Cell index poor quality detection method and device, computer equipment and storage medium

Abstract

The invention provides a cell index poor quality detection method and device, computer equipment and a storage medium, and the method comprises the steps: obtaining time series data of a cell, the time series data comprising a plurality of time points and cell index data corresponding to each time point; performing frequency domain transformation processing on the time sequence data to obtain a plurality of frequency components; mapping processing is carried out on each frequency component to obtain a corresponding data point, and the data point has corresponding target data information; performing clustering processing on the plurality of data points according to the target data information to obtain at least one cluster, each cluster including at least a part of data points; and performing poor quality detection on the at least one cluster to obtain a detection result. By implementing the method disclosed by the invention, the dependence on parameter selection can be reduced through frequency domain transformation processing and clustering processing, the calculation cost is reduced, and the capability of processing high-dimensional data is improved, so that the accuracy and efficiency of anomaly detection are improved.

Description

Technical Field

[0001] The present disclosure relates to the fields of IT business support and artificial intelligence technology, and in particular to a method, device, computer equipment and storage medium for detecting poor quality of cell indicators. Background Art

[0002] With the rapid development of 5G networks, the requirements of different services for network experience are constantly increasing. The traditional single threshold rule quality difference identification method can no longer meet the diversity of different regions, scenarios and business needs.

[0003] In the related technology, when performing quality difference identification, traditional anomaly detection algorithms, sample anomaly detection methods based on distance and density, and related algorithms using machine learning are usually used to identify abnormal samples in community quality difference scenarios.

[0004] In this way, quality difference detection algorithms often rely on statistical assumptions of the data, which limits their effectiveness when faced with complex distributions or data with multimodal characteristics. In addition, such methods often perform poorly in high-dimensional data scenarios and have difficulty capturing complex relationships in the data. Although distance- and density-based methods can better adapt to the distribution characteristics of data, in practical applications, they are sensitive to parameter selection and often face challenges in computational efficiency. Especially when the amount of data is large and the dimension is high, the cost of calculating density and distance will increase significantly. Summary of the invention

[0005] The present disclosure aims to solve one of the technical problems in the related art at least to some extent.

[0006] To this end, the purpose of the present invention is to propose a method, device, computer equipment and storage medium for detecting cell indicator quality differences, which can reduce the dependence on parameter selection through frequency domain transformation processing and clustering processing, reduce computing costs and improve the ability to process high-dimensional data, thereby improving the accuracy and efficiency of anomaly detection.

[0007] To achieve the above-mentioned purpose, the method for detecting poor cell indicator quality proposed in the first aspect of the present disclosure includes:

[0008] Acquire time series data of a cell, wherein the time series data includes: multiple time points and cell indicator data corresponding to each time point;

[0009] Performing frequency domain transformation processing on the time series data to obtain multiple frequency components;

[0010] Performing mapping processing on each of the frequency components to obtain a corresponding data point, wherein the data point has corresponding target data information;

[0011] Performing clustering processing on the plurality of data points according to the target data information to obtain at least one cluster, wherein each of the clusters includes: at least part of the data points; and

[0012] Perform quality difference detection on the at least one cluster to obtain a detection result.

[0013] To achieve the above-mentioned purpose, the cell indicator quality difference detection device proposed in the second aspect of the present disclosure includes:

[0014] An acquisition module, used to acquire time series data of a cell, wherein the time series data includes: multiple time points and cell indicator data corresponding to each time point;

[0015] A first processing module is used to perform frequency domain transformation processing on the time series data to obtain multiple frequency components;

[0016] A second processing module is used to perform mapping processing on each of the frequency components to obtain a corresponding data point, wherein the data point has corresponding target data information;

[0017] a third processing module, configured to perform clustering processing on the plurality of data points according to the target data information to obtain at least one cluster, wherein each cluster includes: at least part of the data points; and

[0018] The detection module is used to perform quality difference detection on the at least one cluster to obtain a detection result.

[0019] The computer device proposed in the third aspect embodiment of the present disclosure includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the cell indicator quality difference detection method proposed in the first aspect embodiment of the present disclosure is implemented.

[0020] The fourth aspect embodiment of the present disclosure proposes a non-temporary computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the cell indicator quality difference detection method proposed in the first aspect embodiment of the present disclosure is implemented.

[0021] The fifth aspect embodiment of the present disclosure proposes a computer program product. When the instructions in the computer program product are executed by a processor, the cell indicator quality difference detection method proposed in the first aspect embodiment of the present disclosure is executed.

[0022] The cell indicator quality difference detection method, device, computer equipment and storage medium provided by the present disclosure obtain the time series data of the cell, wherein the time series data includes: multiple time points and the cell indicator data corresponding to each time point; perform frequency domain transformation processing on the time series data to obtain multiple frequency components; perform mapping processing on each frequency component to obtain the corresponding data point, wherein the data point has corresponding target data information; perform clustering processing on the multiple data points according to the target data information to obtain at least one cluster cluster, wherein each cluster cluster includes: at least part of the data points; and perform quality difference detection on at least one cluster cluster to obtain the detection result. Thus, the dependence on parameter selection can be reduced through frequency domain transformation processing and clustering processing, the computational cost can be reduced, and the ability to process high-dimensional data can be improved, thereby improving the accuracy and efficiency of anomaly detection.

[0023] Additional aspects and advantages of the present disclosure will be given in part in the following description and in part will be obvious from the following description or learned through practice of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The above and / or additional aspects and advantages of the present disclosure will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0025] Figure 1 It is a flowchart of a method for detecting poor cell indicator quality proposed in an embodiment of the present disclosure;

[0026] Figure 2 is a flowchart of a method for detecting poor cell indicator quality proposed in another embodiment of the present disclosure;

[0027] Figure 3 is a flowchart of a method for detecting poor cell indicator quality proposed in another embodiment of the present disclosure;

[0028] Figure 4 It is a structural schematic diagram of a cell indicator quality difference detection device proposed in an embodiment of the present disclosure;

[0029] Figure 5 A block diagram of an exemplary computer device suitable for implementing embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0030] Embodiments of the present disclosure are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present disclosure, and are not to be construed as limitations of the present disclosure. On the contrary, the embodiments of the present disclosure include all changes, modifications, and equivalents that fall within the spirit and connotation of the appended claims.

[0031] Figure 1 It is a flowchart of a method for detecting poor cell indicator quality proposed in an embodiment of the present disclosure.

[0032] Among them, it should be noted that the executor of the cell indicator quality difference detection method of this embodiment is a cell indicator quality difference detection device, which can be implemented by software and / or hardware. The device can be configured in a computer device, and the computer device can include but is not limited to a terminal, a server, etc. For example, the terminal can be a mobile phone, a handheld computer, etc.

[0033] like Figure 1 As shown, the cell indicator quality difference detection method includes:

[0034] S101: Acquire time series data of a cell, wherein the time series data includes: multiple time points and cell indicator data corresponding to each time point.

[0035] The cell refers to the basic coverage area in a cellular communication system, also known as a base station cell.

[0036] Among them, time series data refers to data composed of performance indicators of a cell at different time points.

[0037] The cell indicator data may include key performance indicators (KPI) and key quality indicators (KQI) of the cell, etc., and there is no limitation on this.

[0038] In the disclosed embodiment, when the time series data of a cell is obtained, reliable data support can be provided for subsequent cell indicator quality difference detection.

[0039] S102: Perform frequency domain transformation on the time series data to obtain multiple frequency components.

[0040] Among them, the frequency component can be used to indicate the relevant characteristics of time series data in the frequency domain.

[0041] That is to say, after obtaining the time series data of the cell in the embodiment of the present disclosure, the time domain information can be decomposed into the frequency domain through frequency domain transformation, and more hidden complex high-dimensional information in the original time series can be collected, thereby improving the ability of the present disclosure embodiment to identify abnormal points with poor quality.

[0042] S103: Mapping is performed on each frequency component to obtain a corresponding data point, wherein the data point has corresponding target data information.

[0043] The data point may refer to a point in three-dimensional space corresponding to the frequency component after mapping.

[0044] The target data information may refer to relevant information corresponding to the data point.

[0045] That is to say, after the embodiment of the present disclosure performs frequency domain transformation processing on the time series data to obtain multiple frequency components, each frequency component can be mapped to obtain corresponding data points, wherein the data points have corresponding target data information, thereby providing reliable data support for the subsequent clustering clusters.

[0046] S104: performing clustering processing on the plurality of data points according to the target data information to obtain at least one cluster, wherein each cluster includes: at least part of the data points.

[0047] Among them, the cluster cluster refers to the cluster obtained after multiple data points are clustered.

[0048] In the embodiment of the present disclosure, when clustering multiple data points according to target data information to obtain at least one cluster, the target data information can be input into a pre-trained machine learning model to obtain the corresponding cluster. Alternatively, multiple data points can be clustered according to the target data information based on a method combining numbers and shapes to obtain at least one cluster, and there is no limitation on this.

[0049] In the disclosed embodiment, after clustering multiple data points according to target data information to obtain at least one cluster, clustering of multiple data points can be implemented to ensure that data points in the same cluster have similar characteristics, thereby effectively improving the reliability of subsequent quality difference detection.

[0050] S105: Perform quality difference detection on at least one cluster to obtain a detection result.

[0051] The detection result can be used to indicate the quality difference identification result of the data in the cluster.

[0052] It is understandable that different types of data points may have significant differences. Therefore, the disclosed embodiment can cluster multiple data points according to the target data information to obtain at least one cluster, and then perform quality difference detection on different clusters to obtain detection results.

[0053] In this embodiment, by acquiring the time series data of the cell, wherein the time series data includes: multiple time points and cell indicator data corresponding to each time point; performing frequency domain transformation processing on the time series data to obtain multiple frequency components; performing mapping processing on each frequency component to obtain a corresponding data point, wherein the data point has corresponding target data information; performing clustering processing on the multiple data points according to the target data information to obtain at least one cluster cluster, wherein each cluster cluster includes: at least part of the data points; and performing quality difference detection on at least one cluster cluster to obtain a detection result. Thus, the frequency domain transformation processing and clustering processing can reduce the dependence on parameter selection, reduce the computational cost and improve the ability to process high-dimensional data, thereby improving the accuracy and efficiency of anomaly detection.

[0054] Figure 2 It is a flowchart of a method for detecting poor cell indicator quality proposed in another embodiment of the present disclosure.

[0055] like Figure 2 As shown, the cell indicator quality difference detection method includes:

[0056] S201: Acquire time series data of a cell, wherein the time series data includes: multiple time points and cell indicator data corresponding to each time point.

[0057] The description of S201 can be specifically referred to the above embodiment, which will not be repeated here.

[0058] S202: Determine sample data corresponding to the cell indicator data.

[0059] The sample data refers to the data obtained after sampling and collecting the indicator data of different communities.

[0060] In the disclosed embodiment, after the sample data corresponding to the cell indicator data is determined, reliable reference information can be provided for the subsequent frequency domain transformation processing of the time series data.

[0061] S203: Perform frequency domain transformation processing on the time series data according to multiple time points and multiple sample data to obtain multiple frequency components.

[0062] That is, in the embodiment of the present disclosure, after obtaining the time series data of the cell, the sample data corresponding to the cell index data can be determined; the time series data is subjected to frequency domain transformation processing according to multiple time points and multiple sample data to obtain multiple frequency components. Thus, the indication effect of the obtained frequency components can be effectively improved by combining the sample data corresponding to the cell index data.

[0063] S204: Determine the frequency component based on reference data information under each mapping dimension.

[0064] The mapping dimension refers to the dimension selected when mapping the frequency components.

[0065] In the embodiments of the present disclosure, when selecting the mapping dimension, it can be flexibly selected according to the scene requirements and the characteristics of the frequency components, and there is no limitation on this.

[0066] Optionally, in some embodiments, the mapping dimension includes at least one of the following: a periodicity dimension, a volatility dimension, and a complexity dimension. Thus, it can be ensured that the mapping process can meet personalized application scenarios to ensure the practicality of the obtained reference data information.

[0067] Among them, the periodicity dimension is related to the main frequency of the sequence, the volatility dimension can be measured by the frequency amplitude variability in the FFT results, and the complexity dimension can be measured by the spectral entropy of the sequence.

[0068] Optionally, in some embodiments, the mapping dimension includes: a periodic dimension; when determining the reference data information based on the frequency component under each mapping dimension, it can be: determining a target frequency that makes the frequency component meet a condition, wherein the frequency component meets the condition including: the absolute value of the frequency component based on the target frequency is greater than the absolute value of the frequency component based on other frequencies; and determining the ratio between the preset value and the target frequency as the reference data information under the periodic dimension. Thus, when the mapping dimension includes the periodic dimension, the indication effect of the obtained reference data information on the periodic characteristics of the frequency component can be effectively improved.

[0069] The preset value refers to a preset value used to determine reference data information under a periodic dimension, for example, it can be 1, or it can be flexibly adjusted according to the application scenario, and there is no restriction on this.

[0070] Optionally, in some embodiments, the mapping dimension includes: a volatility dimension; when determining the frequency component based on the reference data information under each mapping dimension, it can be: obtaining the amplitude average of all frequency components; determining the amplitude variation degree of the frequency component according to the total number and the amplitude average of all frequency components; and determining the amplitude variation degree as the reference data information under the volatility dimension. Thus, when the mapping dimension includes the volatility dimension, the accuracy of the indication of the amplitude variation degree of the frequency component by the obtained reference data information can be guaranteed.

[0071] Among them, the amplitude variation degree can be used to indicate the fluctuation of the frequency component amplitude.

[0072] In the embodiments of the present disclosure, when determining the amplitude variation degree of the frequency components based on the total number and the amplitude average value of all frequency components, the total number and the amplitude average value of all frequency components can be input into a pre-trained machine learning model to obtain the corresponding amplitude variation degree. Alternatively, the amplitude variation degree of the frequency components can be determined based on a preset formula, and there is no limitation to this.

[0073] Optionally, in some embodiments, the mapping dimension includes: complexity dimension; when determining the frequency component based on the reference data information under each mapping dimension, it can be: obtaining the amplitude distribution information of the frequency component; normalizing the amplitude distribution information to obtain the target amplitude distribution; determining the spectrum entropy according to the target amplitude distribution, and determining the spectrum entropy as the reference data information under the complexity dimension. Thus, when the mapping dimension includes the complexity dimension, the obtained reference data information can ensure the indication effect of the complexity of the frequency component.

[0074] Among them, the amplitude distribution information can be used to indicate the distribution of frequency components in different dimensions.

[0075] Among them, spectral entropy can be used to reflect the uniformity of frequency component distribution.

[0076] In the embodiment of the present disclosure, when determining the spectrum entropy according to the target amplitude distribution, the target amplitude distribution may be substituted into a preset formula to calculate the corresponding spectrum entropy.

[0077] S205: Determine data points corresponding to the frequency components according to a plurality of reference data information.

[0078] That is, in the embodiments of the present disclosure, after determining that the frequency component is based on the reference data information under each mapping dimension, the data points corresponding to the frequency component can be determined according to the multiple reference data information.

[0079] S206: Using multiple reference data information as target data information.

[0080] That is, after obtaining multiple frequency components, the embodiments of the present disclosure can determine the reference data information of the frequency components based on each mapping dimension; determine the data points corresponding to the frequency components according to the multiple reference data information; and use the multiple reference data information as the target data information. Thus, the obtained target data information can indicate the relevant data of different mapping dimensions, so as to effectively improve the practicality of the target data information.

[0081] S207: Clustering the multiple data points according to the target data information to obtain at least one cluster, wherein each cluster includes: at least part of the data points.

[0082] S208: Perform quality difference detection on at least one cluster to obtain a detection result.

[0083] The description of S207 and S208 can be specifically referred to the above embodiment, which will not be repeated here.

[0084] In this embodiment, by determining sample data corresponding to the cell index data; performing frequency domain transformation processing on the time series data according to multiple time points and multiple sample data, multiple frequency components are obtained. Thus, the indication effect of the obtained frequency components can be effectively improved in combination with the sample data corresponding to the cell index data. By determining the reference data information of the frequency components based on each mapping dimension; determining the data points corresponding to the frequency components according to multiple reference data information; and using multiple reference data information as target data information. Thus, the obtained target data information can indicate the relevant data of different mapping dimensions, so as to effectively improve the practicality of the target data information.

[0085] Figure 3 It is a flowchart of a method for detecting poor cell indicator quality proposed in another embodiment of the present disclosure.

[0086] like Figure 3 As shown, the cell indicator quality difference detection method includes:

[0087] S301: Acquire time series data of a cell, wherein the time series data includes: multiple time points and cell indicator data corresponding to each time point.

[0088] S302: Perform frequency domain transformation on the time series data to obtain multiple frequency components.

[0089] S303: Mapping is performed on each frequency component to obtain a corresponding data point, wherein the data point has corresponding target data information.

[0090] The description of S301 - S303 can be specifically referred to the above embodiment, which will not be repeated here.

[0091] S304: In each iteration process, determine the similarity value between each data point and at least one cluster center.

[0092] Among them, the similarity value can be used to describe the similarity between each data point and at least one cluster center.

[0093] In the embodiment of the present disclosure, when determining the similarity value between each data point and at least one cluster center, the distance between each data point and at least one cluster center may be determined, and then the corresponding similarity value is determined based on the distance (the distance is inversely proportional to the similarity value).

[0094] S305: Determine a maximum similarity value among the multiple similarity values, and cluster each data point into a cluster where the cluster center corresponding to the maximum similarity value is located.

[0095] That is, in the clustering process, the embodiment of the present disclosure can determine the maximum similarity value among multiple similarity values, and cluster each data point into the cluster whose cluster center corresponds to the maximum similarity value, so as to ensure the reliability of the finally obtained cluster.

[0096] S306: updating the cluster center corresponding to the maximum similarity value according to the target data information of the data point, and using the updated cluster center for the next iteration process.

[0097] That is to say, in the embodiment of the present disclosure, after determining the maximum similarity value among multiple similarity values ​​and clustering each data point into the cluster where the cluster center corresponding to the maximum similarity value is located, the cluster center corresponding to the maximum similarity value can be updated according to the target data information of the data point, and the updated cluster center is used for the next iteration process to realize real-time update of the cluster center and ensure the reliability of the cluster center during each round of iteration.

[0098] S307: until the change value of the cluster center is less than a preset threshold and / or the number of iterations reaches a threshold, at least one clustering cluster is determined to have completed the iteration.

[0099] Among them, the specific values ​​of the preset threshold and the number threshold can be flexibly adjusted according to the application scenario, and there is no restriction on this.

[0100] That is to say, in the embodiment of the present disclosure, after mapping each frequency component and obtaining the corresponding data point, the similarity value between each data point and at least one cluster center can be determined in each iteration process; the maximum similarity value among multiple similarity values ​​is determined, and each data point is clustered into the cluster where the cluster center corresponding to the maximum similarity value is located; the cluster center corresponding to the maximum similarity value is updated according to the target data information of the data point, and the updated cluster center is used for the next iteration process; until the change value of the cluster center is less than the preset threshold and / or the number of iterations reaches the number threshold, at least one cluster cluster completed by the iteration is determined. In this way, the clustering processing effect of the data point can be effectively improved by combining the similarity value between each data point and at least one cluster center, and the practicality of the cluster cluster obtained by the iteration is guaranteed based on the preset threshold and / or the number threshold.

[0101] S308: Determine the spectrum characteristics corresponding to each cluster.

[0102] Among them, the spectrum characteristics can be used to indicate the characteristics of multiple data points in the cluster in the spectrum dimension.

[0103] In the embodiment of the present disclosure, when the spectrum characteristics corresponding to each cluster are determined, a reliable execution basis can be provided for the subsequent determination of the quality difference detection algorithm.

[0104] S309: Determine a quality difference detection algorithm corresponding to the spectrum characteristics.

[0105] The quality difference detection algorithm refers to the algorithm used for quality difference detection determined in the embodiments of the present disclosure, such as isolation forest, STL decomposition or DBSCAN, etc., and there is no limitation on this.

[0106] It is understandable that different spectral characteristics of clusters may affect the quality difference detection process. Therefore, in the embodiment of the present disclosure, a quality difference detection algorithm corresponding to the spectral characteristics can be determined to ensure the quality difference detection effect of the obtained quality difference detection algorithm on the corresponding clusters.

[0107] S310: Perform quality difference detection on the corresponding clusters using a quality difference detection algorithm corresponding to the spectrum characteristics to obtain a detection result.

[0108] That is to say, in the embodiment of the present disclosure, after obtaining at least one cluster, the spectrum characteristics corresponding to each cluster can be determined; the quality difference detection algorithm corresponding to the spectrum characteristics can be determined; and the quality difference detection algorithm corresponding to the spectrum characteristics can be used to perform quality difference detection on the corresponding cluster to obtain the detection result. Thus, the quality difference detection algorithm adapted to the cluster can be determined in combination with the spectrum characteristics corresponding to each cluster, thereby ensuring the accuracy and reliability of the obtained detection result.

[0109] In this embodiment, in each iteration process, the similarity value between each data point and at least one cluster center is determined; the maximum similarity value among multiple similarity values ​​is determined, and each data point is clustered into the cluster where the cluster center corresponding to the maximum similarity value is located; the cluster center corresponding to the maximum similarity value is updated according to the target data information of the data point, and the updated cluster center is used for the next iteration process; until the change value of the cluster center is less than the preset threshold and / or the number of iterations reaches the number threshold, at least one cluster cluster completed by iteration is determined. Thus, the clustering processing effect of the data point can be effectively improved in combination with the similarity value between each data point and at least one cluster center, and the practicality of the cluster cluster obtained by iteration is guaranteed based on the preset threshold and / or the number threshold. By determining the spectral characteristics corresponding to each cluster cluster; determining the quality difference detection algorithm corresponding to the spectral characteristics; using the quality difference detection algorithm corresponding to the spectral characteristics to perform quality difference detection on the corresponding cluster cluster, the detection result is obtained. Thus, the quality difference detection algorithm adapted to the cluster cluster can be determined in combination with the spectral characteristics corresponding to each cluster cluster, thereby ensuring the accuracy and reliability of the obtained detection result.

[0110] In summary of the above embodiments, the present disclosure proposes a new method for identifying poor quality of cell indicators based on spectrum clustering, which uses the "SpectraCluster Anomaly Detection" (SCAD) algorithm to achieve efficient, accurate and adaptable identification of dynamic poor quality. First, collect time series data such as key performance indicators (KPI) and key quality indicators (KQI) of the cell, including but not limited to dimensions such as coverage, interference, capacity and structure. The sequences of the key indicators of the cell are all temporal. The Fourier transform can decompose the information in the time domain into the frequency domain, and collect more complex high-dimensional information hidden in the original time series, thereby improving the ability of the poor quality identification algorithm to identify abnormal points. By applying the fast Fourier transform (FFT) to the collected time series data, the SCAD algorithm converts the original sequence data into scattered point coordinates (x, y, z) in three-dimensional space, capturing and quantifying the spectral characteristics of the sequence. Subsequently, the algorithm uses the K-means clustering method to group these scattered points, thereby identifying sequence categories with different quality characteristics. According to the clustering results, this method further selects the most suitable poor quality identification algorithm for each type of sequence from the anomaly detection algorithms such as Isolation Forest, STL (Seasonal and Trend Decomposition Using Residual Detection) decomposition and DBSCAN (Density-based Spatial Clustering Applications with Noise) for analysis. This strategy aims to use the most appropriate detection method for sequences with different quality characteristics to improve the accuracy and efficiency of poor quality identification.

[0111] The SCAD algorithm not only deeply considers the spectral characteristics of time series, significantly reduces the dependence on parameter selection through cluster analysis, but also enhances the ability to process high-dimensional data. By combining advanced data processing technology with machine learning algorithms, the SCAD algorithm provides an innovative solution for identifying poor quality indicators in cell areas.

[0112] (1) Data collection:

[0113] In this proposal, the key performance indicators (KPI) and key quality indicators (KQI) of the communication network cells are comprehensively collected, as well as multi-dimensional data such as coverage, interference, capacity and structure (such as the average number of RRC connections in the KPI indicators). The collected sequence can be expressed as s1, s2, ..., s n , where n is the time point. Each sequence s i Corresponding to a specific time point i, it reflects the cell status and performance indicators at that time point.

[0114] S = {s1, s2, ..., s n}

[0115] Continuous collection of these sequences enables real-time monitoring and analysis of cell operation, allowing accurate identification and assessment of quality differences.

[0116] Through in-depth extraction of cell scene features, such as district, scene, manufacturer, and frequency band features, we can accurately identify and distinguish the characteristics of quality problems in different scenarios. The data collection process can use sliding detection and sliding collection to ensure real-time update and calculation of key indicator data, thereby realizing real-time quality difference identification of cell operation status and service quality.

[0117] (2) Fast Fourier transform and coordinate mapping:

[0118] A. Fast Fourier Transform

[0119] In this proposal, fast Fourier transform (FFT) and coordinate mapping play a vital role. It can extract key spectral features from time series data. According to the key spectral features obtained by fast Fourier transform, the data features can be mapped into scattered points for further K-means clustering and quality difference detection.

[0120] Fast Fourier Transform (FFT) is a key step in processing time series data, especially in converting these data into the frequency domain. Through FFT transformation, the original time series data can be converted into a form that represents its frequency components.

[0121] The core of FFT transformation is to convert time series data into frequency domain representation. For the sequence data S = {s1, s2, ..., s n} (such as the average number of RRC connections in the KPI indicator), its fast Fourier transform (FFT) is expressed as:

[0122] where k = 1,…,n

[0123] Where S(k) represents the conversion of the average number of RRC connections into the kth frequency component in the frequency domain, e -πijk / n is the rotation factor, i is the imaginary unit, representing the phase change in the frequency domain, s j represents the j samples of the average number of RRC connections, and n represents the number of samples.

[0124] B. Coordinate Mapping

[0125] The FFT-transformed data is mapped to coordinates, that is, each frequency component is mapped to scattered point coordinates (x, y, z) in three-dimensional space, so that the spectral characteristics of the data can be intuitively observed and analyzed in geometric space. This step is crucial for identifying sequence categories with different quality characteristics.

[0126] X-coordinate (periodicity): Periodicity is related to the main frequency of the sequence. It can be determined by finding the value of k that maximizes |S(k)|. The X coordinate of the mapping can be expressed as:

[0127]

[0128] Y-axis (volatility): Volatility is measured by the frequency amplitude variability in the FFT results.

[0129]

[0130] Where μ is the average value of the amplitudes of all frequency components and N is the total number of frequency components.

[0131] Z-coordinate (complexity): Complexity is measured by the spectral entropy of the sequence, which reflects how evenly the frequency components are distributed. The calculation of spectral entropy remains the same:

[0132]

[0133] Here P(f) is the amplitude distribution of the frequency components normalized to ensure that the sum is 1.

[0134] After two steps of fast Fourier transform and coordinate mapping, the sequence data to be detected in the cell can be converted into scattered points (x, y, z) through fast Fourier transform and corresponding coordinate mapping, so as to facilitate the subsequent K-means clustering and model quality difference identification operations.

[0135] (3) K-means clustering

[0136] After Fast Fourier Transform (FFT) processing, the key performance and quality indicator sequence of the cell is converted into a series of three-dimensional scattered point coordinates (x, y, z), where each coordinate axis maps the spectral characteristics of the three dimensions of periodicity, volatility and complexity of the sequence. Based on this conversion, the K-means algorithm is further used to perform accurate clustering analysis on these scattered points, realizing the effective classification of sequences with different spectral characteristics. This clustering process not only highlights the natural distribution of data points in multidimensional space, but also provides a clear basis for the division of sequences with different quality characteristics. It is a key step to improve the accuracy and efficiency of sequence quality difference identification.

[0137] In the SCAD algorithm, the K-means clustering algorithm is responsible for the three-dimensional scattered point coordinate data set after the fast Fourier transform (FFT). Each point (x i ,y i , z i ).

[0138] The goal of the K-means clustering algorithm is to Divide into K clusters so that the points within the cluster are as similar as possible (i.e., the distance is close), and the points between clusters are as different as possible. The algorithm can be formalized as an optimization problem, and the objective function is to minimize the sum of the distances of all points to their corresponding cluster centers.

[0139] Mathematically, the K-means algorithm attempts to find a set of cluster centers And each data point d i Assigned to the nearest cluster center, this process can be expressed as:

[0140]

[0141] Among them, ||d i -c j || 2 Represents data point d i To cluster center c j The square of the Euclidean distance. The optimization process is completed through iterations, and each iteration consists of two steps:

[0142] A. Distribution: For each data point d i Find the nearest cluster center c j And d i Assign to the corresponding clusters.

[0143]

[0144] Among them, γ(i) represents the data point d i The index of the cluster to which it belongs.

[0145] B. Update: For each cluster, update the cluster center c j is the mean position of all data points belonging to this cluster.

[0146]

[0147] Among them, S j is assigned to cluster center c j The set of all data points |S j | is the number of points in the set.

[0148] After several iterations, until the convergence condition is met, such as the change of the cluster center is less than the predetermined threshold or the number of iterations reaches the upper limit, the algorithm ends and outputs the final cluster division.

[0149] (4) Model matching and quality difference identification

[0150] The K-means clustering algorithm is applied to these transformed scattered points to identify and classify data sets with similar characteristics. According to the specific spectral characteristics of each cluster, the most suitable quality difference detection algorithm will be applied, including isolation forest, STL decomposition or DBSCAN. Such a strategy will significantly improve the accuracy and efficiency of quality difference identification, thereby providing precise support for the maintenance and optimization of cell indicators.

[0151] A.STL: STL is an algorithm suitable for processing time series data with obvious cyclical and seasonal trends. The data of the first cluster shows a certain degree of cyclicality, but the volatility and complexity are relatively low. Using STL decomposition can help identify these time series-based cyclical and trend anomalies and see the basic patterns and abnormal fluctuations of the data more clearly. Anomaly detection can be performed on the residual part after decomposition. The residual represents the fluctuation after removing the trend and seasonality. The anomaly points usually appear as significant peaks or valleys in the residual. The Z-score anomaly detection method can be combined to identify anomalies.

[0152] In this proposal, for the collected cell time series data S = {s1, s2, ..., s n}, set it to be divided into three types of clusters, then there is S = {S1, S2, S3}, the mapping of the S1 feature sequence in three-dimensional space is characterized by relatively low periodicity, complexity and volatility, then K-means can get the result sequence of the first type of cluster as S1 (the feature sequence contained in the first type of cluster of the cell decomposition result, set the feature sequence contained in S1 as S i-1 ).

[0153] First, STL decomposition is performed, and each time series S i It is broken down into three components:

[0154] s i-1 =T i +S i +R i

[0155] Among them, T i represents the trend component, S i represents the seasonal component, R i represents the residual component.

[0156] Next, calculate the residual sequence R = {R1, R2, ..., R n}Z-score:

[0157]

[0158] Here μ R is the mean value of the residual sequence R. Ris the standard deviation of the residual sequence R. i is time point i.

[0159] Z-score in the residual series. By comparing the Z-score of each residual, outliers that deviate significantly from the average can be identified.

[0160] B.iForest: In the SCAD framework, the characteristic sequence of the second cluster S2 shows moderate periodicity and high volatility in three-dimensional space. This shows that the sequence has a certain regularity but is accompanied by a significant range of variation. Based on these characteristics, the Isolation Forest (iForest) algorithm can be applied. The Isolation Forest algorithm constructs multiple random trees (isolation trees) and uses the degree of isolation of data points as an indicator of anomalies. Abnormal points are usually more easily isolated due to their rarity in the data structure, so the degree of isolation can be used as a basis for anomaly scoring.

[0161] The anomaly detection pipeline of Isolation Forest consists of the following steps:

[0162] Construct isolation trees (iTrees): For the sequence s in the set S2 i-2 ,The algorithm recursively splits the data by randomly selecting features and their corresponding split values ​​,and constructs multiple isolated trees.

[0163] Calculating anomaly score: The algorithm calculates the anomaly score based on the path length of the data point in the isolation tree. A shorter path length means a higher probability of anomaly. The anomaly score calculation formula is:

[0164]

[0165] Among them, E(h(s i-2 )) indicates point s i-2 The average path length on the isolated tree, c(n), is a normalization factor for the average path length of the tree consisting of n points in the data set.

[0166] Marking outliers: Based on the anomaly score, the point s in the sequence i-2 If the score exceeds the preset threshold, it is considered an outlier.

[0167] C. DBSCAN: The DBSCAN algorithm is selected for anomaly detection for the third type of cluster S3. This type of cluster exhibits high periodicity, high volatility, and high complexity in three-dimensional space, indicating that its time series data may follow a complex pattern, and the identification of outliers needs to rely on the spatial density characteristics of the data.

[0168] The DBSCAN algorithm identifies clusters and outliers based on the number of neighbors each data point has within a specific "neighborhood." The algorithm hyperparameters include:

[0169] ∈: Defines the radius size of the neighborhood, that is, neighboring points are searched within this radius. Generally, this hyperparameter can be defined by the distance map method.

[0170] MinPts: The minimum number of neighboring points (including the point itself) required within the ∈-neighborhood of a point to determine whether the point is a core point. The setting of MinPts can generally be determined according to the dimension of the dataset.

[0171] The DBSCAN algorithm process can be described as follows:

[0172] Calculation of the number of points in the neighborhood: For each point s i-3 ∈ S3, the algorithm calculates the number of points within its ∈-neighborhood, that is, the number of points within a circle with a radius of ∈.

[0173] Core point identification: If the number of points within the ∈-neighborhood of point s i-3 reaches or exceeds MinPts, then s i-3 is marked as a core point.

[0174] Classification of points:

[0175] Core points: The number of points within the ∈-neighborhood ≥ MinPts;

[0176] Boundary points: The number of points within the ∈-neighborhood < MinPts but belonging to the ∈-neighborhood of a core point;

[0177] Noise points: Other points that are neither core points nor boundary points are regarded as anomalies.

[0178] That is, the anomaly detection formula can be expressed as:

[0179] Anomalies = {s i-3 ∈ S3 | s i-3 is noise}

[0180] DBSCAN can effectively process datasets with complex spatial distributions. By identifying isolated noise points, DBSCAN provides an effective method for anomaly detection of these time series data with complex distribution characteristics.

[0181] By applying differentiated anomaly detection algorithms to different clusters in the SCAD (Spectral Clustering Anomaly Detection) model, the diverse characteristics of communication network cell indicator sequence data can be adapted. The K-means clustering algorithm is used to cluster the sequence data according to key characteristics such as periodicity, volatility, and complexity. The most appropriate anomaly detection algorithm can be customized for each cluster, including STL decomposition combined with the Z-score method, the isolation forest (iForest), and the DBSCAN algorithm, which are respectively targeted at sequences with obvious periodic and seasonal trends, sequences showing medium periodicity and high volatility, and sequences showing high complexity. In this way, the SCAD model not only improves the accuracy of detecting abnormal behaviors, but also enhances the adaptability and flexibility of the model in dealing with complex and changeable communication network cell indicator data.

[0182] Based on the above embodiments, the technical points proposed in this disclosure include:

[0183] (1) A method for identifying cell quality difference based on spectrum clustering, which extracts frequency domain features through fast Fourier transform (FFT) technology after collecting time series data of key performance indicators KPI, KQI, coverage, interference, capacity and structure of the cell. The method is characterized by further studying the relationship between data, that is, defining the complexity, periodicity and volatility of the data, and classifying the spectrum characteristics in a more fine-grained manner based on the relationship between data in the three dimensions of complexity, periodicity and volatility. A mechanism and theoretical basis for selecting different quality difference identification algorithms for each type of sequence is created, which breaks the deficiency of traditional quality difference detection technology that does not delve into the relationship between data and only uses a single machine learning algorithm for anomaly detection in all scenarios.

[0184] (2) A spectral clustering anomaly detection (SCAD) algorithm converts the FFT results into frequency domain data and performs K-means clustering based on the three dimensions of complexity, periodicity, and volatility. Then, for these frequency domain data, the STL, iForest, and DBSCAN algorithms are used to identify the quality differences of the data, which are characterized by relatively low periodicity, complexity, and volatility in three-dimensional space, medium periodicity, and high volatility in three-dimensional space, and high periodicity, high volatility, and high complexity in three-dimensional space, respectively. This effectively improves the anomaly detection accuracy of the SCAD algorithm.

[0185] The advantages of the cell indicator quality difference detection method proposed in the present disclosure include but are not limited to:

[0186] Compared with traditional quality difference detection technology, the SCAD algorithm fully explores the relationship between data and defines multi-dimensional features such as periodicity, volatility and complexity of time series data. By combining Fast Fourier Transform (FFT) with K-means clustering, the data is classified based on multi-dimensional features such as periodicity, volatility and complexity of time series data, so as to achieve accurate identification and location of anomalies. This multi-dimensional feature comprehensive analysis capability is often missing in traditional methods. Secondly, by selecting appropriate anomaly detection algorithms for data clusters with different characteristics, STL decomposition combined with Z-score method, Isolation Forest (iForest) and DBSCAN, the SCAD model ensures high adaptability and accuracy, surpassing the identification methods based on single thresholds or statistical characteristics. This proposal provides an innovative solution for diverse regions, scenarios and business needs in 5G networks through efficient, accurate and adaptable quality difference identification capabilities.

[0187] By introducing a real-time rolling quality difference detection mechanism into the present invention, timely detection and repair of abnormal indicators of the cell are achieved by continuously monitoring and calculating abnormal points in the feature sequence. This mechanism not only improves the stability of the network, but also responds quickly when abnormal situations occur, ensuring the quality and reliability of communication services.

[0188] SCAD proposes a quality difference identification framework that can efficiently process high-dimensional data, combining spectral feature extraction, cluster analysis, and multi-algorithm anomaly detection. This framework can not only process complex multi-dimensional time series data, but also perform data analysis and anomaly detection at different levels, providing a new quality difference identification method.

[0189] Figure 4 It is a structural diagram of a cell indicator quality difference detection device proposed in an embodiment of the present disclosure.

[0190] like Figure 4 As shown, the cell indicator quality difference detection device 40 includes:

[0191] The acquisition module 401 is used to acquire time series data of a cell, wherein the time series data includes: multiple time points and cell indicator data corresponding to each time point;

[0192] The first processing module 402 is used to perform frequency domain transformation processing on the time series data to obtain multiple frequency components;

[0193] The second processing module 403 is used to perform mapping processing on each frequency component to obtain a corresponding data point, wherein the data point has corresponding target data information;

[0194] The third processing module 404 is used to perform clustering processing on the multiple data points according to the target data information to obtain at least one cluster, wherein each cluster includes: at least part of the data points; and

[0195] The detection module 405 is used to perform quality difference detection on at least one cluster to obtain a detection result.

[0196] It should be noted that the above explanation of the cell indicator quality difference detection method is also applicable to the cell indicator quality difference detection device of this embodiment, and will not be repeated here.

[0197] In this embodiment, by acquiring the time series data of the cell, wherein the time series data includes: multiple time points and cell indicator data corresponding to each time point; performing frequency domain transformation processing on the time series data to obtain multiple frequency components; performing mapping processing on each frequency component to obtain a corresponding data point, wherein the data point has corresponding target data information; performing clustering processing on the multiple data points according to the target data information to obtain at least one cluster cluster, wherein each cluster cluster includes: at least part of the data points; and performing quality difference detection on at least one cluster cluster to obtain a detection result. Thus, the frequency domain transformation processing and clustering processing can reduce the dependence on parameter selection, reduce the computational cost and improve the ability to process high-dimensional data, thereby improving the accuracy and efficiency of anomaly detection.

[0198] Figure 5 A block diagram of an exemplary computer device suitable for implementing embodiments of the present disclosure is shown. Figure 5 The computer device 12 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0199] like Figure 5 As shown, the computer device 12 is in the form of a general-purpose computing device. The components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 that connects various system components (including the system memory 28 and the processing unit 16).

[0200] The bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor or a local bus using any of a variety of bus structures. For example, these architectures include but are not limited to Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus and Peripheral Component Interconnection (PCI) bus.

[0201] The computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0202] The memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 34 may be used to read and write non-removable, non-volatile magnetic media ( Figure 5 Not shown, often called a "hard drive").

[0203] although Figure 5 Not shown, a disk drive for reading and writing a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing a removable non-volatile optical disk (e.g., a compact disc read only memory (Compact Disc Read Only Memory; hereinafter referred to as: CD-ROM), a digital versatile disc read only memory (Digital Video Disc Read Only Memory; hereinafter referred to as: DVD-ROM) or other optical media) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present disclosure.

[0204] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in the memory 28, such program modules 42 including but not limited to an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. The program modules 42 generally perform the functions and / or methods of the embodiments described in the present disclosure.

[0205] The computer device 12 may also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), one or more devices that enable a human body to interact with the computer device 12, and / or any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., network card, modem, etc.). Such communication may be performed via an input / output (I / O) interface 22. In addition, the computer device 12 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 20. As shown, the network adapter 20 communicates with other modules of the computer device 12 via a bus 18. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the computer device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0206] The processing unit 16 executes various functional applications and data processing by running the programs stored in the system memory 28, such as implementing the cell indicator quality difference detection method mentioned in the above embodiment.

[0207] In order to implement the above embodiments, the present disclosure further proposes a non-temporary computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the cell indicator quality difference detection method proposed in the above embodiments of the present disclosure is implemented.

[0208] In order to implement the above embodiments, the present disclosure further proposes a computer program product. When an instruction processor in the computer program product is executed, the cell indicator quality difference detection method proposed in the above embodiments of the present disclosure is executed.

[0209] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. The present disclosure is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The description and examples are to be considered exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0210] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

[0211] It should be noted that, in the description of the present disclosure, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present disclosure, unless otherwise specified, the meaning of "plurality" is two or more.

[0212] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present disclosure includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present disclosure belong.

[0213] It should be understood that the various parts of the present disclosure can be implemented in hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0214] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0215] In addition, each functional unit in each embodiment of the present disclosure may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0216] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0217] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0218] Although the embodiments of the present disclosure have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present disclosure. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present disclosure.

Claims

1. A method for detecting poor cell quality indicators, characterized in that: The method comprises: Acquire time series data of a cell, wherein the time series data includes: multiple time points and cell indicator data corresponding to each time point; Performing frequency domain transformation processing on the time series data to obtain multiple frequency components; Performing mapping processing on each of the frequency components to obtain a corresponding data point, wherein the data point has corresponding target data information; Performing clustering processing on the plurality of data points according to the target data information to obtain at least one cluster, wherein each of the clusters includes: at least part of the data points; and Perform quality difference detection on the at least one cluster to obtain a detection result.

2. The method according to claim 1, characterized in that The frequency domain transformation processing is performed on the time series data to obtain multiple frequency components, including: Determining sample data corresponding to the cell indicator data; The time series data is subjected to frequency domain transformation processing according to the multiple time points and the multiple sample data to obtain the multiple frequency components.

3. The method according to claim 1, characterized in that The mapping process is performed on each of the frequency components to obtain a corresponding data point, wherein the data point has corresponding target data information, including: Determining the frequency component is based on reference data information under each mapping dimension; Determine a data point corresponding to the frequency component according to the plurality of reference data information; A plurality of the reference data information are used as the target data information.

4. The method according to claim 3, characterized in that The mapping dimension includes at least one of the following: a periodicity dimension, a volatility dimension, and a complexity dimension.

5. The method according to claim 3, characterized in that: The mapping dimensions include: a periodic dimension; wherein the determining the frequency component is based on reference data information under each mapping dimension, including: Determine a target frequency that makes the frequency component meet a condition, wherein the frequency component meets the condition includes: an absolute value of the frequency component based on the target frequency is greater than an absolute value of the frequency component based on other frequencies; The ratio between the preset value and the target frequency is determined as the reference data information under the periodic dimension.

6. The method according to claim 3, characterized in that The mapping dimensions include: a volatility dimension; wherein the determining the frequency component is based on reference data information under each mapping dimension, including: Get the average value of the amplitude of all frequency components; Determining the amplitude variation degree of the frequency components according to the total number of all frequency components and the amplitude average value; The amplitude variation degree is determined as reference data information under the volatility dimension.

7. The method according to claim 3, characterized in that The mapping dimensions include: complexity dimensions; wherein the determining the frequency components based on reference data information under each mapping dimension includes: Acquiring amplitude distribution information of the frequency component; Normalizing the amplitude distribution information to obtain a target amplitude distribution; According to the target amplitude distribution, a spectrum entropy is determined, and the spectrum entropy is determined as reference data information under the complexity dimension.

8. The method according to claim 1, characterized in that The clustering of the plurality of data points according to the target data information to obtain at least one cluster includes: During each iteration, the similarity value of each data point with at least one cluster center is determined; Determine a maximum similarity value among the plurality of similarity values, and cluster each data point into a cluster whose cluster center corresponds to the maximum similarity value; The cluster center corresponding to the maximum similarity value is updated according to the target data information of the data point, and the updated cluster center is used for the next iteration process; Until the change value of the cluster center is less than a preset threshold and / or the number of iterations reaches a threshold, at least one clustering cluster that has been iteratively completed is determined.

9. The method according to claim 1, characterized in that: The performing quality difference detection on the at least one cluster to obtain a detection result includes: Determining a spectral characteristic corresponding to each of the clusters; Determining a quality difference detection algorithm corresponding to the spectrum characteristics; A quality difference detection algorithm corresponding to the spectral characteristics is used to perform quality difference detection on the corresponding clusters to obtain the detection result.

10. A cell indicator quality poor detection device, characterized in that: The device comprises: An acquisition module, used to acquire time series data of a cell, wherein the time series data includes: multiple time points and cell indicator data corresponding to each time point; A first processing module is used to perform frequency domain transformation processing on the time series data to obtain multiple frequency components; A second processing module is used to perform mapping processing on each of the frequency components to obtain a corresponding data point, wherein the data point has corresponding target data information; a third processing module, configured to perform clustering processing on the plurality of data points according to the target data information to obtain at least one cluster, wherein each cluster includes: at least part of the data points; and The detection module is used to perform quality difference detection on the at least one cluster to obtain a detection result.

11. A computer device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 9.

12. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: in, The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 9.

13. A computer program product, characterized in that The invention comprises a computer program which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 9.

Images