；
8) To mark the correlation index corr of the cell1Correlation index corr2And correlation index corr3As training data, training a random forest model to obtain a cell classification model M;
9) the correlation index corr of the abnormal cell1Correlation index corr2And correlation index corr3Inputting the test data into the cell classification model M, and outputting the abnormal reason of each abnormal cell as pattern1、pattern2… and patternnA probability matrix A of (2);
10) if the maximum element a of the probability matrix A is not more than epsilon, the abnormal cell is a dormant cell; epsilon is a preset probability threshold.
Example 3:
the detection method of the dormant cell based on the abnormal detection and the integrated learning in the mobile communication network comprises the following steps of embodiment 2:
1) and differentiating the historical data sequence and the test data sequence by using a seasonal difference method. And inputting the differentiated historical data sequence and the differentiated test data sequence into an addition model to obtain a historical trend data set and a test trend data set.
Further, the addition model Y [ T ] + S [ T ] + e [ T ]. Wherein, T [ T ], ST, e [ T ] are key performance index KPI data after difference. Y [ t ] represents historical trend data or test trend data output by the additive model.
2) Based on the historical trend data set, respectively calculating the mean value mu and the variance sigma of different key performance index KPI data sequences in different cells, and determining the normal interval I of each key performance index KPI data.
Further, the normal interval I of the key performance indicator KPI data is (μ - ω × σ, μ + ω × σ). Omega is a constant and can be specifically set according to parameters such as geographic environment, base station setting and the like.
3) And judging whether each element in the test trend data set is positioned in a normal interval I of the corresponding key performance indicator KPI data sequence, if not, marking the corresponding key performance indicator KPI data as an abnormal indicator, and marking a cell where the abnormal indicator is positioned as an abnormal cell.

Claims (7)

