Method and device for detecting abnormal state of cell

By dimensionality reduction and clustering processing of key performance indicators of cells in mobile communication networks, abnormal state evaluation information is generated, and the problem of inaccurate cell load status judgment in the prior art is solved, and more efficient detection and optimization are achieved.

CN114079957BActive Publication Date: 2024-11-12CHINA MOBILE COMM LTD RES INST +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010794979.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-08-10
Publication Date
2024-11-12
Estimated Expiration
2040-08-10

AI Technical Summary

Technical Problem

The determination of cell load status in prior art in mobile communication networks depends on manual experience, resulting in inaccurate judgment results and large workloads, and possible missed cells that are not judged to be high load.

Method used

The self-encoder is used to reduce the dimensionality of the cell's key performance indicators (KPI) parameters, and generate abnormal state evaluation information through clustering algorithms, simplifying the detection of network abnormal load states and improving the accuracy of detection results.

Benefits of technology

The calculation amount of subsequent classification processing is simplified, detection efficiency is improved, cell abnormal state can be more accurately identified, and the dependence on manual judgment is reduced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114079957B_ABST
    Figure CN114079957B_ABST
Patent Text Reader

Abstract

The present invention provides a method and device for detecting abnormal state of a cell, the method comprising: obtaining at least one key performance indicator KPI parameter group of a first cell collected within a preset time period, the KPI parameter group comprising values ​​of multiple KPI parameters collected at corresponding sampling time points; performing dimensionality reduction processing on each KPI parameter group of the first cell respectively, generating at least one first dimensionality reduction vector of the first cell; inputting the at least one first dimensionality reduction vector into an evaluator, generating evaluation information of the abnormal state of the first cell. The present invention can simplify the detection of abnormal load state of a network and improve the accuracy of the detection result.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of mobile communication technology, and in particular to a method and device for detecting abnormal state of a cell. Background Art

[0002] When performing network operation and maintenance, mobile communication network operators usually define multiple high-load related statistical data that need to be counted when the cell is busy when the network is under high load, and artificially divide the threshold. The time period that meets the threshold condition is judged as the high-load period. For example, if a cell in the existing network is under high load for more than three days a week during the busy period, a work order will be issued and assigned to the network optimization engineer of the existing network for processing. This threshold-based judgment and processing of cell load is relatively rough, requiring network optimization engineers to conduct continuous observation for a period of time, and combine multiple parameters such as key performance indicator (KPI) parameters and neighboring cell information for manual judgment. It may also miss cells that are fully loaded in certain parameters but are not judged to be under high load.

[0003] On the other hand, when optimizing high-load cells, existing network engineers usually judge the type of high load and find the cause of the high load based on their own experience. The above process is still mainly completed by analyzing KPI parameters. However, since there are many KPI parameters in the cell, even if the secondary parameters are simplified, the joint state space of their mapping is huge, which is difficult for engineers to handle. Therefore, existing network engineers generally only track a few commonly used parameters and optimize them in combination with other empirical methods. The workload of the above network optimization solutions is usually very large, and the actual effect is also limited by expert experience. Summary of the invention

[0004] At least one embodiment of the present invention provides a method, a terminal and a network device for detecting abnormal cell status, which are used to simplify the detection of abnormal network load status and improve the accuracy of the detection result.

[0005] According to one aspect of the present invention, at least one embodiment provides a method for detecting an abnormal state of a cell, including:

[0006] Acquire at least one key performance indicator KPI parameter group of the first cell collected within a preset time period, where the KPI parameter group includes values ​​of multiple KPI parameters collected at corresponding sampling time points;

[0007] Performing dimensionality reduction processing on each KPI parameter group of the first cell respectively to generate at least one first dimensionality reduction vector of the first cell;

[0008] The at least one first dimension reduction vector is input into an evaluator to generate evaluation information of the abnormal state of the first cell.

[0009] Optionally, performing dimensionality reduction processing on each KPI parameter group of the first cell respectively includes:

[0010] Each KPI parameter group of the first cell is input into a pre-trained autoencoder to obtain a first dimension reduction vector output by the autoencoder.

[0011] Optionally, the evaluator is used to perform clustering processing on the at least one first dimensionality reduction vector, determine and output the first category to which the at least one first dimensionality reduction vector belongs and its corresponding labeling information, wherein the first category is one of multiple pre-generated categories, and the labeling information is used to indicate the abnormal status of the cell of the corresponding category.

[0012] Optionally, before acquiring at least one KPI parameter group of the first cell, the method further includes:

[0013] Acquire multiple KPI parameter groups of multiple cells collected at multiple sampling time points, wherein each KPI parameter group includes the multiple KPI parameters collected by the same cell at a sampling time point;

[0014] Using the multiple KPI parameter groups, an autoencoder is trained until a preset training end condition is reached; wherein the autoencoder includes an encoder and a decoder, and the encoder and the decoder both include multi-layer neural networks.

[0015] Optionally, the method further includes:

[0016] Obtaining multiple dimensionality reduction vectors obtained by inputting the multiple KPI parameter groups into the trained autoencoder;

[0017] Clustering is performed on the multiple dimension reduction vectors to obtain multiple categories.

[0018] Optionally, after obtaining the multiple categories, the method further includes:

[0019] For each category, perform the following processing:

[0020] Selecting a parameter group belonging to the category from the multiple KPI parameter groups, training a classifier using the selected parameter group, and calculating the weight of each KPI parameter in the parameter group according to the training result;

[0021] Repeat the following steps until no KPI parameter exists in the parameter group: select a target KPI parameter with the largest weight from the remaining KPI parameters in the parameter group, add it to a parameter set, and delete the target KPI parameter and KPI parameters whose correlation with the target KPI parameter is greater than a preset threshold from the parameter group;

[0022] Selecting a preset number of KPI parameters from the parameter set in descending order of weight as the KPI parameters of the category, and calculating the deviation of the statistical features of the selected KPI parameters in the category relative to the statistical features in all samples;

[0023] The KPI parameters, weights and deviations of the category are used as feature outputs of the category;

[0024] The annotation information input for the category is received, and a corresponding relationship between the annotation information and the category is established.

[0025] Optionally, the annotation information includes at least one of the following information: an evaluation value of the cell abnormality level, an evaluation value of the cell load level, an evaluation value of the cell optimization level, and remark information.

[0026] Optionally, the number of the multiple categories is Wherein, N is the number of the multiple KPI parameter groups.

[0027] Optionally, determining the first category to which the at least one first dimensionality reduction vector belongs and its corresponding labeling information includes:

[0028] Performing clustering processing on the at least one first reduced dimensionality vector, determining a first center point of the at least one first reduced dimensionality vector, calculating the Euclidean distance between the first center point and center points of each category, and selecting a category corresponding to the shortest Euclidean distance as the first category to which the at least one first reduced dimensionality vector belongs;

[0029] The labeling information corresponding to the first category is determined according to a pre-established correspondence between labeling information and categories.

[0030] Optionally, before the step of performing clustering processing on the at least one first dimensionality reduction vector, the method further includes:

[0031] Outlier detection is performed on the at least one first dimensionality reduction vector to remove outliers therein to obtain an updated first dimensionality reduction vector.

