Call quality poor cause positioning method and system, computer device and storage medium
By constructing a knowledge graph of poor call quality and a Bayesian network, and combining real-time data, the root cause of poor call quality is located digitally and probabilistically, solving the problems of high time and manpower costs in existing technologies, and achieving efficient and accurate root cause location of poor call quality.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA UNITED NETWORK COMM GRP CO LTD
- Filing Date
- 2023-06-02
- Publication Date
- 2026-04-17
AI Technical Summary
Current technologies for locating the root cause of poor call quality incur significant time and manpower costs, making it difficult to efficiently analyze the reasons for poor call quality.
By acquiring the quality defect analysis manual and historical case data, a quality defect knowledge graph is constructed. By using Bayesian network technology combined with real-time data, the root causes of quality defect experience entities are determined, realizing the digitalization and probabilistic positioning of expert experience.
It improves the efficiency and accuracy of identifying the root causes of poor network quality, reduces manpower and time costs, and enhances network operation and maintenance efficiency and user experience.
Smart Images
Figure CN116680411B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of communication technology, and in particular to a method, system, computer device, and storage medium for locating the root cause of poor call quality. Background Technology
[0002] Wireless voice call service is a basic service provided by telecommunications operators, and the quality of wireless voice call service directly affects users' evaluation of the telecommunications operators' services. In actual use, poor voice call quality, including connection failures, dropped calls, and mumbled words, still causes certain inconveniences for users.
[0003] Wireless call quality poor root cause localization refers to identifying the cause of poor call quality as quickly as possible when such issues occur. This assists frontline network optimization engineers in resolving the root cause of the poor call quality in the shortest possible time, thereby restoring call quality. Clearly, continuously improving the efficiency and accuracy of voice call quality poor root cause localization is of great significance for operators to improve network operation and maintenance, enhance the user's voice call experience, and reduce user churn.
[0004] Currently, the primary approach is to leverage the expertise of frontline network optimization specialists to analyze specific cases and pinpoint the root causes of performance issues. However, given the limited time and energy available to network optimization experts, requiring them to analyze every case individually would undoubtedly incur significant time and manpower costs. Summary of the Invention
[0005] The technical problem to be solved by this invention is that the existing technology for locating the root cause of poor quality results in expensive time and manpower costs.
[0006] To address the aforementioned shortcomings of existing technologies, the following solutions are provided:
[0007] In a first aspect, the present invention provides a method for locating the root cause of poor call quality, comprising: acquiring historical data of a poor call quality analysis manual and poor call quality cases; acquiring a poor call quality knowledge graph based on the historical data of the poor call quality analysis manual and poor call quality cases; acquiring, based on the poor call quality knowledge graph, poor call quality cause entities related to a first poor call quality experience entity as first poor call quality cause entities, to form a set of first poor call quality cause entities, the set of first poor call quality cause entities including multiple first poor call quality cause entities; and determining the root cause of the first poor call quality experience entity in the set of first poor call quality cause entities based on the probability of each first poor call quality cause entity in the set of first poor call quality cause entities. Wherein, the poor call quality entities in the poor call quality knowledge graph include poor call quality experience entities and poor call quality cause entities. The first poor call quality experience entity is any one of the poor call quality experience entities.
[0008] Specifically, a quality defect knowledge graph is obtained based on historical data from the quality defect analysis manual and quality defect cases. This includes: identifying quality defect entities and the relationships between them based on the historical data from the quality defect analysis manual and quality defect cases; and constructing the quality defect knowledge graph based on the quality defect entities and the relationships between them. The relationships between quality defect entities include inclusion relationships and causal relationships.
[0009] Specifically, the process involves obtaining entities related to the first poor-quality experience entity from the poor-quality knowledge graph as first poor-quality cause entities, forming a set of first poor-quality cause entities. This includes: obtaining one or more poor-quality cause entities with a causal relationship to the first poor-quality experience entity from the poor-quality knowledge graph as second poor-quality cause entities; and obtaining third poor-quality cause entities from the poor-quality knowledge graph. The relationship between the third poor-quality cause entities is a causal relationship, or the relationship between the third poor-quality cause entity and the second poor-quality cause entity is a causal relationship. Here, the first poor-quality cause entity is either the second or the third poor-quality cause entity. The set of first poor-quality cause entities includes both the second and third poor-quality cause entities.