1. The dormant cell detection method based on anomaly detection and ensemble learning is characterized by comprising the following steps of:
1) acquiring key performance indicator KPI data sequences of all cells of a mobile communication network, and cleaning the key performance indicator KPI data sequences of each cell;
2) establishing a historical data sequence and a test data sequence based on the cleaned key performance index KPI data sequence;
3) based on the historical data sequence and the test data sequence, screening out all cells with normal key performance index KPI data sequences, and reserving abnormal cells;
the steps for determining the abnormal cell are as follows:
3.1) differentiating the historical data sequence and the test data sequence by using a seasonal difference method; inputting the differentiated historical data sequence and the differentiated test data sequence into an addition model to obtain a historical trend data set and a test trend data set;
3.2) respectively calculating the mean value mu and the variance sigma of different key performance index KPI data sequences in different cells based on a historical trend data set, and determining the normal interval I of each key performance index KPI data;
3.3) judging whether each element in the test trend data set is positioned in a normal interval I of a corresponding key performance indicator KPI data sequence, if not, marking the corresponding key performance indicator KPI data as an abnormal indicator, and marking a cell where the abnormal indicator is positioned as an abnormal cell;
4) acquiring abnormal reasons of the abnormal cells within h time, classifying the abnormal reasons and recording the classified reasons as patterns1,pattern2,…,patternn；
5) Respectively labeling the cell data corresponding to each type of abnormal reason to obtain labeled cell data and a corresponding labeled cell;
6) selecting the average received signal strength AVG _ RSSI _ PUCCH (RSSI1) of the physical uplink control channel of the marked cell and the average received signal strength AVG _ RSSI _ PUSCH (RSSI2) of the physical uplink shared channel, and calculating an IFP index; IFP ═ max (AVG _ RSSI _ PUCCH (RSSI1), AVG _ RSSI _ PUSCH (RSSI 2));
7) calculating correlation indexes corr of IFP indexes of the abnormal cell and the marked cell and the average received signal strength AVG _ RSSI _ PUCCH (RSSI1) of the physical uplink control channel1Correlation index corr between the IFP index and the average received signal strength AVG _ RSSI _ PUSCH (RSSI2) of the physical uplink shared channel2Correlation index corr of average received signal strength AVG _ RSSI _ PUCCH (RSSI1) of physical uplink control channel and average received signal strength AVG _ RSSI _ PUSCH (RSSI2) of physical uplink shared channel3；
8) To mark the correlation index cor of the cellr1Correlation index corr2And correlation index corr3As training data, training a random forest model to obtain a cell classification model M;
9) the correlation index corr of the abnormal cell1Correlation index corr2And correlation index corr3Inputting the test data into the cell classification model M, and outputting the abnormal reason of each abnormal cell as pattern1、pattern2… and patternnA probability matrix A of (2);
10) if the maximum element a of the probability matrix A is not more than epsilon, the abnormal cell is a dormant cell; epsilon is a preset probability threshold.
2. The dormant cell detection method based on anomaly detection and ensemble learning according to claim 1, wherein the step of performing data cleaning on the Key Performance Indicator (KPI) data sequence of each cell is as follows:
1) based on the set granularity g, screening out cells with unqualified granularity;
2) screening out cells with unconventional time length based on the set data length S being T/g; t is a dormant cell detection period;
3) filling the remaining cell absence values with 0;
4) and screening out the service alarm cell based on the service alarm information.
3. The dormant cell detection method based on anomaly detection and ensemble learning of claim 1, wherein: the historical data sequence is t0-t1Key performance indicator KPI data sequences of time segments; the test data sequence is t2-t3Key performance indicator KPI data sequences of time segments; t is t0Starting time of key performance index KPI data sequence; t is t1∈(t0,t3) The end time of the historical data sequence; t is t2∈(t0,t3) Is the test data sequence start time; t is t3The end time of the key performance index KPI data sequence is obtained.
4. The dormant cell detection method based on anomaly detection and ensemble learning of claim 1, wherein: the addition model Y [ T ] + S [ T ] + e [ T ]; wherein, T [ T ], ST, e [ T ] are key performance index KPI data after difference; y [ t ] represents historical trend data or test trend data output by the additive model.
5. The dormant cell detection method based on anomaly detection and ensemble learning of claim 1, wherein: the normal interval I of the key performance indicator KPI data is (μ - ω × σ, μ + ω × σ); ω is a constant.
6. The dormant cell detection method based on anomaly detection and ensemble learning according to claim 1 or 2, characterized in that: the abnormal reasons comprise uplink interference, adjacent cell abnormity and line abnormity.
7. The abnormal detection and integrated learning-based dormant cell detection method according to claim 1, wherein a correlation index corr1As follows:
Figure FDA0003668872120000031
correlation index corr2As follows:
Figure FDA0003668872120000032
correlation index corr3As follows:
Figure FDA0003668872120000033
in the formula, cov () is covariance and var () is variance.
CN201911408236.8A 2019-12-31 2019-12-31 Dormant cell detection method based on anomaly detection and integrated learning Active CN111148142B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911408236.8A CN111148142B (en) 2019-12-31 2019-12-31 Dormant cell detection method based on anomaly detection and integrated learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911408236.8A CN111148142B (en) 2019-12-31 2019-12-31 Dormant cell detection method based on anomaly detection and integrated learning

Publications (2)

Publication Number Publication Date
CN111148142A CN111148142A (en) 2020-05-12
CN111148142B true CN111148142B (en) 2022-07-12

Family

ID=70522360

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911408236.8A Active CN111148142B (en) 2019-12-31 2019-12-31 Dormant cell detection method based on anomaly detection and integrated learning

Country Status (1)