[0032] According to another aspect of the present invention, at least one embodiment provides a device for detecting abnormal state of a cell, including:

[0033] A parameter acquisition module, used to acquire at least one key performance indicator KPI parameter group of the first cell collected within a preset time period, wherein the KPI parameter group includes values ​​of multiple KPI parameters collected at corresponding sampling time points;

[0034] A dimensionality reduction processing module, used to perform dimensionality reduction processing on each KPI parameter group of the first cell respectively, and generate at least one first dimensionality reduction vector of the first cell;

[0035] The labeling processing module is used to input the at least one first dimensionality reduction vector into an evaluator to generate evaluation information of the abnormal state of the first cell.

[0036] Optionally, the dimensionality reduction processing module is further used to input each KPI parameter group of the first cell into a pre-trained autoencoder to obtain a first dimensionality reduction vector output by the autoencoder.

[0037] Optionally, the labeling processing module is also used to perform clustering processing on the at least one first dimensionality reduction vector through the evaluator, determine and output the first category to which the at least one first dimensionality reduction vector belongs and its corresponding labeling information, wherein the first category is one of multiple pre-generated categories, and the labeling information is used to indicate the abnormal status of the cell of the corresponding category.

[0038] Optionally, the detection device further includes:

[0039] An autoencoder training module is used to obtain multiple KPI parameter groups of multiple cells collected at multiple sampling time points before obtaining at least one KPI parameter group of the first cell, wherein each KPI parameter group includes the multiple KPI parameters collected by the same cell at a sampling time point; using the multiple KPI parameter groups, training an autoencoder until a preset training end condition is reached; wherein the autoencoder includes an encoder and a decoder, and the encoder and the decoder both include multi-layer neural networks.

[0040] Optionally, the detection device further includes:

[0041] A clustering module is used to obtain a plurality of reduced-dimensionality vectors obtained after inputting the plurality of KPI parameter groups into the trained autoencoder; and cluster the plurality of reduced-dimensionality vectors to obtain a plurality of categories.

[0042] Optionally, the detection device further includes:

[0043] The labeling information establishment module is used to perform the following processing for each category after obtaining multiple categories:

[0044] Selecting a parameter group belonging to the category from the multiple KPI parameter groups, training a classifier using the selected parameter group, and calculating the weight of each KPI parameter in the parameter group according to the training result;

[0045] Repeat the following steps until no KPI parameter exists in the parameter group: select a target KPI parameter with the largest weight from the remaining KPI parameters in the parameter group, add it to a parameter set, and delete the target KPI parameter and KPI parameters whose correlation with the target KPI parameter is greater than a preset threshold from the parameter group;

[0046] Selecting a preset number of KPI parameters from the parameter set in descending order of weight as the KPI parameters of the category, and calculating the deviation of the statistical features of the selected KPI parameters in the category relative to the statistical features in all samples;

[0047] The KPI parameters, weights and deviations of the category are used as feature outputs of the category;

[0048] The annotation information input for the category is received, and a corresponding relationship between the annotation information and the category is established.

[0049] Optionally, the labeling processing module is also used to perform clustering processing on the at least one first dimensionality reduction vector, determine the first center point of the at least one first dimensionality reduction vector, calculate the Euclidean distance between the first center point and the center points of each category, and select the category corresponding to the shortest Euclidean distance as the first category to which the at least one first dimensionality reduction vector belongs; and determine the labeling information corresponding to the first category based on the pre-established correspondence between the labeling information and the category.

[0050] Optionally, the detection device further includes:

[0051] The outlier removal module is used to perform outlier detection on the at least one first dimensionality reduction vector before clustering the at least one first dimensionality reduction vector, remove outliers therein, and obtain an updated first dimensionality reduction vector.

[0052] According to another aspect of the present invention, at least one embodiment provides a device for detecting abnormal cell status, including a transceiver and a processor, wherein:

[0053] The transceiver is used to obtain at least one key performance indicator KPI parameter group of the first cell collected within a preset time period, wherein the KPI parameter group includes values ​​of multiple KPI parameters collected at corresponding sampling time points;

[0054] The processor is used to perform dimensionality reduction processing on each KPI parameter group of the first cell respectively to generate at least one first dimensionality reduction vector of the first cell; input the at least one first dimensionality reduction vector into the evaluator to generate evaluation information of the abnormal state of the first cell.

[0055] According to another aspect of the present invention, at least one embodiment provides a computer-readable storage medium having a program stored thereon, and when the program is executed by a processor, the steps of the method described above are implemented.

[0056] Compared with the prior art, the method and device for detecting abnormal cell status provided by the embodiments of the present invention can simplify the amount of computation required for subsequent classification processing, simplify the detection of abnormal network load status, and improve detection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0058] Figure 1 A schematic diagram of an application scenario of an embodiment of the present invention;

[0059] Figure 2 A flowchart of a method for detecting abnormal cell status provided by an embodiment of the present invention;

[0060] Figure 3 A schematic diagram of the overall process of a method for detecting abnormal cell status provided by an embodiment of the present invention;

[0061] Figure 4 A schematic diagram of the structure of an autoencoder provided in an embodiment of the present invention;

[0062] Figure 5 A schematic diagram of a structure of a cell abnormal state detection device provided by an embodiment of the present invention;

[0063] Figure 6 Another structural schematic diagram of a cell abnormal status detection device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0064] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to enable the scope of the present invention to be fully communicated to those skilled in the art.

[0065] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and need not be used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable in appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, the process, method, system, product or equipment comprising a series of steps or units need not be limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment. "And / or" in the specification and claims represents at least one of the connected objects.

[0066] The technology described herein is not limited to NR systems and Long Time Evolution (LTE) / LTE-Advanced (LTE-A) systems, and can also be used for various wireless communication systems, such as Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal Frequency Division Multiple Access (OFDMA), Single-carrier Frequency Division Multiple Access (SC-FDMA) and other systems. The terms "system" and "network" are often used interchangeably. A CDMA system can implement radio technologies such as CDMA2000, Universal Terrestrial Radio Access (UTRA). UTRA includes Wideband Code Division Multiple Access (WCDMA) and other CDMA variants. A TDMA system can implement radio technologies such as Global System for Mobile Communication (GSM). OFDMA systems can implement radio technologies such as Ultra Mobile Broadband (UMB), Evolution-UTRA (E-UTRA), IEEE 802.21 (Wi-Fi), IEEE 802.16 (WiMAX), IEEE 802.20, Flash-OFDM, etc. UTRA and E-UTRA are parts of Universal Mobile Telecommunications System (UMTS). LTE and more advanced LTE (such as LTE-A) are new UMTS versions that use E-UTRA. UTRA, E-UTRA, UMTS, LTE, LTE-A, and GSM are described in documents from an organization named "3rd Generation Partnership Project" (3GPP). CDMA2000 and UMB are described in documents from an organization named "3rd Generation Partnership Project 2" (3GPP2).The techniques described herein may be used for the systems and radio technologies mentioned above as well as for other systems and radio technologies. However, the following description describes an NR system for example purposes and uses NR terminology in much of the following description, although the techniques may also be applicable to applications other than NR system applications.

[0067] The following description provides examples and does not limit the scope, applicability, or configuration set forth in the claims. Changes may be made to the functions and arrangements of the elements discussed without departing from the spirit and scope of the present disclosure. Various examples may appropriately omit, replace, or add various procedures or components. For example, the described methods may be performed in an order different from that described, and various steps may be added, omitted, or combined. In addition, features described with reference to certain examples may be combined in other examples.

[0068] See also Figure 1 , Figure 1 A block diagram of a wireless communication system applicable to an embodiment of the present invention is shown. The wireless communication system includes a terminal 11 and a network device 12. The terminal 11 may also be referred to as a user terminal or user equipment (UE, User Equipment), and the terminal 11 may be a terminal side device such as a mobile phone, a tablet personal computer (Tablet Personal Computer), a laptop computer (Laptop Computer), a personal digital assistant (Personal Digital Assistant, PDA), a mobile Internet device (Mobile Internet Device, MID), a wearable device (Wearable Device) or a vehicle-mounted device. It should be noted that the specific type of the terminal 11 is not limited in the embodiment of the present invention. The network device 12 can be a base station and / or a core network element, wherein the above-mentioned base station can be a base station of 5G and later versions (for example: gNB, 5G NR NB, etc.), or a base station in other communication systems (for example: eNB, WLAN access point, or other access point, etc.), wherein the base station can be called a node B, an evolved node B, an access point, a base transceiver station (Base Transceiver Station, BTS), a radio base station, a radio transceiver, a basic service set (Basic Service Set, BSS), an extended service set (Extended Service Set, ESS), a B node, an evolved B node (eNB), a home B node, a home evolved B node, a WLAN access point, a WiFi node or some other suitable term in the field. As long as the same technical effect is achieved, the base station is not limited to specific technical vocabulary. It should be noted that in the embodiment of the present invention, only the base station in the NR system is taken as an example, but the specific type of the base station is not limited.

[0069] The base station can communicate with the terminal 11 under the control of a base station controller, and in various examples, the base station controller can be part of the core network or certain base stations. Some base stations can communicate control information or user data with the core network via a backhaul. In some examples, some of these base stations can communicate with each other directly or indirectly via a backhaul link, which can be a wired or wireless communication link. The wireless communication system can support operations on multiple carriers (waveform signals of different frequencies). A multi-carrier transmitter can transmit modulated signals on these multiple carriers at the same time. For example, each communication link can be a multi-carrier signal modulated according to various radio technologies. Each modulated signal can be sent on a different carrier and can carry control information (e.g., reference signals, control channels, etc.), overhead information, data, etc.

[0070] The base station can communicate wirelessly with the terminal 11 via one or more access point antennas. Each base station can provide communication coverage for its respective coverage area. The coverage area of ​​the access point can be divided into sectors that constitute only a portion of the coverage area. The wireless communication system may include different types of base stations (e.g., macro base stations, micro base stations, or micro-micro base stations). The base station may also utilize different radio technologies, such as cellular or WLAN radio access technologies. The base station may be associated with the same or different access networks or operator deployments. The coverage areas of different base stations (including coverage areas of the same or different types of base stations, coverage areas utilizing the same or different radio technologies, or coverage areas belonging to the same or different access networks) may overlap.

[0071] The communication link in the wireless communication system may include an uplink for carrying an uplink (UL) transmission (e.g., from the terminal 11 to the network device 12), or a downlink for carrying a downlink (DL) transmission (e.g., from the network device 12 to the terminal 11). UL transmission may also be referred to as reverse link transmission, and DL transmission may also be referred to as forward link transmission. Downlink transmission may be performed using a licensed frequency band, an unlicensed frequency band, or both. Similarly, uplink transmission may be performed using a licensed frequency band, an unlicensed frequency band, or both.

[0072] As described in the background technology, in the prior art, the judgment of the network load status in the operation and maintenance of the mobile network usually relies on the expert experience of the network optimization engineers, and the accuracy of the judgment result is not high. In order to solve at least one of the above problems, an embodiment of the present invention provides a method for detecting abnormal cell status, which can simplify the detection of abnormal network load status and improve the accuracy of the detection result.

[0073] Please refer to Figure 2 , a method for detecting abnormal cell status provided by an embodiment of the present invention can be applied to network side equipment, such as Figure 1 The base station shown in can also be applied to a core network element. The method includes:

[0074] Step 21: obtaining at least one key performance indicator KPI parameter group of the first cell collected within a preset time period, wherein the KPI parameter group includes values ​​of multiple KPI parameters collected at corresponding sampling time points.

[0075] Specifically, each KPI parameter group corresponds to a sampling time point, and the KPI parameter group is the values ​​of the multiple KPI parameters of the first cell collected at the corresponding sampling time point, and different KPI parameter groups are the values ​​of the multiple KPI parameters of the first cell collected at different time points.

[0076] Here, when the load of the first cell needs to be detected, the embodiment of the present invention obtains at least one KPI parameter group of the first cell collected within a preset time period, and each KPI parameter group includes multiple pre-selected KPI parameters. At the end of this article, specific examples of the multiple KPI parameters are given.

[0077] Step 22: Perform dimensionality reduction processing on each KPI parameter group of the first cell to generate at least one first dimensionality reduction vector of the first cell.

[0078] Here, in order to simplify the subsequent amount of computation, the embodiment of the present invention can use an autoencoder to perform dimensionality reduction processing on each KPI parameter group, and input each KPI parameter group of the first cell into a pre-trained autoencoder to obtain the first dimensionality reduction vector output by the autoencoder, so that each KPI parameter group can obtain a corresponding first dimensionality reduction vector, and a corresponding number of first dimensionality reduction vectors can be obtained for at least one KPI parameter group. Through dimensionality reduction processing, the embodiment of the present invention can simplify the amount of computation for subsequent classification processing, simplify the detection of abnormal network load status, and improve detection efficiency.

[0079] Step 23: input the at least one first dimension reduction vector into an evaluator to generate evaluation information of the abnormal state of the first cell.

[0080] Here, the embodiment of the present invention can use a pre-generated evaluator to generate evaluation information of abnormal cell status. The evaluator is used to perform clustering processing on the at least one first dimensionality reduction vector, determine and output the first category to which the at least one first dimensionality reduction vector belongs and its corresponding annotation information, wherein the first category is one of a plurality of pre-generated categories, and the annotation information is used to indicate the abnormal cell load status of the corresponding category.

[0081] The embodiment of the present invention can use various clustering algorithms to obtain multiple classifications in advance. Then, in step 23, the at least one first dimensionality reduction vector obtained in step 22 is clustered using the same clustering algorithm to determine the center point of the at least one first dimensionality reduction vector (referred to as the first center point here for ease of description), calculate the Euclidean distance between the first center point and the center points of each pre-generated category, and select the category corresponding to the shortest Euclidean distance as the first category to which the at least one first dimensionality reduction vector belongs. In addition, the labeling information corresponding to the first category can also be determined based on the correspondence between the pre-established labeling information and the category. Here, the labeling information corresponding to each category can be provided in advance by a network optimization engineer and saved locally on the device. The labeling information can specifically include at least one of the following information: an evaluation value of the degree of abnormality of the cell, an evaluation value of the degree of load of the cell, an evaluation value of the degree of optimization of the cell, and remarks information.

[0082] Optionally, before executing step 23, outlier detection may be performed on the at least one first dimensionality reduction vector to remove outliers therein to obtain an updated first dimensionality reduction vector.

[0083] Through the above steps, the embodiment of the present invention can use multiple KPI parameters to detect abnormal cell load. Due to the use of dimensionality reduction processing and clustering processing, the embodiment of the present invention can use more KPI parameters for detection, thereby improving the accuracy of the detection result. In addition, due to the use of autoencoders to perform dimensionality reduction processing on multiple KPI parameters and the use of clustering algorithms for category prediction, the embodiment of the present invention does not need to rely on manual judgment of network optimization engineers, and can simplify the detection of abnormal network load status.

[0084] Before the above step 21, the embodiment of the present invention may pre-train the autoencoder and generate the multiple categories. The training of the autoencoder and the generation of multiple categories will be described below.

[0085] The embodiment of the present invention can obtain multiple KPI parameter groups of multiple cells collected at multiple sampling time points, wherein each KPI parameter group includes the multiple KPI parameters collected from the same cell at a sampling time point.

[0086] For example, the KPI parameters of multiple cells collected from the network management are assumed to have been formatted in a data table format, and the i-th row and j-th column represent the j-th KPI parameter of the i-th sampling point (i.e., cell). The unique identifier of each row is in the form of the cell unique name + data collection timestamp. The above method of the embodiment of the present invention finally outputs the recommendation information related to a cell and load inferred based on the KPI parameters of a certain period of time, and can also output the load recommendation information of a cell at any KPI collection time.

[0087] The overall process of the solution of the embodiment of the present invention is as follows Figure 3 As shown in the figure, it can be divided into two stages: training stage and inference stage (sometimes referred to as the actual application stage in this article). In the training stage, the dimension reduction step requires training a neural network of an autoencoder, and the state mapping step requires manual intervention to complete the labeling of category information. The dotted line shows the additional steps included in the training stage.

[0088] like Figure 3 As shown, before the dimensionality reduction process is performed, the embodiment of the present invention can filter out some noise data or data of low quality through a data cleaning step. For example, KPI parameters with too single values ​​are removed according to the following calculation method: when the frequency of the value with the most occurrences of a certain KPI parameter is greater than a certain preset threshold, the parameter is deleted. Then, the KPI parameters after data cleaning can also be normalized to a distribution form of zero mean and unit variance.

[0089] Then, using the multiple KPI parameter groups, an autoencoder is trained until a preset training end condition is reached (eg, the training duration reaches a preset duration threshold, or the training data volume reaches a preset data volume threshold). Figure 3 As shown, the autoencoder generally includes an encoder and a decoder, and the encoder and the decoder both include a multi-layer neural network, for example, Figure 3 The encoder and decoder in the example include three layers of neural networks, respectively. The first and second layers include a fully connected layer, a PRelu activation layer, and a batch normalization layer, respectively, and the third layer is a fully connected network. The three layers (Layer1 to Layer3) of the encoder have N1, N2, and N3 neurons, respectively, while the three layers (Layer1 to Layer3) of the decoder have N2, N1, and N0 neurons, where N0 is the data dimension of the input encoder, that is, the number of KPI parameters in each KPI data group. The typical values ​​of N1, N2, and N3 are 64, 16, and 2.

[0090] The dimensionality reduction process of the autoencoder is different in the actual application stage of step 22 and in the training stage before the application. In the training stage, the encoder and decoder of the autoencoder are regarded as one, and the decoder finally outputs the estimated value of the original data. The original data is regarded as the labeled value, and the loss function is calculated according to the estimated value and the labeled value. The loss function is Smooth l2 loss with l1 regularization term (as shown in Formula 1 below). The Adam algorithm is used to train the encoder and decoder for the loss function. During training, batch normalization uses the mean and variance calculated by batch.

[0091] The loss function is calculated as follows:

[0092]

[0093] Where x is the labeled value, is the estimate and M is the dimension of x. is the l1 regularization term.

[0094] In the actual application stage, only the encoder works, the encoder processes the input data, and the encoder outputs the data dimensionality reduction result. The data dimensionality reduction result can be further processed by the outlier detection component to filter out a small number of outliers and send them to the next step. In the actual application stage, batch normalization uses the mean and variance calculated by the data as a whole. Here, outliers are sometimes also called flying points, outliers or abnormal points, which refer to individual values ​​in the sample, whose data deviates significantly from the rest of the observations. Outliers can be detected using outlier detection algorithms, such as box plots, simple statistics, 3σ principle and other algorithms.

[0095] For more details about the autoencoder, please refer to the introduction of related existing technologies, which will not be described in detail in this article.

[0096] After obtaining the trained autoencoder, the embodiment of the present invention can obtain multiple dimensionality reduction vectors obtained after the multiple KPI parameter groups are input into the trained autoencoder, and cluster the multiple dimensionality reduction vectors to obtain multiple categories. The clustering algorithm can adopt the K-Means algorithm. For example, the number of divided categories can be Wherein N is the number of the plurality of KPI parameter groups. For example, in the above data table, each row of data represents a plurality of KPI parameters of a certain cell at a certain data collection time, and in this case, N is the number of rows of the data table.

[0097] After obtaining multiple categories, you can also perform the following processing for each category:

[0098] 1) Select a parameter group belonging to the category from the multiple KPI parameter groups, and use the selected KPI parameter group to train a classifier, and calculate the weight of each KPI parameter in the KPI parameter group according to the training result. The classifier here can be a base classifier, specifically a CART decision tree.

[0099] Here, assuming that K categories are obtained after clustering, the category labels obtained by clustering can be mapped to each KPI parameter group. In this way, the KPI parameter groups under each category can be screened out by the category labels to obtain all KPI parameter groups under the category. If there are fewer samples (i.e., KPI parameter groups) under a certain category, the samples can be balanced to generate a sufficient number of samples.

[0100] In addition, here, the weight of each KPI parameter is calculated based on the training results. The weight of a KPI parameter represents the importance of a certain KPI parameter in distinguishing the category from other categories. The weight of a KPI parameter is obtained by calculating the sum of the values ​​of purity reduction caused by all nodes containing the KPI parameter.

[0101] 2) Repeat the following steps until no KPI parameter exists in the KPI parameter group: select a target KPI parameter with the largest weight from the remaining KPI parameters in the parameter group, add it to a parameter set, and delete the target KPI parameter and KPI parameters whose correlation with the target KPI parameter is greater than a preset threshold from the parameter group.

[0102] Here, the weights can be sorted, and the Pearson correlation coefficients of all KPI parameters in the category can be calculated. Then, starting from the KPI parameter with the highest weight, the KPI parameter with the highest weight is added to a parameter set, and then, other KPI parameters whose correlation coefficients with the KPI parameter with the highest weight are greater than a certain threshold (such as 0.8) are removed in turn. By repeating the above process, a parameter set containing one or more KPI parameters can be finally obtained.

[0103] 3) Selecting a preset number of KPI parameters from the parameter set in descending order of weight as the KPI parameters of the category, and calculating the deviation of the statistical features of the selected KPI parameters in the category relative to the statistical features in all samples. Here, all samples refer to the multiple KPI parameter groups used when training the autoencoder.

[0104] Here, considering that there may be more KPI parameters in the parameter set, a preset number (for example, 10) can be set, and then all or part of the parameters are selected from the parameter set as the KPI parameters of this category, and the deviation of the statistical characteristics of the selected KPI parameters in this category relative to the statistical characteristics in all samples is calculated.

[0105] 4) The KPI parameters, weights and deviations of the category are output as the features of the category.

[0106] Here, the KPI parameters, weights, and deviations of each category are output as features of the category, which can be provided to network optimization engineers for scoring.

[0107] 5) Receive the annotation information input for the category and establish a corresponding relationship between the annotation information and the category.

[0108] Here, the scoring results (i.e., annotation information) of each category by the network optimization engineer can be received. For example, the scoring items include three items, namely, the load abnormality degree, the load degree, and the degree to be optimized. Each item has five options (extremely low, low, medium, high, and extremely high), corresponding to 1 to 5 points respectively. A possible sample table format is shown in Table 1.

[0109] Abnormal load Load level Degree to be optimized Reason Note

[0110] Table 1

[0111] The scoring results of the existing network engineers are mapped to categories. In this way, if there is work order information, the work order information can be extracted for comprehensive mapping.

[0112] Here are some specific examples of analysis:

[0113] Cell analysis example:

[0114] By using the method of the embodiment of the present invention, the category of a cell and its corresponding labeling information can be output, and this information can be provided to network optimization engineers for decision making.

[0115] For example, for a certain cell, cell analysis includes the following steps:

[0116] 1) Count the proportion of states with abnormality degree>=3 in the cell time axis, and when the above proportion>10%, output the indication information of the abnormal cell, and output the marked reason corresponding to the abnormal state with the largest proportion.

[0117] 2) Count the proportion of states with abnormal degree>=3 and high load degree>=3. When the proportion is >10%, output the indication information of the high load cell, and output the marked reasons corresponding to the high load abnormal state with the largest proportion.

[0118] 3) Count the proportion of states with abnormality degree>=3 and degree to be optimized>=3. When the proportion is>10%, output the indication information of the cell to be optimized, and output the marked reasons corresponding to the abnormal state to be optimized with the largest proportion.

[0119] 4) The intermediate calculation results such as the category's KPI parameter weights, status timing, etc. are also provided as options for engineers to review.

[0120] Cell timing analysis example:

[0121] Since the calculated state has a timing characteristic, its timing information can be used for further abnormal classification.

[0122] Example 1: Before clustering, the states of P consecutive time series are regarded as joint states for clustering. P is a preset positive integer, for example, 2.

[0123] Example 2: When determining that a cell is abnormal, the cell is determined to be abnormal only when an abnormal state occurs Q times consecutively, where Q is a preset positive integer, for example, 4.

[0124] It can be seen from the above that the embodiments of the present invention have at least the following advantages:

[0125] 1) The embodiment of the present invention introduces an unsupervised method for detecting abnormal cell status. Compared with the existing threshold-based solution, it can use any number of parameters, adapt to network changes, and detect more accurately without manual work.

[0126] 2) Compared with the output of the existing threshold-based cell anomaly detection, the embodiment of the present invention can locate multiple problems at one time, and can also output multiple candidate anomalies and sort them by weight, and can adjust the precision or recall rate by the distance threshold.

[0127] 3) Compared with the existing methods based on manual experience, the embodiments of the present invention automatically map human experience in the actual application stage without manual processing.

[0128] 4) Compared with the existing method of screening key KPIs based on statistical mean deviation, the embodiment of the present invention can detect abnormal scenarios where statistical deviation is irrelevant to the category. For example, if some parameters have large fluctuations, it may be that the parameter of a certain type of abnormality behaves abnormally but is not the main cause of the abnormality.

[0129] 5) Compared with the existing statistical methods, the embodiments of the present invention can improve the accuracy of problem location and output multiple potential candidate causes.

[0130] 6) Compared with existing methods, the number of clustering categories is The determination is made to match the actual number of abnormal categories of the cell-level KPI parameters in the existing network.

[0131] 7) Compared with the existing manual labeling scheme, the embodiments of the present invention are more structured than the existing operation and maintenance labeling data, which is conducive to data training.

[0132] The above describes various methods of the embodiments of the present invention. The following further provides devices for implementing the above methods.

[0133] Please refer to Figure 5 The embodiment of the present invention provides a detection device 50 for detecting abnormal state of a cell, comprising:

[0134] A parameter acquisition module 51 is used to acquire at least one key performance indicator KPI parameter group of the first cell collected within a preset time period, wherein the KPI parameter group includes values ​​of multiple KPI parameters collected at corresponding sampling time points;

[0135] A dimensionality reduction processing module 52, configured to perform dimensionality reduction processing on each KPI parameter group of the first cell respectively to generate at least one first dimensionality reduction vector of the first cell;

[0136] The labeling processing module 53 is used to input the at least one first dimension reduction vector into an evaluator to generate evaluation information of the abnormal state of the first cell.

[0137] Optionally, the parameter acquisition module 51 is further used to input each KPI parameter group of the first cell into a pre-trained autoencoder to obtain a first dimensionality reduction vector output by the autoencoder.

[0138] Optionally, the labeling processing module 53 is also used to perform clustering processing on the at least one first dimensionality reduction vector through the evaluator, determine and output the first category to which the at least one first dimensionality reduction vector belongs and its corresponding labeling information, wherein the first category is one of multiple pre-generated categories, and the labeling information is used to indicate the abnormal status of the cell of the corresponding category.

[0139] Optionally, the detection device further includes:

[0140] An autoencoder training module is used to obtain multiple KPI parameter groups of multiple cells collected at multiple sampling time points before obtaining at least one KPI parameter group of the first cell, wherein each KPI parameter group includes the multiple KPI parameters collected by the same cell at a sampling time point; using the multiple KPI parameter groups, training an autoencoder until a preset training end condition is reached; wherein the autoencoder includes an encoder and a decoder, and the encoder and the decoder both include multi-layer neural networks.

[0141] Optionally, the detection device further includes:

[0142] A clustering module is used to obtain a plurality of reduced-dimensionality vectors obtained after inputting the plurality of KPI parameter groups into the trained autoencoder; and cluster the plurality of reduced-dimensionality vectors to obtain a plurality of categories.

[0143] Optionally, the detection device further includes:

[0144] The labeling information establishment module is used to perform the following processing for each category after obtaining multiple categories:

[0145] Selecting a parameter group belonging to the category from the multiple KPI parameter groups, training a classifier using the selected parameter group, and calculating the weight of each KPI parameter in the parameter group according to the training result;

[0146] Repeat the following steps until no KPI parameter exists in the parameter group: select a target KPI parameter with the largest weight from the remaining KPI parameters in the parameter group, add it to a parameter set, and delete the target KPI parameter and KPI parameters whose correlation with the target KPI parameter is greater than a preset threshold from the parameter group;

[0147] Selecting a preset number of KPI parameters from the parameter set in descending order of weight as the KPI parameters of the category, and calculating the deviation of the statistical features of the selected KPI parameters in the category relative to the statistical features in all samples;

[0148] The KPI parameters, weights and deviations of the category are used as feature outputs of the category;

[0149] The annotation information input for the category is received, and a corresponding relationship between the annotation information and the category is established.

[0150] Optionally, the annotation information includes at least one of the following information: an evaluation value of the cell abnormality level, an evaluation value of the cell load level, an evaluation value of the cell optimization level, and remark information.

[0151] Optionally, the number of the multiple categories is Wherein, N is the number of the multiple groups of KPI parameters.

[0152] Optionally, the labeling processing module is also used to perform clustering processing on the at least one first dimensionality reduction vector, determine the first center point of the at least one first dimensionality reduction vector, calculate the Euclidean distance between the first center point and the center points of each category, and select the category corresponding to the shortest Euclidean distance as the first category to which the at least one first dimensionality reduction vector belongs; and determine the labeling information corresponding to the first category based on the pre-established correspondence between the labeling information and the category.

[0153] Optionally, the detection device further includes:

[0154] The outlier removal module is used to perform outlier detection on the at least one first dimensionality reduction vector before clustering the at least one first dimensionality reduction vector, remove outliers therein, and obtain an updated first dimensionality reduction vector.

[0155] It should be noted that the device in this embodiment is the same as the above Figure 2The device corresponding to the method shown, the implementation methods in the above embodiments are all applicable to the embodiments of the device, and can also achieve the same technical effects. The above device provided by the embodiment of the present invention can implement all the method steps implemented by the above method embodiment, and can achieve the same technical effects. The parts and beneficial effects that are the same as those in the method embodiment will not be described in detail here.

[0156] Please refer to Figure 6 The embodiment of the present invention provides a schematic diagram of a structure of a detection device 600 for detecting abnormal state of a cell, including: a processor 601, a transceiver 602, a memory 603 and a bus interface, wherein:

[0157] In the embodiment of the present invention, the cell abnormal state detection device 600 further includes: a program stored in the memory 603 and executable on the processor 601, and the program is executed by the processor 601 to implement the following steps:

[0158] Acquire at least one key performance indicator KPI parameter group of the first cell collected within a preset time period, the KPI parameter group including values ​​of multiple KPI parameters collected at corresponding sampling time points; perform dimensionality reduction processing on each KPI parameter group of the first cell respectively to generate at least one first dimensionality reduction vector of the first cell; input the at least one first dimensionality reduction vector into an evaluator to generate evaluation information of the abnormal state of the first cell.

[0159] Optionally, when the processor executes the program, the processor further implements the following steps:

[0160] Each KPI parameter group of the first cell is input into a pre-trained autoencoder to obtain a first dimension reduction vector output by the autoencoder.

[0161] Optionally, when the processor executes the program, the processor further implements the following steps:

[0162] The evaluator performs clustering processing on the at least one first dimensionality reduction vector, determines and outputs a first category to which the at least one first dimensionality reduction vector belongs and its corresponding labeling information, wherein the first category is one of a plurality of pre-generated categories, and the labeling information is used to indicate an abnormal state of a cell of the corresponding category.

[0163] Optionally, when the processor executes the program, the processor further implements the following steps:

[0164] Acquire multiple KPI parameter groups of multiple cells collected at multiple sampling time points, wherein each KPI parameter group includes the multiple KPI parameters collected by the same cell at a sampling time point;

[0165] Using the multiple KPI parameter groups, an autoencoder is trained until a preset training end condition is reached; wherein the autoencoder includes an encoder and a decoder, and the encoder and the decoder both include multi-layer neural networks.

[0166] Optionally, when the processor executes the program, the following steps are further implemented: obtaining multiple dimensionality reduction vectors obtained after inputting the multiple KPI parameter groups into the trained autoencoder; and clustering the multiple dimensionality reduction vectors to obtain multiple categories.

[0167] Optionally, when executing the program, the processor further implements the following steps: after obtaining multiple categories, for each category, respectively perform the following processing:

[0168] Selecting a parameter group belonging to the category from the multiple KPI parameter groups, training a classifier using the selected parameter group, and calculating the weight of each KPI parameter in the parameter group according to the training result;

[0169] Repeat the following steps until no KPI parameter exists in the parameter group: select a target KPI parameter with the largest weight from the remaining KPI parameters in the parameter group, add it to a parameter set, and delete the target KPI parameter and KPI parameters whose correlation with the target KPI parameter is greater than a preset threshold from the parameter group;

[0170] Selecting a preset number of KPI parameters from the parameter set in descending order of weight as the KPI parameters of the category, and calculating the deviation of the statistical features of the selected KPI parameters in the category relative to the statistical features in all samples;

[0171] The KPI parameters, weights and deviations of the category are used as feature outputs of the category;

[0172] The annotation information input for the category is received, and a corresponding relationship between the annotation information and the category is established.

[0173] Optionally, the annotation information includes at least one of the following information: an evaluation value of the cell abnormality level, an evaluation value of the cell load level, an evaluation value of the cell optimization level, and remark information.

[0174] Optionally, the number of the multiple categories is Wherein, N is the number of the multiple KPI parameter groups.

[0175] Optionally, when executing the program, the processor also implements the following steps: clustering the at least one first dimensionality reduction vector, determining the first center point of the at least one first dimensionality reduction vector, calculating the Euclidean distance between the first center point and the center points of each category, and selecting the category corresponding to the shortest Euclidean distance as the first category to which the at least one first dimensionality reduction vector belongs; and determining the labeling information corresponding to the first category based on the pre-established correspondence between the labeling information and the category.

[0176] Optionally, when executing the program, the processor further implements the following steps: performing outlier detection on the at least one first dimensionality reduction vector, removing outliers therein, and obtaining an updated first dimensionality reduction vector.

[0177] exist Figure 6 In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically linking together various circuits of one or more processors represented by processor 601 and memory represented by memory 603. The bus architecture may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface. The transceiver 602 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices over a transmission medium.

[0178] The processor 601 is responsible for managing the bus architecture and general processing, and the memory 603 can store data used by the processor 601 when performing operations.

[0179] It should be noted that the device in this embodiment is the same as the above Figure 2 The device corresponding to the method shown, the implementation methods in the above embodiments are all applicable to the embodiments of the device, and can also achieve the same technical effect. In the device, the transceiver 602 and the memory 603, as well as the transceiver 602 and the processor 601 can be connected through the bus interface communication, the function of the processor 601 can also be implemented by the transceiver 602, and the function of the transceiver 602 can also be implemented by the processor 601. It should be noted that the above device provided by the embodiment of the present invention can implement all the method steps implemented by the above method embodiment, and can achieve the same technical effect, and the parts and beneficial effects that are the same as the method embodiment in this embodiment will not be specifically repeated here.

[0180] In some embodiments of the present invention, a computer-readable storage medium is further provided, on which a program is stored, and when the program is executed by a processor, the following steps are implemented:

[0181] Acquire at least one key performance indicator KPI parameter group of the first cell collected within a preset time period, the KPI parameter group including values ​​of multiple KPI parameters collected at corresponding sampling time points; perform dimensionality reduction processing on each KPI parameter group of the first cell respectively to generate at least one first dimensionality reduction vector of the first cell; input the at least one first dimensionality reduction vector into an evaluator to generate evaluation information of the abnormal state of the first cell.

[0182] When the program is executed by the processor, it can implement all the implementation methods in the above-mentioned detection method of the abnormal state of the cell applied to the detection device, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0183] The following is a list of cell-level KPI parameters that can be used in the embodiments of the present invention. The present invention can use some or all of the following parameters:

[0184] Uplink average MCS;

[0185] RSSI_PUCCH_AVG;

[0186] Downstream dual flow ratio;

[0187] Downlink average MCS;

[0188] SINR_PUSCH_AVG;

[0189] Downlink transmission efficiency (bytes / prb);

[0190] AVERAGE_CQI;

[0191] Downlink retransmission rate LTE_5208a;

[0192] RSSI_PUSCH_AVG;

[0193] Downstream traffic (MB);

[0194] PHR mean;

[0195] Switch execution attempt;

[0196] Number of downlink packets in the cell;

[0197] Uplink packet loss rate on the user side of the cell;

[0198] VoLTE uplink packet loss rate;

[0199] RRC / RACH ratio;

[0200] Maximum downlink rate (Mbit / s);

[0201] Average downlink rate (Mbit / s);

[0202] PDCCH_CCE average utilization;

[0203] MAC layer uplink block error rate;

[0204] Number of PRACH requests;

[0205] Uplink retransmission rate LTE_5207b;

[0206] Number of uplink packets in the cell;

[0207] SINR_PUCCH_AVG;

[0208] Average number of RRC connections;

[0209] CA-capable terminals;

[0210] AGG8 proportion (%);

[0211] AGG1_USED_PDCCH;

[0212] AGG4_USED_PDCCH;

[0213] Number of successful DRB establishment;

[0214] UE_PER_DL_TTI_AVG;

[0215] Average number of valid RRC connections;

[0216] Maximum number of RRC connections;

[0217] PRACH establishment success rate;

[0218] DRB establishment success rate;

[0219] Wireless drop rate;

[0220] Uplink traffic (MB);

[0221] AGG2_USED_PDCCH;

[0222] Switch-in success rate;

[0223] Average uplink PRB utilization;

[0224] Uplink transmission efficiency (bytes / prb);

[0225] Number of failed switch-in attempts;

[0226] Number of uplink packet losses in the cell (QCI=1);

[0227] PRACH load;

[0228] Number of PRACH successes;

[0229] Uplink interference value;

[0230] AGG8_USED_PDCCH;

[0231] Number of uplink packet losses in the cell;

[0232] UE_PER_UL_TTI_AVG;

[0233] Number of successful switch-in times;

[0234] UE_PER_UL_TTI_MAX;

[0235] Number of RRC establishment requests;

[0236] Downlink packet loss rate of the cell user plane;

[0237] MAC layer downlink block error rate;

[0238] Number of DRB establishment requests;

[0239] HO_PREP_IN_SUCC;

[0240] UE_PER_DL_TTI_MAX;

[0241] Including preparation switching success rate;

[0242] Number of ERAB establishment requests;

[0243] Downlink PRB average utilization;

[0244] Maximum number of valid RRC connections;

[0245] Average uplink rate (Mbit / s);

[0246] Including preparation switching attempt;

[0247] Including the success rate of preparing to switch in;

[0248] Number of switch-in attempts;

[0249] E_RAB successful establishment times QCI_1;

[0250] Number of downlink packets in the cell (QCI=1);

[0251] The switch was executed successfully;

[0252] Wireless drop rate_denominator;

[0253] VoLTE downlink packet loss rate;

[0254] Switch execution failed;

[0255] Number of successful RRC establishments;

[0256] Number of downlink packet losses in the cell;

[0257] Switch execution success rate;

[0258] VoLTE voice traffic (Erl);

[0259] Including preparation switch failure;

[0260] Maximum uplink rate (Mbit / s);

[0261] ENB_INIT_TO_IDLE_RNL;

[0262] E_RAB establishment success rate QCI_2;

[0263] FAIL_ENB_HO_PREP_AC;

[0264] SIGN_CONN_ESTAB_REJ_NEMG;

[0265] Number of ERAB establishment failures;

[0266] Number of successful ERAB establishments;

[0267] ERAB_ADD_SETUP_FAIL_RNL_MOB;

[0268] eSVRCC switching ratio;

[0269] ERAB_INI_SETUP_FAIL_TNL_TRU;

[0270] ERAB_ADD_SETUP_FAIL_TNL_TRU;

[0271] ENB_INIT_TO_IDLE_OTHER;

[0272] Number of successful eSVRCC switching;

[0273] Number of eSVRCC handover requests;

[0274] eSVRCC handover success rate;

[0275] FAIL_ENB_HO_PREP_QCI;

[0276] E_RAB abnormal release times QCI_1;

[0277] E_RAB establishment request number QCI_1;

[0278] E_RAB drop rate QCI_1;

[0279] E_RAB establishment request number QCI_2;

[0280] VoLTE video traffic (Erl);

[0281] E_RAB successful establishment times QCI_2;

[0282] RRC establishment success rate;

[0283] Congestion times;

[0284] Wireless drop rate_numerator;

[0285] Number of load-related switching requests;

[0286] Number of successful switching due to load reasons;

[0287] PDCCH congestion rate (%);

[0288] Number of failed connections;

[0289] FAIL_ENB_HO_PREP_TIME;

[0290] Number of CA configuration users;

[0291] Number of downlink packet losses in the cell (QCI=1);

[0292] Number of cell uplink packets (QCI=1);

[0293] Number of CA activated users;

[0294] E_RAB establishment success rate QCI_1;

[0295] Including the number of attempts to prepare for switching in;

[0296] Wireless connection rate;

[0297] E_RAB drop rate denominator QCI_1;

[0298] Number of RRC establishment failures;

[0299] SIGN_EST_F_RRCCOMPL_MISSING;

[0300] FAIL_ENB_HO_PREP_OTHER;

[0301] Success rate of ERAB establishment;

[0302] Cell availability.

[0303] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0304] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0305] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0306] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of the present invention.

[0307] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0308] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, ROM, RAM, magnetic disks, or optical disks.

[0309] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for detecting abnormal state of a cell, characterized in that: include: Acquire at least one key performance indicator KPI parameter group of the first cell collected within a preset time period, where the KPI parameter group includes values ​​of multiple KPI parameters collected at corresponding sampling time points; Performing dimensionality reduction processing on each KPI parameter group of the first cell respectively to generate at least one first dimensionality reduction vector of the first cell; Inputting the at least one first dimension reduction vector into an evaluator to generate evaluation information of the abnormal state of the first cell; The evaluator is used to perform clustering processing on the at least one first dimensionality reduction vector, determine and output the first category to which the at least one first dimensionality reduction vector belongs and its corresponding annotation information, wherein the first category is one of multiple categories generated in advance, and the annotation information is used to indicate the abnormal state of the cell of the corresponding category; the multiple categories are obtained by clustering multiple dimensionality reduction vectors, and the multiple dimensionality reduction vectors correspond to multiple KPI parameter groups of multiple cells collected at multiple sampling time points, and the correspondence between the categories and the annotation information is established in the following manner: For each category, perform the following processing: Selecting a parameter group belonging to the category from the multiple KPI parameter groups, training a classifier using the selected parameter group, and calculating the weight of each KPI parameter in the parameter group according to the training result; Repeat the following steps until no KPI parameter exists in the parameter group: select a target KPI parameter with the largest weight from the remaining KPI parameters in the parameter group, add it to a parameter set, and delete the target KPI parameter and KPI parameters whose correlation with the target KPI parameter is greater than a preset threshold from the parameter group; Selecting a preset number of KPI parameters from the parameter set in descending order of weight as the KPI parameters of the category, and calculating the deviation of the statistical features of the selected KPI parameters in the category relative to the statistical features in all samples; The KPI parameters, weights and deviations of the category are used as feature outputs of the category; The annotation information input for the category is received, and a corresponding relationship between the annotation information and the category is established.

2. The method according to claim 1, characterized in that Performing dimensionality reduction processing on each KPI parameter group of the first cell respectively includes: Each KPI parameter group of the first cell is input into a pre-trained autoencoder to obtain a first dimension reduction vector output by the autoencoder.

3. The method according to any one of claims 1 to 2, characterized in that: Before acquiring at least one KPI parameter group of the first cell, the method further includes: Acquire multiple KPI parameter groups of multiple cells collected at multiple sampling time points, wherein each KPI parameter group includes the multiple KPI parameters collected at a sampling time point for the same cell; Using the multiple KPI parameter groups, an autoencoder is trained until a preset training end condition is reached; wherein the autoencoder includes an encoder and a decoder, and the encoder and the decoder both include multi-layer neural networks.

4. The method according to claim 3, characterized in that Also includes: Obtaining multiple dimensionality reduction vectors obtained by inputting the multiple KPI parameter groups into the trained autoencoder; Clustering is performed on the multiple dimension reduction vectors to obtain multiple categories.

5. The method according to claim 1, characterized in that The annotation information includes at least one of the following information: an evaluation value of the abnormality degree of the cell, an evaluation value of the load degree of the cell, an evaluation value of the degree of the cell to be optimized, and remarks information.

6. The method according to claim 4, characterized in that The number of the multiple categories is Wherein, N is the number of the multiple KPI parameter groups.

7. The method according to claim 1, characterized in that Determining the first category to which the at least one first dimensionality reduction vector belongs and its corresponding labeling information includes: Performing clustering processing on the at least one first reduced dimensionality vector, determining a first center point of the at least one first reduced dimensionality vector, calculating the Euclidean distance between the first center point and center points of each category, and selecting a category corresponding to the shortest Euclidean distance as the first category to which the at least one first reduced dimensionality vector belongs; The labeling information corresponding to the first category is determined according to a pre-established correspondence between labeling information and categories.

8. The method according to claim 1, characterized in that Before the step of performing clustering processing on the at least one first dimensionality reduction vector, the method further includes: Outlier detection is performed on the at least one first dimensionality reduction vector to remove outliers therein to obtain an updated first dimensionality reduction vector.

9. A detection device for abnormal state of a cell, characterized in that: include: A parameter acquisition module, used to acquire at least one key performance indicator KPI parameter group of the first cell collected within a preset time period, wherein the KPI parameter group includes values ​​of multiple KPI parameters collected at corresponding sampling time points; A dimensionality reduction processing module, used to perform dimensionality reduction processing on each KPI parameter group of the first cell respectively, and generate at least one first dimensionality reduction vector of the first cell; A labeling processing module, used for inputting the at least one first dimension reduction vector into an evaluator to generate evaluation information of the abnormal state of the first cell; The annotation processing module is further used to perform clustering processing on the at least one first dimensionality reduction vector through the evaluator, determine and output a first category to which the at least one first dimensionality reduction vector belongs and its corresponding annotation information, wherein the first category is one of a plurality of categories generated in advance, and the annotation information is used to indicate an abnormal state of a cell of the corresponding category; the plurality of categories are obtained by clustering a plurality of dimensionality reduction vectors, and the plurality of dimensionality reduction vectors correspond to a plurality of KPI parameter groups of a plurality of cells collected at a plurality of sampling time points; The detection device also includes: The labeling information establishment module is used to perform the following processing for each category after obtaining multiple categories: Selecting a parameter group belonging to the category from the multiple KPI parameter groups, training a classifier using the selected parameter group, and calculating the weight of each KPI parameter in the parameter group according to the training result; Repeat the following steps until no KPI parameter exists in the parameter group: select a target KPI parameter with the largest weight from the remaining KPI parameters in the parameter group, add it to a parameter set, and delete the target KPI parameter and KPI parameters whose correlation with the target KPI parameter is greater than a preset threshold from the parameter group; Selecting a preset number of KPI parameters from the parameter set in descending order of weight as the KPI parameters of the category, and calculating the deviation of the statistical features of the selected KPI parameters in the category relative to the statistical features in all samples; The KPI parameters, weights and deviations of the category are used as feature outputs of the category; The annotation information input for the category is received, and a corresponding relationship between the annotation information and the category is established.

10. The detection device according to claim 9, characterized in that The dimension reduction processing module is further used to input each KPI parameter group of the first cell into a pre-trained autoencoder to obtain a first dimension reduction vector output by the autoencoder.

11. The detection device according to any one of claims 9 to 10, characterized in that: Also includes: An autoencoder training module is used to obtain multiple KPI parameter groups of multiple cells collected at multiple sampling time points before obtaining at least one KPI parameter group of the first cell, wherein each KPI parameter group includes the multiple KPI parameters collected by the same cell at a sampling time point; using the multiple KPI parameter groups, training an autoencoder until a preset training end condition is reached; wherein the autoencoder includes an encoder and a decoder, and the encoder and the decoder both include multi-layer neural networks.

12. The detection device according to claim 11, characterized in that Also includes: A clustering module, used to obtain a plurality of dimension reduction vectors obtained by inputting the plurality of KPI parameter groups into the trained autoencoder; Clustering is performed on the multiple dimension reduction vectors to obtain multiple categories.

13. The detection device according to claim 9, characterized in that: The labeling processing module is also used to perform clustering processing on the at least one first dimensionality reduction vector, determine the first center point of the at least one first dimensionality reduction vector, calculate the Euclidean distance between the first center point and the center points of each category, and select the category corresponding to the shortest Euclidean distance as the first category to which the at least one first dimensionality reduction vector belongs; determine the labeling information corresponding to the first category based on the pre-established correspondence between the labeling information and the category.

14. The detection device according to claim 9, characterized in that: Also includes: The outlier removal module is used to perform outlier detection on the at least one first dimensionality reduction vector before clustering the at least one first dimensionality reduction vector, remove outliers therein, and obtain an updated first dimensionality reduction vector.

15. A detection device for abnormal state of a cell, characterized in that: comprising a transceiver and a processor, wherein: The transceiver is used to obtain at least one key performance indicator KPI parameter group of the first cell collected within a preset time period, wherein the KPI parameter group includes values ​​of multiple KPI parameters collected at corresponding sampling time points; The processor is configured to perform dimensionality reduction processing on each KPI parameter group of the first cell respectively to generate at least one first dimensionality reduction vector of the first cell; input the at least one first dimensionality reduction vector into an evaluator to generate evaluation information of an abnormal state of the first cell; The evaluator is used to perform clustering processing on the at least one first dimensionality reduction vector, determine and output the first category to which the at least one first dimensionality reduction vector belongs and its corresponding annotation information, wherein the first category is one of multiple categories generated in advance, and the annotation information is used to indicate the abnormal state of the cell of the corresponding category; the multiple categories are obtained by clustering multiple dimensionality reduction vectors, and the multiple dimensionality reduction vectors correspond to multiple KPI parameter groups of multiple cells collected at multiple sampling time points, and the correspondence between the categories and the annotation information is established in the following manner: For each category, perform the following processing: Selecting a parameter group belonging to the category from the multiple KPI parameter groups, training a classifier using the selected parameter group, and calculating the weight of each KPI parameter in the parameter group according to the training result; Repeat the following steps until no KPI parameter exists in the parameter group: select a target KPI parameter with the largest weight from the remaining KPI parameters in the parameter group, add it to a parameter set, and delete the target KPI parameter and KPI parameters whose correlation with the target KPI parameter is greater than a preset threshold from the parameter group; Selecting a preset number of KPI parameters from the parameter set in descending order of weight as the KPI parameters of the category, and calculating the deviation of the statistical features of the selected KPI parameters in the category relative to the statistical features in all samples; The KPI parameters, weights and deviations of the category are used as feature outputs of the category; The annotation information input for the category is received, and a corresponding relationship between the annotation information and the category is established.

16. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for detecting an abnormal state of a cell as described in any one of claims 1 to 8 are implemented.