[0010] Specifically, based on the probability of each first poor quality cause entity in the set of first poor quality cause entities, the root cause of the first poor quality experience entity is determined in the set of first poor quality cause entities. This includes: constructing a causal chain for the first poor quality experience entity based on the relationships between third poor quality cause entities, the relationships between the third poor quality cause entity and the second poor quality cause entity, and the relationships between the second poor quality cause entity and the first poor quality experience entity; obtaining a set of potential root causes of the first poor quality experience entity based on the causal chain; obtaining the probability of occurrence of each potential root cause in the set of potential root causes of the first poor quality experience entity; and determining the root cause of the first poor quality experience entity based on the probability of occurrence of each potential root cause in the set of potential root causes of the first poor quality experience entity. Wherein, the poor quality entity at the very end of each causal chain of the first poor quality experience entity is the first poor quality experience entity. The potential root cause of the first poor quality experience entity is the third poor quality cause entity at the very beginning of each causal chain of the first poor quality experience entity.
[0011] Specifically, obtaining the probability of each potential root cause in the set of potential root causes of the first poor quality experience entity includes: constructing a Bayesian network for the first poor quality experience entity based on the causal chain of all first poor quality experience entities; obtaining initial values for the probabilities of each leaf node and each intermediate node in the Bayesian network of the first poor quality experience entity based on historical data from the poor quality analysis manual and poor quality cases; updating the probabilities of each leaf node and each intermediate node in the Bayesian network of the first poor quality experience entity based on real-time data related to the first poor quality experience entity; and determining the probability of each potential root cause of the first poor quality experience entity based on the updated probabilities of each leaf node and each intermediate node in the Bayesian network of the first poor quality experience entity. Here, the root node of the Bayesian network corresponds to the first poor quality experience entity, the leaf nodes of the Bayesian network correspond to the potential root causes of the first poor quality experience entity, and the intermediate nodes of the Bayesian network correspond to the first poor quality cause entity between the first poor quality experience entity and its potential root causes. The real-time data related to the first poor quality experience entity includes real-time data of the data collection base station operating parameters, real-time data of performance management, real-time data of configuration management, and real-time data of measurement reports related to the first poor quality experience entity.
[0012] Specifically, based on historical data from the quality difference analysis manual and quality difference cases, initial values for the probabilities of each leaf node and each intermediate node in the Bayesian network of the first quality difference experience entity are obtained. This includes: obtaining the prior probabilities of the leaf nodes and the conditional probability tables corresponding to the intermediate nodes in the Bayesian network of the first quality difference experience entity, based on historical data from the quality difference analysis manual and quality difference cases. The prior probability values of the leaf nodes in the Bayesian network of the first quality difference experience entity are used as the initial values for the probabilities of each leaf node in the Bayesian network of the first quality difference experience entity. Furthermore, the probability values in the conditional probability tables corresponding to the intermediate nodes in the Bayesian network of the first quality difference experience entity are used as the initial values for the probabilities of each intermediate node in the Bayesian network of the first quality difference experience entity.
[0013] Specifically, determining the root cause of the first poor-quality experience entity based on the probability of occurrence of each potential root cause in the set of potential root causes includes: setting a first probability threshold; and determining the root cause of the first poor-quality experience entity from its potential root causes based on the first probability threshold. Wherein, the sum of the probabilities of the root causes of the first poor-quality experience entity is not less than the first probability threshold.
[0014] Secondly, the present invention provides a system for locating the root cause of poor quality, comprising a first acquisition module, a second acquisition module, a third acquisition module, and a determination module. The first acquisition module is configured to acquire historical data of a poor quality analysis manual and poor quality cases. The second acquisition module is configured to acquire a poor quality knowledge graph based on the historical data of the poor quality analysis manual and poor quality cases. The third acquisition module is configured to acquire, based on the poor quality knowledge graph, poor quality cause entities related to a first poor quality experience entity as first poor quality cause entities, to form a set of first poor quality cause entities, the set of first poor quality cause entities including multiple first poor quality cause entities. The poor quality entities in the poor quality knowledge graph include poor quality experience entities and poor quality cause entities. The first poor quality experience entity is any one of the poor quality experience entities. The determination module is configured to determine the root cause of the first poor quality experience entity in the set of first poor quality cause entities based on the probability of each first poor quality cause entity in the set of first poor quality cause entities. The poor quality entities in the poor quality knowledge graph include poor quality experience entities and poor quality cause entities. The first poor quality experience entity is any one of the poor quality experience entities.
[0015] Specifically, the second acquisition module is configured to acquire poor-quality entities and the relationships between them based on historical data from the poor-quality analysis manual and poor-quality cases. The relationships between poor-quality entities include inclusion relationships and causal relationships. Furthermore, a poor-quality knowledge graph is constructed based on the poor-quality entities and their relationships.
[0016] Specifically, the third acquisition module is configured to acquire one or more quality-poor cause entities that have a causal relationship with the first quality-poor experience entity, based on the quality-poor knowledge graph, as the second quality-poor cause entity. It also acquires a third quality-poor cause entity based on the quality-poor knowledge graph. The relationship between the third quality-poor cause entities is a causal relationship, or the relationship between the third quality-poor cause entity and the second quality-poor cause entity is a causal relationship. The first quality-poor cause entity is either the second or the third quality-poor cause entity. The set of first quality-poor cause entities includes both the second and third quality-poor cause entities.
[0017] Specifically, the determination module includes a causal relationship chain construction unit, a potential root cause set acquisition unit, a probability acquisition unit, and a root cause determination unit. The causal relationship chain construction unit is configured to construct a causal relationship chain for the first poor quality experience entity based on the relationships between the third poor quality cause entities, the relationships between the third poor quality cause entity and the second poor quality cause entity, and the relationships between the second poor quality cause entity and the first poor quality experience entity. The poor quality entity at the very end of each causal relationship chain for the first poor quality experience entity is the first poor quality experience entity. The potential root cause set acquisition unit is configured to acquire a set of potential root causes for the first poor quality experience entity based on the causal relationship chain. The potential root cause of the first poor quality experience entity is the third poor quality cause entity at the very beginning of each causal relationship chain. The probability acquisition unit is configured to acquire the probability of each potential root cause appearing in the set of potential root causes for the first poor quality experience entity. The root cause determination unit is configured to determine the root cause of the first poor quality experience entity based on the probability of each potential root cause appearing in the set of potential root causes for the first poor quality experience entity.
[0018] Specifically, the determination module also includes a Bayesian network construction unit. This unit is configured to construct a Bayesian network for the first poor quality experience entity based on the causal chain of all first poor quality experience entities. The root node of the Bayesian network corresponds to the first poor quality experience entity, the leaf nodes correspond to the potential root causes of the first poor quality experience entity, and the intermediate nodes correspond to the first poor quality cause entities between the first poor quality experience entity and its potential root causes. The probability acquisition unit is configured to acquire the initial value of the probability of each node in the Bayesian network of the first poor quality experience entity based on historical data from the poor quality analysis manual and poor quality cases. Furthermore, it updates the probability of each node in the Bayesian network of the first poor quality experience entity based on real-time data related to the first poor quality experience entity. This data includes real-time data on the base station operating parameters, performance management, configuration management, and measurement reports related to the first poor quality experience entity. The root cause determination unit is configured to determine the probability of the occurrence of the potential root cause of each first-quality-poor-experience entity based on the probability of each node in the updated Bayesian network of the first-quality-poor-experience entity.
[0019] Specifically, the probability acquisition unit is configured to obtain the prior probabilities of the leaf nodes in the Bayesian network of the first poor-quality experience entity and the conditional probability table corresponding to the intermediate nodes in the Bayesian network of the first poor-quality experience entity, based on the historical data of the poor-quality analysis manual and poor-quality cases. The prior probability values of the leaf nodes in the Bayesian network of the first poor-quality experience entity are used as the initial values of the probability of each leaf node in the Bayesian network of the first poor-quality experience entity. Furthermore, the probability values in the conditional probability table corresponding to the intermediate nodes in the Bayesian network of the first poor-quality experience entity are used as the initial values of the probability of each intermediate node in the Bayesian network of the first poor-quality experience entity.
[0020] Specifically, the root cause determination unit is configured to: set a first probability threshold; and determine the root cause of the first poor quality experience entity from among its potential root causes based on the first probability threshold. The sum of the probabilities of the root causes of the first poor quality experience entity is not less than the first probability threshold.
[0021] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the above-described method for locating the root cause of poor call quality.
[0022] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor performs the above-described method for locating the root cause of poor call quality.
[0023] The method, system, computer equipment, and storage medium for locating the root causes of poor call quality provided by this invention utilize knowledge graph technology to digitize expert experience related to locating the root causes of poor call quality, forming digital experts and improving the efficiency of using expert experience. Simultaneously, by combining Bayesian networks, the expert experience is probabilistically processed, and possible root causes of poor call quality are ranked according to probability. The effective combination of these two methods ultimately achieves a data- and knowledge-driven technology for locating the root causes of poor call quality. Attached Figure Description
[0024] Figure 1 This is a flowchart of a method for locating the root cause of poor call quality in an embodiment of the present invention;
[0025] Figure 2 This is a schematic diagram illustrating the relationship between entities with poor quality of experience in an embodiment of the present invention;
[0026] Figure 3 This is a schematic diagram illustrating the relationship between entities causing poor quality in an embodiment of the present invention;
[0027] Figure 4 This is a flowchart of another method for locating the root cause of poor call quality in an embodiment of the present invention;
[0028] Figure 5 This is a flowchart of another method for locating the root cause of poor call quality in an embodiment of the present invention;
[0029] Figure 6 This is a flowchart of another method for locating the root cause of poor call quality in an embodiment of the present invention;
[0030] Figure 7 This is a structural block diagram of a call quality poor cause localization system according to an embodiment of the present invention;
[0031] Figure 8 This is a structural block diagram of another call quality poor cause localization system in an embodiment of the present invention;
[0032] Figure 9 This is a structural block diagram of another call quality poor cause localization system in an embodiment of the present invention;
[0033] Figure 10 This is a structural block diagram of a computer device according to an embodiment of the present invention. Detailed Implementation
[0034] To enable those skilled in the art to better understand the technical solution of the present invention, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.
[0035] It is understood that the specific embodiments and accompanying drawings described herein are merely for explaining the invention and are not intended to limit the invention.
[0036] It is understood that, without conflict, the various embodiments and features in the embodiments of the present invention can be combined with each other.
[0037] It is understood that, for ease of description, only the parts related to the present invention are shown in the accompanying drawings, while the parts unrelated to the present invention are not shown in the drawings.
[0038] It is understood that each unit or module involved in the embodiments of the present invention may correspond to only one entity structure, or may be composed of multiple entity structures, or multiple units or modules may be integrated into one entity structure.
[0039] It is understood that, without conflict, the functions and steps marked in the flowcharts and block diagrams of this invention may occur in a different order than that marked in the accompanying drawings.
[0040] It is understood that the flowcharts and block diagrams of this invention illustrate the possible architecture, functions, and operations of systems, apparatuses, devices, and methods according to various embodiments of this invention. Each block in the flowchart or block diagram may represent a unit, module, program segment, or code, containing executable instructions for implementing the specified function. Furthermore, each block or combination of blocks in the block diagram and flowchart can be implemented using a hardware-based system to achieve the specified function, or using a combination of hardware and computer instructions.
[0041] It is understood that the units and modules involved in the embodiments of the present invention can be implemented by software or by hardware. For example, the units and modules can be located in a processor.
[0042] Embodiments of the present invention provide a method for locating the root cause of poor call quality, such as... Figure 1 As shown, the method includes steps 101 to 104.
[0043] Step 101: Obtain historical data from the quality difference analysis manual and quality difference cases.
[0044] It's understandable that a call quality analysis manual is a summary of call quality analysis by experts. Historical data on call quality issues refers to the complete record of the experts' root cause analysis, diagnostic process, diagnostic results, and feedback on each case. Based on this data, it's possible to extract valuable expert knowledge and experience.
[0045] The historical data in the Quality Deviation Analysis Manual and Quality Deviation Cases records a large number of quality deviation phenomena and the causes of these phenomena. Therefore, the root cause of quality deviation phenomena can be located based on the historical data in the Quality Deviation Analysis Manual and Quality Deviation Cases.
[0046] Step 102: Obtain a quality defect knowledge graph based on the historical data of the quality defect analysis manual and quality defect cases.
[0047] In some embodiments, poor-quality entities and the relationships between them can be obtained based on historical data from the poor-quality analysis manual and poor-quality cases, and a poor-quality knowledge graph can be constructed based on these entities and their relationships. Poor-quality entities include poor-quality experience entities and poor-quality cause entities. Relationships between poor-quality entities include inclusion relationships and causal relationships.
[0048] Understandably, to reuse expert knowledge, it needs to be stored according to a data structure that computers can understand. There are two equivalent formal representations of knowledge graphs. The first approach considers a knowledge graph to be represented as a directed acyclic graph, i.e. in, It is a collection of entities, recording all entities involved in the knowledge graph. The first approach is to define a set of relations, recording all relations involved in the knowledge graph. The second approach considers a knowledge graph to be represented as a set of triples, i.e.
[0049] Understandably, the numerous quality-poor phenomena recorded in the historical data of the quality-poor analysis manual and quality-poor case studies can be considered as entities representing quality-poor experiences, while the causes of these phenomena can be considered as entities representing quality-poor causes. In other words, this constitutes the entity set of the quality-poor knowledge graph. It mainly includes two types of entities: the first type is entities with poor quality experience, denoted as... This includes instances of poor call quality experienced by users, such as mumbled words, dropped calls, etc. The second category is entities related to the reasons for poor call quality, denoted as... It includes potential causes of various types of poor quality, such as 3D interference and weak coverage.
[0050] Understandably, the relationships between different entities representing poor quality experiences can be inclusion relationships, and the relationships between different entities representing causes of poor quality can be either causal or inclusion relationships. That is, the set of relationships in the quality-poor knowledge graph. There are two types of relationships: containment relationships (denoted as `contains`) and causes relationships (denoted as `causes`). For containment relationships, there are two possible scenarios: first, a poor quality experience entity may contain another poor quality experience entity; second, a poor quality cause entity may also contain another poor quality cause entity. For causes relationships, there are also only two scenarios: first, a poor quality cause entity may cause a poor quality experience entity; second, a poor quality cause entity may cause another poor quality cause entity. In the embodiments of this invention, there is no potential cause relationship between poor quality experience entities.
[0051] For example, based on the historical data of the quality defect analysis manual and quality defect cases, we can uncover, for instance, Figure 2 The entities exhibiting poor quality experiences and the relationships between them, as well as... Figure 3 The entities that cause poor quality and the relationships between them are shown.
[0052] like Figure 2 As shown, the entity "poor call quality experience" can include entities "poor call setup quality experience", "poor call quality experience during the call", and "poor call end quality experience". The entity "poor call setup quality experience" can include the entity "unable to connect", the entity "poor call quality experience during the call" can include the entity "swallowing words and sentences", and the entity "poor call end quality experience" can include the entity "dropped call".
[0053] like Figure 3 As shown, the entity "Coverage Issues" can include the entities "Weak Coverage," "Overlapping Coverage," and "Coverage Outside the Designated Area." The entity "Inappropriate Antenna Downtilt Angle Setting" can lead to the entities "Weak Coverage," "Overlapping Coverage," and "Coverage Outside the Designated Area."
[0054] Step 103: Obtain the entities that are related to the first poor quality experience entity from the poor quality knowledge graph as the first poor quality cause entities, so as to form a set of first poor quality cause entities.
[0055] Understandably, the first poor quality experience entity is any one of the poor quality experience entities. The set of first poor quality cause entities includes multiple first poor quality cause entities.
[0056] In some embodiments, such as Figure 4 As shown, the implementation method of step 103 may include steps 401 to 402.
[0057] Step 401: Obtain one or more quality-poor cause entities that are related to the first quality-poor experience entity based on the quality-poor knowledge graph, and use them as the second quality-poor cause entities.
[0058] Understandably, the number of entities causing the second quality defect can be one or more.
[0059] Step 402: Obtain the third entity responsible for poor quality based on the knowledge graph of poor quality.
[0060] Understandably, the number of third-quality causal entities can be one or more. When there is only one third-quality causal entity, the relationship between the third-quality causal entity and the second-quality causal entity is a causal relationship. When there are multiple third-quality causal entities, the relationship between different third-quality causal entities is a causal relationship, or the relationship between one of the multiple third-quality causal entities and the second-quality causal entity is a causal relationship.
[0061] In this case, the first poor quality cause entity can be either the second or the third poor quality cause entity. The set of first poor quality cause entities can include both the second and third poor quality cause entities.
[0062] For example, as shown in equation (1), taking the first poor quality experience entity as the poor quality experience entity "dropped call" and the second poor quality cause entity as the poor quality cause entity "loss of connection with terminal during handover", the third poor quality cause entity can include the poor quality cause entity "over-coverage" that directly causes the second poor quality cause entity (poor quality cause entity "loss of connection with terminal during handover") and the poor quality cause entity "insufficient site resources" that directly causes the poor quality cause entity "over-coverage". It can be understood that the relationship between the poor quality cause entity "insufficient site resources" and the second poor quality cause entity (poor quality cause entity "loss of connection with terminal during handover") is an indirect cause relationship, that is, the poor quality cause entity "insufficient site resources" indirectly causes the second poor quality cause entity (poor quality cause entity "loss of connection with terminal during handover").
[0063]
[0064] According to equation (1), the set of entities causing the first poor quality includes entities causing poor quality such as “loss of connection with terminal during handover”, “overlapping coverage”, and “insufficient site resources”.
[0065] Step 104: Based on the probability of each first poor quality cause entity in the set of first poor quality cause entities, determine the root cause of the first poor quality experience entity in the set of first poor quality cause entities.
[0066] In some embodiments, such as Figure 5 As shown, the implementation method of step 104 may include steps 501 to 504.
[0067] Step 501: Based on the relationships between the third poor quality cause entities, the relationships between the third poor quality cause entities and the second poor quality cause entities, and the relationships between the second poor quality cause entities and the first poor quality experience entities, construct a causal relationship chain for the first poor quality experience entity.
[0068] For example, based on the aforementioned poor quality knowledge graph, the following three rules can be defined.
[0069] Rule 1: For containment and causation relations, define their corresponding inverse relations. That is: Containment -1 =Included; resulting in -1 =Attribution. Where for any inverse relation r, the following holds:
[0070]
[0071] Rule 2: If Then a causal chain can be constructed. Understandably, entities representing poor-quality experiences will only appear at the very end of the causal chain, while the other positions will be entities representing the causes of poor-quality experiences.
[0072] Rule 3: For causal chains If there are no nodes that can cause A, and Z is a node representing a poor experience, then this causal chain can be called a complete causal chain. The node A at the very beginning of this causal chain is the potential root cause of Z.
[0073] In the embodiments of the present invention, after obtaining the complete causal relationship chain of rule three, the corresponding potential root cause node A can be inferred based on the known final poor quality experience node Z.
[0074] Understandably, in step 501, the lowest quality entity in the causal chain of each first quality-poor experience entity is the first quality-poor experience entity.
[0075] Step 502: Obtain the set of potential root causes of the first poor quality experience entity based on the causal relationship chain of the first poor quality experience entity.
[0076] Understandably, the potential root cause of the first poor quality experience entity is the third poor quality cause entity at the very beginning of the causal relationship chain of each first poor quality experience entity. For example, if Equation (1) is the complete causal relationship chain of the first poor quality experience entity, then the poor quality cause entity "insufficient site resources" is the potential root cause of the first poor quality experience entity "dropped call".
[0077] Based on the steps described above, the complete set of causal relationship chains CC for the first poor-quality experience entity can be obtained. And it satisfies equation (2).
[0078] For a complete causal chain, and its final end is the first poor quality experience entity (2).
[0079] After obtaining the complete set of causal relationship chains CC of the first poor quality experience entity, iterate through CC. i At the very front, we can obtain the set of potential root causes of the first poor quality experience entity, which is denoted as RC.
[0080] Step 503: Obtain the probability of each potential root cause appearing in the set of potential root causes of the first poor quality experience entity.
[0081] Understandably, in order to determine the root cause of the first poor quality experience entity, in addition to knowing the set RC of potential root causes of the first poor quality experience entity, it is also necessary to know the probability of the occurrence of all potential root causes in the set RC, and sort them according to the magnitude of the probability, and give the top k potential root causes with the highest probability, so as to determine the root cause of the first poor quality experience entity.
[0082] In this case, knowledge graph technology can be combined with Bayesian network technology. Knowledge-driven methods can be used to graph expert knowledge, and data-driven Bayesian network technology can be combined to probabilistically generate low-quality knowledge graphs and rank potential root causes.
[0083] In some embodiments, such as Figure 6 As shown, the implementation method of step 503 may include steps 601 to 604.
[0084] Step 601: Construct a Bayesian network of the first-quality-poor-experience entities based on the causal relationship chains of all first-quality-poor-experience entities.
[0085] Understandably, the root node of the Bayesian network corresponds to the first poor quality experience entity. The leaf nodes of the Bayesian network correspond to the latent root causes of the first poor quality experience entity. The intermediate nodes of the Bayesian network correspond to the first poor quality cause entity between the first poor quality experience entity and its latent root causes. Understandably, intermediate nodes can correspond to the third or second poor quality cause entity.
[0086] For example, traversing cc i And obtain cc i After considering all the entities and relationships involved, these elements can be used to construct subgraphs of a quality-discretionary knowledge graph. And utilize subgraphs Building Bayesian Networks The network structure. Due to the subgraph All edges in the subgraph represent relations; therefore, the subgraph... It can be directly converted into a Bayesian network. Bayesian networks The set of nodes is a subgraph. A collection of entities, The set of edges is The set of relationships that lead to [something].
[0087] Step 602: Obtain the initial values of the probability of each leaf node and each intermediate node in the Bayesian network of the first poor quality experience entity based on the poor quality analysis manual and historical data of poor quality cases.
[0088] Understandably, for Bayesian networks, in addition to defining their structure, it is also necessary to initialize the parameters (i.e., probabilities) of each node. In some embodiments, step 602 can be implemented as follows: Based on the quality difference analysis manual and historical data of quality difference cases, obtain the prior probabilities of the leaf nodes in the Bayesian network of the first quality difference experience entity and the conditional probability table corresponding to the intermediate nodes in the Bayesian network of the first quality difference experience entity. Use the prior probability values of the leaf nodes in the Bayesian network of the first quality difference experience entity as the initial value of the probability of each leaf node in the Bayesian network of the first quality difference experience entity. And, use the probability values in the conditional probability table corresponding to the intermediate nodes in the Bayesian network of the first quality difference experience entity as the initial value of the probability of each intermediate node in the Bayesian network of the first quality difference experience entity.
[0089] For example, in the subgraph In the context of the Bayesian network, the entity "unreasonable antenna angle setting" can serve as a potential root cause of the poor quality experience entity "dropped calls." Therefore, when reasoning about the potential root causes of "dropped calls," the entity "unreasonable antenna angle setting" becomes a leaf node in the Bayesian network. Initializing this leaf node requires knowing the prior probability of the entity "unreasonable antenna angle setting," which can be denoted as P(unreasonable antenna angle setting), representing the number of times the entity "unreasonable antenna angle setting" occurs in the subgraph. The proportion of all nodes in the complete set U is as shown in equation (3).