Country Link
CN (1) CN111148142B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113379176B (en) * 2020-03-09 2023-12-19 中国移动通信集团设计院有限公司 Method, device, equipment and readable storage medium for detecting abnormal data of telecommunication network
WO2022013635A1 (en) * 2020-07-16 2022-01-20 Rakuten Symphony Singapore Pte. Ltd. Detecting of sleeping cell in a mobile network
CN113515678A (en) * 2021-05-13 2021-10-19 上海梯之星信息科技有限公司 Abnormal data screening method
CN113556773A (en) * 2021-06-04 2021-10-26 杭州红岭通信息科技有限公司 Method for detecting sleeping cell based on AI model

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101594622A (en) * 2009-06-15 2009-12-02 上海华为技术有限公司 The detection method of dormant cell, device and system
CN102176780A (en) * 2009-06-15 2011-09-07 上海华为技术有限公司 Dormant cell detection method, device and system
WO2014094591A1 (en) * 2012-12-17 2014-06-26 电信科学技术研究院 Cell sleeping, sleeping information transmitting and processing method and device
CN105072060A (en) * 2015-07-03 2015-11-18 京信通信技术(广州)有限公司 DTX determination method and apparatus of physical uplink control channel, and base station

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9204315B2 (en) * 2011-06-20 2015-12-01 Alcatel Lucent Method of coordinating fault detection responses by access nodes of a network
US9374796B2 (en) * 2014-10-10 2016-06-21 Qualcomm Incorporated Channel structure for a cellular internet of things system

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101594622A (en) * 2009-06-15 2009-12-02 上海华为技术有限公司 The detection method of dormant cell, device and system
CN102176780A (en) * 2009-06-15 2011-09-07 上海华为技术有限公司 Dormant cell detection method, device and system
WO2014094591A1 (en) * 2012-12-17 2014-06-26 电信科学技术研究院 Cell sleeping, sleeping information transmitting and processing method and device
CN105072060A (en) * 2015-07-03 2015-11-18 京信通信技术(广州)有限公司 DTX determination method and apparatus of physical uplink control channel, and base station

Also Published As

Publication number Publication date
CN111148142A (en) 2020-05-12

Similar Documents

Publication Publication Date Title
CN111148142B (en) Dormant cell detection method based on anomaly detection and integrated learning
Rajendran et al. SAIFE: Unsupervised wireless spectrum anomaly detection with interpretable features
CN104683984B (en) The real-time monitoring process method of wireless communication signals and system
CN114358152A (en) Intelligent power data anomaly detection method and system
CN110062410B (en) Cell interruption detection positioning method based on self-adaptive resonance theory
KR20170026228A (en) Identification using spectroscopy
CN109495327B (en) User activity anomaly detection and flow prediction method based on big data analysis
CN110493720B (en) Terminal positioning method and device and storage medium
CN107395301A (en) A kind of frequency spectrum sensing method based on K mean algorithms
CN112188532A (en) Training method of network anomaly detection model, network detection method and device
CN107809766B (en) Method and device for generating machine learning sample for network optimization
CN112784690B (en) Broadband signal parameter estimation method based on deep learning
CN116308958A (en) Carbon emission online detection and early warning system and method based on mobile terminal
CN114019139B (en) Method for detecting heavy metal abnormal data of agricultural land soil
CN111314113A (en) Internet of things node fault detection method and device, storage medium and computer equipment
CN111065106B (en) Index mutation cell detection method based on anomaly detection and kernel density estimation KDE in mobile communication network
EP3849231B1 (en) Configuration of a communication network
KR102014234B1 (en) Method and Apparatus for automatic analysis for Wireless protocol
Saffar et al. Mobile user environment detection using deep learning based multi-output classification
FI129551B (en) Analyzing operation of communications network
CN111798237B (en) Abnormal transaction diagnosis method and system based on application log
CN111385116B (en) Multidimensional correlation feature analysis method and device for high-interference cells
CN112398706B (en) Data evaluation standard determining method and device, storage medium and electronic equipment
CN109993556B (en) User behavior analysis method and device, computing equipment and storage medium
US20200026983A1 (en) Method and apparatus for signal matching

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant