Network problem discovery method, apparatus, and computer-readable storage medium

Abstract

The application discloses a network problem discovery method, device and computer readable storage medium. The network problem discovery method comprises the following steps: acquiring network problem data, determining the degradation problem type of the network problem data; determining the data mining mode corresponding to each degradation problem type, mining the work order corresponding to the network problem data according to the data mining mode to obtain a degradation problem cell; acquiring the influence value of each degradation data in the degradation problem cell, determining a target work order according to the influence value, and discovering the network problem according to the target work order. According to the application, the target work order is determined according to the influence value of each degradation data in the degradation problem cell, and then the network problem is discovered according to the target work order. In this way, the effectiveness of network problem mining is improved.

Description

Technical Field

[0001] This application relates to the field of network technology, and in particular to a method, apparatus and computer-readable storage medium for discovering network problems. Background Technology

[0002] Currently, 5G network optimization primarily focuses on centralized optimization, analyzing and addressing network performance issues. This involves dispatching work orders for degraded cells, processing these orders, and resolving the problems. Problem discovery, as the source and target of network optimization efforts, is particularly crucial. Existing problem discovery methods include: non-real-time work orders mainly originate from the network optimization platform; real-time work orders mainly come from the EMOS quality management platform; complaint-related work orders primarily originate from customer service; drive test work orders mainly come from drive test software and drive test analysis platforms; perception-related work orders mainly come from OMC data collection and the network optimization platform; and there are also specialized work orders related to capacity, interference, and specific scenarios.

[0003] However, existing problem detection methods are limited in their means and have low quality of problem detection. Furthermore, the same community may have multiple types of work orders, and these different types are processed separately, potentially leading to inconsistencies or even conflicts in the handling of different types of work orders within the same community. Additionally, they consume significant human and material resources. Therefore, existing problem detection methods cannot effectively uncover network problems. Summary of the Invention

[0004] This application provides a method, apparatus, and computer-readable storage medium for discovering network problems, aiming to address the problem of how to effectively uncover network issues.

[0005] To achieve the above objectives, this application provides a method for discovering network problems, the method comprising:

[0006] Acquire network problem data and determine the type of degradation problem in the network problem data;

[0007] Determine the data mining method corresponding to each of the aforementioned degradation problem types, and mine the work orders corresponding to the network problem data according to the data mining method to obtain the degradation problem cells;

[0008] Obtain the impact value of each degraded data in the degraded problem cell, determine the target work order based on the impact value, and perform network problem discovery based on the target work order.

[0009] Optionally, the step of determining the data mining method corresponding to each of the degradation problem types includes:

[0010] Obtain the work order characteristics of the network problem data corresponding to each of the aforementioned degradation problem types;

[0011] The data mining method corresponding to each degradation problem type is determined based on the work order characteristics and the preset correspondence, where the preset correspondence is the correspondence between the work order characteristics and the data mining method.

[0012] Optionally, the step of obtaining the impact value of each degradation data in the degradation problem cell includes:

[0013] Obtain the frequency of co-occurrence between each of the degraded data and the target degraded data;

[0014] The coefficients and constants in the impact value prediction function for each of the deteriorated data points are determined based on the correlation frequency.

[0015] The impact value of each of the deterioration data is determined based on the coefficients and the constants.

[0016] Optionally, the step of obtaining the correlation frequency of the simultaneous occurrence of each of the degraded data and the target degraded data includes:

[0017] Construct a tree diagram based on the degradation data;

[0018] The frequency of co-occurrence of each deteriorated data point and the target deteriorated data point is determined based on the tree diagram.

[0019] Optionally, the step of determining the impact value of each of the degradation data based on the coefficient and the constant includes:

[0020] Obtain the product of the deteriorated data and the coefficient;

[0021] Obtain the sum of the product and the constant, and use the sum as the impact value of each of the degraded data.

[0022] Optionally, the step of determining the target work order based on the influence value includes:

[0023] The problem type of the work order is determined based on the impact value;

[0024] The work orders are merged according to the problem type to obtain the target work order.

[0025] Optionally, the step of specifying the duration of occurrence of each of the degraded data includes:

[0026] Obtain the occurrence duration of each of the aforementioned deterioration data points, as well as the deterioration cycle duration of the work order;

[0027] The problem type of the work order corresponding to the network problem data is determined based on the occurrence duration, the degradation cycle duration, and the impact value.

[0028] Furthermore, to achieve the above objectives, this application also provides a network problem detection device, which includes a first acquisition module, a determination module, and a second acquisition module, wherein:

[0029] The first acquisition module is used to acquire network problem data and determine the degradation problem type of the network problem data;

[0030] The determining module is used to determine the data mining method corresponding to each of the degradation problem types, and to mine the work orders corresponding to the network problem data according to the data mining method to obtain the degradation problem cells.

[0031] The second acquisition module is used to acquire the impact value of each degraded data in the degraded problem cell, determine the target work order based on the impact value, and perform network problem discovery based on the target work order.

[0032] In addition, to achieve the above objectives, this application also provides a network problem detection device, which includes a memory, a processor, and a network problem detection program stored in the memory and running on the processor. When the processor executes the network problem detection program, it implements the steps of the network problem detection method described above.

[0033] In addition, to achieve the above objectives, this application also provides a computer-readable storage medium storing a network problem detection program, which, when executed by a processor, implements the steps of the network problem detection method described above.

[0034] This application proposes a method for network problem discovery. It involves acquiring network problem data and determining the degradation type of the data; identifying the corresponding data mining method for each degradation type; mining work orders corresponding to the network problem data using the data mining method to obtain degradation problem areas; obtaining the impact value of each degradation data point in the degradation problem area; determining target work orders based on the impact values; and then performing network problem discovery based on the target work orders. This application improves the effectiveness of network problem discovery by determining target work orders based on the impact value of each degradation data point in the degradation problem area and then performing network problem discovery based on the target work orders. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of the terminal structure of the hardware operating environment involved in the embodiments of this application;

[0036] Figure 2 This is a flowchart illustrating the first embodiment of the network problem discovery method of this application;

[0037] Figure 3This is a detailed flowchart of step S30 in the second embodiment of the network problem discovery method of this application;

[0038] Figure 4 This is a schematic diagram illustrating the operation flow of the network problem detection method in this application;

[0039] Figure 5 A diagram illustrating data grouping and classification;

[0040] Figure 6 This is a schematic diagram illustrating the matching process in big data mining algorithms.

[0041] Figure 7 A schematic diagram illustrating the effect of AI-driven problem mining in 5G big data.

[0042] Figure 8 This is a diagram illustrating data mining.

[0043] Figure 9 This is a schematic diagram of an FP-tree;

[0044] Figure 10 This is a schematic diagram of a set of associated sample points based on a shared set.

[0045] Figure 11 This is a schematic diagram of the network problem detection device of this application.

[0046] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0047] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of this application.

[0048] The main solution of this application embodiment is: to obtain network problem data and determine the degradation problem type of the network problem data; to determine the data mining method corresponding to each degradation problem type, and to mine the work orders corresponding to the network problem data according to the data mining method to obtain degradation problem cells; to obtain the impact value of each degradation data in the degradation problem cell, to determine the target work order according to the impact value, and to discover network problems according to the target work order.

[0049] Existing problem detection methods suffer from limited means and low quality. Furthermore, the same neighborhood may contain multiple types of work orders, each processed separately, leading to inconsistencies and even conflicts in the handling of different types of work orders within the same neighborhood. Additionally, these methods consume significant human and material resources. Therefore, existing problem detection methods are ineffective in uncovering network problems.

[0050] This application obtains network problem data, determines the degradation type of the network problem data, identifies the corresponding data mining method for each degradation problem type, and mines the corresponding work orders based on the data mining method to obtain degradation problem areas; obtains the impact value of each degradation data in the degradation problem area, determines the target work order based on the impact value, and then discovers network problems based on the target work order. This application improves the effectiveness of network problem discovery by determining the target work order based on the impact value of each degradation data in the degradation problem area and then discovering network problems based on the target work order.

[0051] like Figure 1 As shown, Figure 1 This is a schematic diagram of the terminal device structure of the hardware operating environment involved in the embodiments of this application.

[0052] like Figure 1 As shown, the terminal device may include: a processor 1001, such as a CPU; a network interface 1004; a user interface 1003; a memory 1005; and a communication bus 1002. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen and an input unit such as a keyboard. Optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be high-speed RAM or non-volatile memory, such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0053] Those skilled in the art will understand that Figure 1 The terminal device structure shown does not constitute a limitation on the terminal device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0054] like Figure 1 As shown, the memory 1005, which is a computer-readable storage medium, may include a network problem detection program.

[0055] exist Figure 1 In the terminal device shown, network interface 1004 is mainly used for data communication with the backend server; user interface 1003 is mainly used for data communication with the client (user terminal); processor 1001 can be used to call the network problem detection program in memory 1005 and perform the following operations:

[0056] Acquire network problem data and determine the type of degradation problem in the network problem data;

[0057] Determine the data mining method corresponding to each of the aforementioned degradation problem types, and mine the work orders corresponding to the network problem data according to the data mining method to obtain the degradation problem cells;

[0058] Obtain the impact value of each degraded data in the degraded problem cell, determine the target work order based on the impact value, and perform network problem discovery based on the target work order.

[0059] refer to Figure 2 , Figure 2 This is a flowchart illustrating the first embodiment of the network problem discovery method of this application.

[0060] This application provides a method for discovering network problems. It should be noted that although the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than that shown here.

[0061] The network problem detection method in this embodiment is applied to a control server and includes the following steps:

[0062] Step S10: Obtain network problem data and determine the degradation problem type of the network problem data;

[0063] It should be noted that the degradation in this embodiment involves network optimization performance indicators. For example, if users in two base station cells handover 100 times, and all 100 handovers are successful, then the handover performance indicator is normal; if 50 handovers fail, then the handover performance indicator is abnormal, affecting users, which is also called degradation.

[0064] In this embodiment, the terminal collects and accesses network problem data from the network layer. This network problem data includes PM performance data, CM parameter data, MR coverage data, asset management engineering parameter data, complaint data, drive test data, OMC alarm data, end-to-end perception data, etc. Then, the degradation problem type of the network problem data is determined according to degradation rules. The degradation problem types include major problem categories (e.g., complaint, coverage, interference, drive test / test, performance, perception) and minor problem categories (e.g., dropped calls, handover, connection). For example, refer to... Figure 5 , Figure 5 The diagram illustrates the grouping and classification of data. The network problem data is organized into m horizontal problem categories. Each horizontal problem category includes multiple vertical problem subcategories. The vertical problem subcategories of each horizontal problem category are then classified so that vertical problem subcategories in the same row belong to the same category.

[0065] In one embodiment, the terminal pre-organizes degradation rules based on all work orders (including non-real-time work orders, real-time work orders, complaint work orders, drive test work orders, perception work orders, coverage work orders, capacity / interference / scenario work orders, etc.). For example, refer to Table 1, which lists performance indicator work order items.

[0066] Table 1

[0067] capacity Uplink PRB average utilization 5G Accessibility 5G low RRC connection establishment success rate 5G Accessibility 5G low QoS Flow establishment success rate 5G Accessibility 5G low wireless connection rate 5G Retention High 5G wireless call drop rate (cell-level) 5G Retention 5G high flow drop rate 5G Mobility 5G low handover success rate 5G Mobility 5G low gNB handover success rate 5G quality Downlink packet loss rate at cell RLC layer 5G quality High MAC layer uplink block error rate cell 5G data 5G low user downlink speed 5G

[0068] Refer to Table 2, which lists the issue items for perception-related work orders.

[0069] Table 2

[0070]

[0071]

[0072] Refer to Table 3, which lists the issue items for complaint-related work orders.

[0073] Table 3

[0074] Multi-party calling (5G) sound volume issue 5G Multi-party calling (5G) Call dropped 5G Multi-party calling (5G) Call initiation failed 5G Multi-party calling (5G) Unable to hear the call sound 5G International and Hong Kong, Macao and Taiwan roaming (5G) Unable to roam internationally 5G International and Hong Kong, Macao and Taiwan roaming (5G) Unable to make international long-distance calls 5G Domestic roaming (5G) No signal in the area 5G Domestic roaming (5G) Area signal is weak 5G Domestic roaming (5G) Slow internet speed or webpages cannot be opened 5G Domestic roaming (5G) Unable to roam within or between provinces 5G Mobile Internet (5G) No signal in the area 5G Mobile Internet (5G) Area signal is weak 5G Mobile Internet (5G) Internet connection dropped 5G Mobile Internet (5G) Slow internet speed or webpages cannot be opened 5G

[0075] In one embodiment, after the terminal collects and accesses network problem data from the network layer, it uses this degradation problem data as a data source for degradation problem mining, and then performs data access, cleaning, and storage. The terminal can configure its requirements for the collected and accessed network layer data. For example, referring to Table 4, the network layer data requirements are as follows:

[0076] Table 4

[0077]

[0078]

[0079] Step S20: Determine the data mining method corresponding to each of the degradation problem types, and mine the work orders corresponding to the network problem data according to the data mining method to obtain the degradation problem cells;

[0080] In this embodiment, the terminal determines the data mining method corresponding to each degradation problem type. Here, the degradation problem type refers to a major problem category. Then, based on the determined data mining method, data mining is performed on the work orders corresponding to the network problem data to obtain the degradation problem areas. For example, different data mining methods (including threshold method, averaging method, top method, clustering method, primary / secondary method, prediction method, induction method, iterative method, etc.) are needed to match and calculate different work order categories corresponding to different problem categories.

[0081] In one embodiment, based on the data and work order type, the data is mainly divided into continuous numerical data and discrete data, requiring different algorithms for data mining. These data mining algorithms include threshold methods, k-means algorithms, inductive methods, TOP methods, clustering methods, and maximum / minimum methods, among others.

[0082] 1. Threshold method: Primarily used for identifying degradation rules based on continuous attribute indicators. This involves setting a degradation threshold; once the threshold is reached, a problematic degraded cell is identified. For example, the radio drop rate (cell level) = (number of gNB requesting context releases - number of normal gNB requesting context releases) / (number of successful initial context establishments + number of legacy contexts + number of successful handovers + number of successful RRC connection reconstructions (non-source cells)) * 100%. Then, if the radio drop rate > 5%, the cell is considered degraded during that period, forming a degraded problem cell. Here, gNB refers to a 5G base station, and RRC (Radio Resource Control) handles information between user equipment and the base station.

[0083] 2. The k-means algorithm uses the mean as the average of multiple data sets over a period of time, reflecting the average level within that interval. For example, for RPB (physical resource block) interference data, the power levels of 273 PRBs need to be converted, summed, averaged, and then converted to the correct level, as shown below:

[0084] The set of 273 PRB interference levels in cell A during hour X is as follows:

[0085] A = {a0 a1 a2 a3 a4 a5...a} 272 dBm

[0086] The power of the i-th PRB is converted to:

[0087] The average voltage level of cell A over X hours is:

[0088]

[0089] 3. Inductive method: Based on the description of the problem type, classify the problem into different types. For example, dropped calls in the complaint category belong to the "keep" category, while slow internet speed or inability to open web pages belong to the "quality" category.

[0090] 4. The TOP method mainly performs linear normalization of the indicator K, makes TOP judgments, selects TOP problems and eliminates other problems, and mainly controls the number of deterioration.

[0091] 5. Clustering method: Density clustering is used, mainly examining the continuity of geographical locations such as MR (Measurement Report) raster coverage, and continuously expanding clusters based on connectable samples to obtain the final clustering result. The DBSCAN density clustering algorithm is used, which characterizes the density of sample distribution based on a set of "neighborhoods". Then, based on the distance location distribution, degradation judgment is made, and then continuous degradation raster clustering is performed based on the degraded raster. The steps are as follows:

[0092] Define the degradation object: Degradation problem A is designated as the degradation object;

[0093] Density direct reach: If B is located in the neighborhood of A at a certain distance R, it is considered that B has direct density reach from A;

[0094] Density reachability: If B is located in a neighborhood of A at a certain distance R, and C is located in a neighborhood of B at a certain distance R, then C is said to be density reachable from A.

[0095] Density connectivity: Density connectivity is achieved by obtaining the neighborhood correlation of A, B, and C through diffusion. Assuming the latitude and longitude of the degraded cell / grid are A(x, y), B(m, n), and C(j, i), and the diffusion radius is R centered on A, if R is greater than the distance between A and B, then distance is considered, and diffusion problems within radius R are searched around B. This is:

[0096]

[0097] Where z is the conversion constant between latitude and longitude and distance (meters). If R>Q, then clustering is performed between A and B, and further diffusion is carried out using B. It is checked whether the distance between B and C is less than R. If the condition is met, diffusion continues until there is no deterioration.

[0098] 6. Maximum-Minimum Method: For a given degradation problem, the maximum value among multiple data points within a certain time period is taken as the numerical value of the degradation problem. For example, assuming the maximum number of users between 8:00 AM and 12:00 PM is {133, 121, 200, 305}, the maximum value method yields a maximum user count of 305 during the morning peak hours.

[0099] In one embodiment, the work order characteristics of network problem data corresponding to each degradation problem type are obtained. Then, the data mining method corresponding to each degradation problem type is determined according to the work order characteristics and a preset correspondence, wherein the preset correspondence is the correspondence between work order characteristics and data mining methods. For example, refer to... Figure 6 , Figure 6 This is a schematic diagram illustrating the matching process in big data mining algorithms. Figure 6It can be seen that the data mining methods (i.e., big data mining algorithms) corresponding to each work order feature may be the same or different. Furthermore, based on the determined big data mining algorithm, problem mining is performed on non-real-time work orders, real-time work orders, complaint work orders, drive test work orders, perception work orders, coverage work orders, capacity / interference / scenario-specific work orders, etc., to identify deterioration problem areas. Then, these deterioration problem areas are summarized, organized, and presented according to problem category, problem subcategory, and deterioration granularity, and rendered using GIS.

[0100] This embodiment employs algorithms such as thresholding, averaging, topology, clustering, primary / secondary methods, prediction, induction, and iterative methods, along with the characteristics of various network problem types, to perform optimized algorithm-driven matching, resulting in more accurate network problem discovery. Simultaneously, through the discovery of degraded problem types via algorithm matching, comprehensive network problem detection can be effectively achieved, improving network quality and optimizing user experience.

[0101] Step S30: Obtain the impact value of each degraded data in the degraded problem cell, determine the target work order based on the impact value, and perform network problem discovery based on the target work order.

[0102] In this embodiment, the impact value of each degraded data in the degraded problem cell is obtained, the target work order is determined based on the impact value, and network problem discovery is performed based on the target work order. Each degraded data corresponds to an impact value prediction function, which determines the impact value of each degraded data in the degraded problem cell. This impact value is the impact value of each degraded data, not the final impact value. Therefore, it is also necessary to obtain the occurrence duration of each degraded data and the degrade cycle duration of the work order. Then, the problem type of the work order corresponding to the network problem data is determined based on the occurrence duration, degrade cycle duration, and impact value. In one embodiment, the primary and secondary degrade calculation of the work order is performed: a model trained using AI modeling is used to calculate the newly added problem degraded elements, obtaining the primary and secondary degraded problem judgment of the work order. Then, based on the degrade cycle, the sample point set of the work order to be generated is determined, and the impact value of each involved element on other elements is calculated, i.e., F(K). ij And perform the calculations as follows:

[0103]

[0104] Wherein D(K) ij The final impact value is D(K). ij The corresponding problem subclass is taken as the main deterioration problem, and the others D(K) are treated as the main deterioration problems. ij The corresponding problem subclass is considered as a secondary degradation problem, where P is the duration of the merge degradation cycle, and o is K. ij The duration of occurrence is determined by comparing the occurrence of element D(K) within the degradation cycle. ijBy analyzing the size and interrelated values, the merging and association status can be determined, thereby generating the target work order.

[0105] In one embodiment, the work order rules are configured and generated as follows: the main degradation problem of the work order is optimized, the work order merging pressure is reduced, and the work order list is output and rendered by GI.3S.

[0106] Based on the results of degradation problem mining and the parameters of the trained model, degradation time period and rules are configured, degradation elements of newly added problems are calculated, the main degradation problems leading to the optimization work orders are obtained, work order merging pressure reduction is performed, and guidance on the correlation impact and frequent itemset value is given.

[0107] Work order degradation rules:

[0108] 1) Cycle requirements: Divided into hourly, daily, weekly, and monthly levels;

[0109] 2) Degradation cycle configuration: XX days / hours / weeks / months;

[0110] 3) Degradation rules: The number of degradation cycles can be flexibly configured, and can be configured directly or in flexible segments;

[0111] 4) Merging rules: For problematic cells with multiple indicators of deterioration, merging can be performed within horizontal groups of work order categories or all indicators can be merged. The main deterioration indicator and related deterioration indicators should be noted.

[0112] 5) Filtering rules: Filter and intercept work orders based on the work orders already dispatched and processed by the system, whitelist, etc.

[0113] Furthermore, the 5G big data problem AI mining system design in this embodiment includes: a problem mining and configuration module, an AI training module, and a work order generation module, as referenced. Figure 7 , Figure 7 This is a schematic diagram illustrating the effect of AI mining of 5G big data problems, where:

[0114] Problem Mining and Configuration Module: This module primarily utilizes underlying data sources such as PM performance data, CM parameter data, MR coverage data, asset management engineering parameter data, complaint data, drive test data, OMC alarm data, and end-to-end sensing data. It employs algorithms including thresholding, averaging, topology, clustering, primary / secondary methods, prediction, induction, and iterative methods to identify and resolve issues in non-real-time work orders, real-time work orders, complaint work orders, drive test work orders, sensing work orders, coverage work orders, and specific work orders related to capacity / interference / scenario.

[0115] AI training module: First, vertical classification of similar performance degradation features and horizontal classification of work order categories are performed to obtain grid degradation elements. Then, incremental collection of sample points is carried out based on the FP-growth algorithm, and AI modeling training is performed based on linear logistic regression to calculate the coefficients of frequent itemsets and their interrelationships. Finally, the merging of multiple sample points within the degradation cycle and joint work order dispatch are performed.

[0116] Work order generation module: Based on work order rules, it generates work orders, performs value matching and merging filtering, and provides work order preprocessing functions and GIS presentation.

[0117] This embodiment acquires network problem data and determines the degradation problem type of the network problem data; it determines the data mining method corresponding to each degradation problem type, and mines the corresponding work orders based on the data mining method to obtain degradation problem areas; it obtains the impact value of each degradation data in the degradation problem area, determines the target work order based on the impact value, and then discovers network problems based on the target work order. This application improves the effectiveness of network problem discovery by determining the target work order based on the impact value of each degradation data in the degradation problem area and then discovering network problems based on the target work order.

[0118] Further, refer to Figure 3 This paper presents a second example of the network problem discovery method proposed in this application.

[0119] The difference between the second embodiment of the network problem detection method in this application and the first embodiment is that the step of obtaining the impact value of each degraded data in the degraded problem cell includes:

[0120] Step S31: Obtain the correlation frequency of each of the degraded data and the target degraded data occurring simultaneously;

[0121] Step S32: Determine the coefficients and constants in the influence value prediction function for each of the deteriorated data based on the correlation frequency;

[0122] Step S33: Determine the impact value of each of the deterioration data based on the coefficients and the constants.

[0123] In this embodiment, the terminal mines the set of frequent problem items to find the correlation frequency between pairs of problem subclasses (i.e., two deteriorated data) in the sample point set of the deteriorated problem area. Then, based on the correlation frequency, it solves the coefficients and constants in the influence value prediction function of each problem subclass corresponding to each other problem subclass (i.e. target deteriorated data). Then, when a new work order is obtained, the terminal uses the corresponding influence value prediction function to predict the influence value of each problem subclass in the new work order.

[0124] Specifically, the terminal categorizes and organizes problematic cells, then introduces the FP-growth algorithm to incrementally collect sample points, performs logical modeling and training for linear regression, obtains the parameter matrix of pairwise influences, and applies it to model adjustment. The specific steps include:

[0125] 1. Data grouping, including data mining and group training, where:

[0126] Data Mining: Based on the underlying data of all cells in the network and all access work orders, feature data mining is performed on the full range of indicators K for all cells in the network. This data is then grouped horizontally by broad categories (m groups) and vertically by degradation issues (n groups). Reference Figure 8 , Figure 8 This is a diagram illustrating data mining.

[0127] Group Training: An AI model is introduced, with each subclass of degradation problems treated as a sample point. Training is then conducted in a one-to-one correspondence among subclasses within the same area and time period. This enables the mining of problem rules based on the association between problem subclasses and the mining of frequent itemsets of problem subclasses. Simultaneously, reference... Figure 5 , Figure 5 This diagram illustrates the data grouping and classification. The data is grouped and classified according to the aforementioned similar performance degradation features. Figure 5 The system is configured with a total of m groups of horizontal work order categories and a total of n groups of vertical degradation problem classification groups.

[0128] 2. FP Sample Iterative Collection: The terminal needs to perform horizontal problem subclass influence training and vertical problem class association training, as well as frequent itemset training for single problem subclasses. To address computational redundancy caused by the increasing number of data samples, this embodiment introduces the FP-growth algorithm to incrementally stack sample points and train the mutual influence of work order subclasses and frequent itemsets.

[0129] Since AI training involves batch iterative calculations over a certain period of time, an FP-tree is first built to reduce redundant calculations of historical data.

[0130] First, sample points were collected and categorized. Each cell and each time period was designated as a sample point. Each sample point would present a variable number of degradation indicators (i.e., degradation data). Each degradation indicator was a sub-category of problems. The resulting data samples are shown in Table 5 below:

[0131] Table 5

[0132] 1 Community 1 2020-6-28 <![CDATA[{K 1a 、K 2a 、K 3b 、K 1c }]]> 2 Community 1 2020-6-29 <![CDATA[{K 2b 、K 2a 、K 1c }]]> 3 Community 1 2020-6-30 <![CDATA[{K 1a 、K 2a 、K 3b 、K 1c 、K 4c 、K 3c }]]> 4 Community 2 2020-6-28 <![CDATA[{K 3b 、K 1c }]]> 5 Community 2 2020-6-22 <![CDATA[{K 2a 、K 3b 、K 1c }]]>

[0133] Then, the FP-tree is constructed until the final FP-tree is completed. Initially, the FP-tree contains only one root node, marked with the symbol null. Subsequently, it is divided into K-trees. 2aAs a shared set, FP is gradually expanded, referencing Figure 9 , Figure 9 This is a schematic diagram of an FP-tree.

[0134] Furthermore, according to Figure 9 The FP-growth algorithm can be used to obtain a set of associated sample points based on shared sets: (See reference) Figure 10 , Figure 10 This is a schematic diagram of a set of associated sample points based on a shared set. The sample point elements are incrementally collected and summarized in the above manner, and then the logical training of linear regression is performed.

[0135] Then, linear regression is trained. The least squares method is introduced to train the linear logistic regression of the sample points, realizing the mutual influence of work order subclasses and the results of frequent itemsets.

[0136] Collect the elements from the sample points, organize the entire set, and set A is shown below:

[0137] A = {K 1a K 2a K 3a ... K na K 1b K 2b K 3b ... K nb ...K 1m K 2m K 2m ... K 3m}

[0138] The formula specifies a total of m groups for horizontal work order categories and a total of n groups for vertical degradation problem classification.

[0139] After performing mutual influence processing on the sample point element set A, we obtain the set B of the two quadrants, as shown below:

[0140]

[0141] 3. Linear Logistic Regression Modeling Training

[0142] To perform AI linear modeling, a prediction function is built using a linear combination of attributes. Each element is modeled and trained, as shown below.

[0143] Task 1:

[0144] F(K 1a ) = w 11 *K 1a +w 12 *K 1b+…+w mn *K mn +b;

[0145] Task Two:

[0146] F(K 2a ) = w 11 *K 1a +w 12 *K 1b +…+w mn *K mn +b;

[0147] Task mn:

[0148] F(K mn ) = w 11 *K 1a +w 12 *K 1b +…+w mn *K mn +b;

[0149] The elements are trained by pairwise association. Sample points with associations are marked as 1, and those without associations are marked as 0. The pairwise associations are shown below:

[0150] F(K ij ) = K ij *w ij +b ij =y ij

[0151] We need to obtain the parameter matrices w and b for use in applying the training results.

[0152] Parameter matrix X:

[0153]

[0154] w was trained using linear regression. ij With b ij This makes the two parameters closely correlated and affect the actual value.

[0155]

[0156] In this embodiment, the variance error is calculated using Euclidean distance, and the least squares method is employed, which is based on minimizing the mean square error to solve the model. In linear regression, the least squares method is used to determine the actual standard threshold, minimizing the sum of the distances from all training sample points to the actual standard threshold, thus approximating the actual impact value.

[0157] Solve for w ijWith b ij Make E (w,b) The process of minimizing P is called the least squares "parameter estimation" of the linear regression model, where P is:

[0158]

[0159] E (w,b) For w respectively ij With b ij Taking the derivative, we get:

[0160]

[0161]

[0162] Then, setting the above two equations to 0 will yield the closed-form solution for the optimal solutions of w and b.

[0163]

[0164]

[0165] in, For K 1a For K ij The average frequency of occurrence of associations.

[0166] From this, we can obtain K. 1a The parameter matrix X that affects w and b for all other elements is used to complete the training for Task 1.

[0167] By analogy, we can obtain the model results for tasks 1 to mn.

[0168] Based on the above algorithm, the coefficient matrix of the mutual influence between each pair of elements can be obtained, i.e., w ij b ij .

[0169] In determining w ij and b ij After that, w ij and b ij Substitute into the influence value prediction function F(K) ij This allows us to obtain the impact value of each deteriorated data point. In one embodiment, the product of the deteriorated data point and the coefficient is obtained, and then the sum of the product and the constant is obtained. This sum is used as the impact value of each deteriorated data point.

[0170] F(K ij ) = K ij *w ij +b ij =y ij .

[0171] This embodiment employs the FP-growth algorithm for incremental sample stacking, effectively reducing the time required for correlation calculations. By establishing an FP-tree, it achieves incremental accumulation of large amounts of data, significantly reducing the computational load of AI training iterations and improving operational efficiency. Furthermore, it utilizes a linear logistic regression algorithm to effectively mine frequent itemsets and key influencing correlation issues across multiple network problems. This algorithm also assists in generating high-value work orders and making other merging decisions, effectively optimizing the network optimization problem discovery process and greatly improving resource allocation for centralized optimization efforts.

[0172] To better illustrate the network problem detection method of this application, refer to... Figure 4 , Figure 4 This is a schematic diagram illustrating the operation flow of the network problem detection method in this application.

[0173] In this embodiment, network problem discovery incorporates data such as PM performance data, CM parameter data, MR coverage data, engineering parameter data, complaint data, drive test data, OMC alarm data, and end-to-end sensing data as the basic data sources for problem discovery. Then, the types of work orders to be discovered are determined, including non-real-time work orders, real-time work orders, complaint work orders, drive test work orders, sensing work orders, coverage work orders, and capacity / interference / scenario-specific work orders. Algorithms such as thresholding, averaging, topology, clustering, primary / secondary methods, prediction, induction, and iterative methods are used to discover degradation problems from 5G big data. The discovered degradation problems are then subjected to phased AI modeling training to obtain frequent itemset problems and the mutual influence and correlation between pairwise problems. For newly discovered degradation problems, the influence and management relationships trained by the AI ​​are invoked to merge and organize multiple types of degradation problems, including those related to cells, and identify frequent itemset problems leading to multiple types of degradation, forming optimization work orders for use in network optimization work. Specifically, the following steps are included:

[0174] 1. Degradation issue mining: First, we need to collect and access underlying data such as PM performance data, CM parameter data, MR coverage data, asset management engineering parameter data, complaint data, drive test data, OMC alarm data, and end-to-end perception data as the data source for degradation issue mining.

[0175] This research focuses on big data mining methods, applying different approaches to various work order categories. Key data mining methods include thresholding, averaging, topology, clustering, primary / secondary methods, prediction, induction, and iterative algorithms. These algorithms are used to identify and resolve issues in various work order categories, including non-real-time work orders, real-time work orders, complaint work orders, drive test work orders, perception work orders, coverage work orders, and work orders related to capacity / interference / scenario issues.

[0176] 2. AI modeling correlation research: Introducing big data and AI algorithms, the degradation problem is divided into continuous and discrete categories, and then grouped according to network optimization specifications. AI modeling is then performed to calculate the degradation correlation impact of problem subcategories and calculate the frequent itemsets of degradation problems, which are used as the logical integration of optimization work order dispatching and merging processing rules.

[0177] Degradation mining and sorting: Based on the underlying data of all cells in the network and the full number of access work orders, feature data mining of the full index K of all cells in the network is carried out, and horizontal major category grouping (m groups) and vertical degradation problem classification grouping (n groups) are performed to obtain problem elements.

[0178] Iterative sample point collection: Based on the mined correlations and frequent itemsets, the main degradation of the community degradation index is determined, and work orders are formed through merging and combining. Simultaneously, the FP-growth algorithm is introduced to incrementally overlay sample points, and to train the work order subclasses on inter-correlation influences and frequent itemsets.

[0179] Logical modeling and training of linear regression: The system performs logical training for linear regression based on the least squares method to train linear logistic regression on sample points, realizing the interrelationships and frequent itemsets of work order subcategories.

[0180] 3. Optimize work order generation: Based on the results of deterioration problem mining and the results of trained model parameters, configure the deterioration time period and rules, calculate the deterioration elements of newly added problems, obtain the main deterioration problems that lead to the optimized work orders, and perform work order merging and pressure reduction to obtain high-value comprehensive work orders.

[0181] 4. System Design and Effects: The system's functional design mainly consists of three parts: problem discovery and configuration, AI modeling and training module, and work order generation module.

[0182] Problem discovery and configuration mainly involves identifying underlying data flows and problems in the system architecture, as well as configuring algorithms and functions.

[0183] The AI ​​modeling training module performs AI modeling training based on the mined sample points;

[0184] The work order generation module is mainly responsible for generating work orders, as well as calculating and judging primary and secondary degradation, merging processing, and GIS presentation.

[0185] This embodiment employs various algorithms, including thresholding, averaging, topology, clustering, primary / secondary methods, prediction, induction, and iterative methods, to identify different types of optimization work orders. Then, AI modeling and correlation training are performed to model the impact and correlation of different major and minor problem types. The trained impact relationships and correlation results are then merged and correlated with the work orders to generate high-value network optimization work orders that significantly impact user perception, enabling efficient network optimization work. The technical effects achieved by this embodiment through the above-mentioned techniques include:

[0186] Diverse discovery methods: Traditional network problem discovery methods are relatively simple, while this embodiment uses algorithms such as threshold method, averaging method, top method, clustering method, primary and secondary method, prediction method, induction method, and iterative method to achieve efficient problem discovery of 5G big data.

[0187] Comprehensive Problem Discovery: Currently, the discovery of wireless network problems in different provinces is limited by provincial resources, resulting in significant differences in skill levels and an inability to identify major network issues affecting users. This embodiment, however, leverages a large number of 5G data sources to achieve efficient and comprehensive network problem discovery.

[0188] Problem type deduplication: Currently, the discovery, processing, and resolution of various work orders are independent. Multiple types of work orders exist within the same community, potentially leading to inconsistent or even conflicting processing approaches, resulting in multiple adjustments to the processing strategy for that community. This embodiment, through AI modeling training, identifies the most frequent and impactful problems for each work order category, as well as the main degradation factors affecting other problems. These are then merged and used for auxiliary judgment, reducing redundant processing of major work order categories and improving resource allocation.

[0189] Network optimization involves source configuration optimization: Currently, various work orders are grouped and managed separately within the province, resulting in numerous work order categories and duplicate work order cells, requiring significant manpower and resources. This embodiment integrates various substandard work orders, achieving resource integration and configuration optimization from the source of network optimization, thereby improving the efficiency of network optimization work.

[0190] Furthermore, this application also provides a network problem detection device, which includes a memory, a processor, and a network problem detection program stored in the memory and running on the processor. The device acquires network problem data, determines the degradation problem type of the network problem data, determines the data mining method corresponding to each degradation problem type, mines the work orders corresponding to the network problem data according to the data mining method to obtain degradation problem areas, obtains the impact value of each degradation data in the degradation problem area, determines the target work order based on the impact value, and performs network problem detection based on the target work order. This application improves the effectiveness of network problem detection by determining the target work order based on the impact value of each degradation data in the degradation problem area and then performing network problem detection based on the target work order.

[0191] Further, refer to Figure 11 , Figure 11 This is a schematic diagram of the network problem detection device of this application;

[0192] The network problem detection device 100 includes a first acquisition module 10, a determination module 20, and a second acquisition module 30, wherein:

[0193] The first acquisition module 10 is used to acquire network problem data and determine the degradation problem type of the network problem data;

[0194] The determining module 20 is used to determine the data mining method corresponding to each of the degradation problem types, and to mine the work orders corresponding to the network problem data according to the data mining method to obtain the degradation problem cells.

[0195] The second acquisition module 30 is used to acquire the impact value of each degraded data in the degraded problem cell, determine the target work order based on the impact value, and perform network problem discovery based on the target work order.

[0196] Furthermore, the determining module 20 includes an acquisition unit and a determining unit;

[0197] The acquisition unit is used to acquire the work order characteristics of the network problem data corresponding to each of the degradation problem types;

[0198] The determining unit is used to determine the data mining method corresponding to each of the degradation problem types based on the work order characteristics and the preset correspondence relationship, wherein the preset correspondence relationship is the correspondence relationship between the work order characteristics and the data mining method.

[0199] Furthermore, the second acquisition module 30 includes a second acquisition unit and a second determination unit;

[0200] The second acquisition unit is used to acquire the correlation frequency of each of the degraded data and the target degraded data occurring simultaneously;

[0201] The second determining unit is used to determine the coefficients and constants in the influence value prediction function for each of the deteriorated data based on the correlation frequency;

[0202] The second determining unit is further configured to determine the impact value of each of the deteriorated data based on the coefficient and the constant.

[0203] Furthermore, the second acquisition unit includes a construction subunit and a second determination word unit;

[0204] The construction subunit is used to construct a tree diagram based on the deterioration data;

[0205] The second determining subunit is used to determine the frequency of co-occurrence of each deteriorated data and the target deteriorated data based on the tree diagram.

[0206] Furthermore, the second determining unit includes a second acquiring subunit;

[0207] The second acquisition subunit is used to acquire the product of the degraded data and the coefficient;

[0208] The second acquisition subunit is further configured to acquire the sum of the product and the constant, and use the sum as the impact value of each of the degraded data.

[0209] Furthermore, the second acquisition module 30 includes a merging unit;

[0210] The second determining unit is further configured to determine the problem type of the work order based on the influence value;

[0211] The merging unit is used to merge the work orders according to the problem type to obtain the target work order.

[0212] Furthermore, the second acquisition subunit is also used to acquire the occurrence duration of each of the deteriorated data and the deterioration cycle duration of the work order;

[0213] The second acquisition subunit is further configured to determine the problem type of the work order corresponding to the network problem data based on the occurrence duration, the degradation cycle duration, and the impact value.

[0214] The implementation of the functions of each module of the network problem detection device described above is similar to the process in the above method embodiment, and will not be described in detail here.

[0215] In addition, this application also provides a computer-readable storage medium storing a network problem detection method program, which, when executed by a processor, implements the steps of the network problem detection method described above.

[0216] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0217] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0218] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0219] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0220] It should be noted that any reference signs placed between parentheses in the claims should not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. This application can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0221] Although alternative embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make further changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the alternative embodiments as well as all changes and modifications falling within the scope of this application.

[0222] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for detecting network problems, characterized in that, The method includes: Acquire network problem data and determine the type of degradation problem in the network problem data; Obtain the work order characteristics of the network problem data corresponding to each of the aforementioned degradation problem types; The data mining method corresponding to each degradation problem type is determined based on the work order characteristics and the preset correspondence, where the preset correspondence is the correspondence between the work order characteristics and the data mining method. The data mining method described above is used to mine the work orders corresponding to the network problem data to obtain the degraded problem cells. Obtain the impact value and occurrence duration of each degraded data in the degraded problem cell, as well as the degrade cycle duration of the corresponding work order; determine the problem type of the work order corresponding to the network problem data based on the impact value, occurrence duration, and degrade cycle duration; merge the work orders according to the problem type to obtain a target work order, the target work order including primary degraded problems and secondary degraded problems, as well as the corresponding associated impacts; and perform network problem discovery based on the target work order. The step of obtaining the impact value of each degraded data in the degraded problem cell includes: Using problem cells and time periods as sample points, the degradation data in the degradation problem cells are incrementally superimposed using the FP-growth algorithm to construct an FP tree, and the correlation frequency of each degradation data and the target degradation data occurring simultaneously is obtained based on the FP tree. Based on the correlation frequency, the deteriorated data is trained using linear regression with the least squares method to obtain the coefficients and constants in the influence value prediction function for each deteriorated data. Obtain the product of the deteriorated data and the coefficient; Obtain the sum of the product and the constant, and use the sum as the influence value of each of the degraded data, wherein the influence value is used to represent the influence value of each degraded data on the target degraded data.

2. A network problem detection device, characterized in that, The network problem detection device includes a first acquisition module, a determination module, and a second acquisition module, wherein: The first acquisition module is used to acquire network problem data and determine the degradation problem type of the network problem data; The determining module is used to obtain the work order features of the network problem data corresponding to each of the degradation problem types; determine the data mining method corresponding to each of the degradation problem types according to the work order features and the preset correspondence relationship, wherein the preset correspondence relationship is the correspondence relationship between the work order features and the data mining method; and mine the work orders corresponding to the network problem data according to the data mining method to obtain the degradation problem cells. The second acquisition module is used to acquire the impact value and occurrence duration of each degraded data in the degraded problem cell, as well as the degrade cycle duration of the corresponding work order; determine the problem type of the work order corresponding to the network problem data based on the impact value, occurrence duration, and degrade cycle duration; merge the work orders according to the problem type to obtain a target work order, the target work order including primary degraded problems and secondary degraded problems, as well as corresponding associated impacts, and perform network problem discovery based on the target work order; the step of acquiring the impact value of each degraded data in the degraded problem cell includes: using the problem cell and time period as sample points, through FP The -growth algorithm incrementally overlays sample points of degraded data in the degraded problem cell to construct an FP-tree, and obtains the association frequency of each degraded data and target degraded data occurring simultaneously based on the FP-tree; based on the association frequency, linear regression training is performed on the degraded data using the least squares method to obtain the coefficients and constants in the influence value prediction function of each degraded data; the product of the degraded data and the coefficients is obtained; the sum of the product and the constant is obtained, and the sum is used as the influence value of each degraded data, wherein the influence value is used to represent the influence value of each degraded data on the target degraded data.

3. A network problem detection device, characterized in that, The network problem detection device includes a memory, a processor, and a network problem detection program stored in the memory and running on the processor. When the processor executes the network problem detection program, it implements the steps of the method as described in claim 1.

4. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a network problem detection program, which, when executed by a processor, implements the steps of the method as described in claim 1.

Images