[0090]
[0091] For example, for intermediate nodes, it is necessary to initialize their corresponding conditional probability table. For instance, if there are three direct causes (i.e., the node preceding the "dropped call" node in three complete causal chains, denoted as A, B, and C) that can lead to a "dropped call," then P(dropped call | A, B, C) needs to be initialized in the conditional probability table.
[0092] There are a total of 8 values. Understandably, P(dropped call | A, B, C) represents the probability of a dropped call occurring given that A, B, and C all exist. This represents the probability of a dropped call occurring when both A and B exist and C does not exist.
[0093] Step 603: Update the probability values of each leaf node and each intermediate node in the Bayesian network of the first poor quality experience entity based on the real-time data related to the first poor quality experience entity.
[0094] Understandably, the relationships between the first poor quality experience entity, the first poor quality cause entity, the first poor quality experience entity and the first poor quality cause entity, as well as the relationships between the first poor quality cause entities, can be extracted from the real-time data related to the first poor quality experience entity. The extracted entities and relationships (also known as evidence) are then input into the Bayesian network of the first poor quality experience entity, and belief propagation calculations are performed. This allows the probability values of each leaf node and each intermediate node in the Bayesian network of the first poor quality experience entity to be updated.
[0095] Understandably, real-time data related to the first entity with poor quality experience includes real-time data on the base station's operating parameters, performance management, configuration management, and measurement reports. Real-time data refers to data from the day or the moment the probability update is performed.
[0096] Step 604: Determine the probability of the occurrence of the potential root cause of each first-quality-poor experience entity based on the probability of each leaf node and each intermediate node in the updated Bayesian network of the first-quality-poor experience entity.
[0097] Understandably, updating the probability of each intermediate node may affect the probability of the leaf nodes. Therefore, after obtaining the updated probability of each leaf node and each intermediate node in the Bayesian network, the probability of each leaf node is determined, which is also the probability of the potential root cause of each first-quality poor experience entity. Thus, the root cause of the first-quality poor experience entity can be located based on the probability of the potential root cause of each first-quality poor experience entity.
[0098] Step 504: Determine the root cause of the first poor quality experience entity based on the probability of occurrence of each potential root cause in the set of potential root causes of the first poor quality experience entity.
[0099] In some embodiments, the implementation method of step 504 may include: setting a first probability threshold; and determining the root cause of the first poor quality experience entity from among the potential root causes of the first poor quality experience entity based on the first probability threshold. Wherein, the sum of probabilities of the root causes of the first poor quality experience entity is not less than the first probability threshold.
[0100] Understandably, the implementation method of step 504 may also include: directly pre-selecting one or more potential root causes as the root cause of the first poor quality experience entity according to actual needs or specific circumstances, such as pre-selecting the potential root cause with the highest probability as the root cause of the first poor quality experience entity, or pre-selecting the potential root cause with the highest or second highest probability as the root cause of the first poor quality experience entity.
[0101] Embodiments of the present invention provide a call quality poor cause localization system, such as... Figure 7As shown, the call quality poor root cause localization system 700 includes a first acquisition module 701, a second acquisition module 702, a third acquisition module 703, and a determination module 704. The first acquisition module 701 is configured to acquire historical data from the call quality poor analysis manual and call quality poor cases. The second acquisition module 702 is configured to acquire call quality poor cause entities related to the first call quality poor experience entity as first call quality poor cause entities based on a call quality poor knowledge graph, forming a set of first call quality poor cause entities. The set of first call quality poor cause entities includes multiple first call quality poor cause entities. The call quality poor entities in the call quality poor knowledge graph include call quality poor experience entities and call quality poor cause entities. The first call quality poor experience entity is any one of the call quality poor experience entities. The determination module 704 is configured to determine the root cause of the first call quality poor experience entity in the set of first call quality poor cause entities based on the probability of each first call quality poor cause entity in the set of first call quality poor cause entities.
[0102] In some embodiments, the second acquisition module 702 is configured to acquire poor-quality entities and the relationships between them based on historical data from the poor-quality analysis manual and poor-quality cases. The relationships between poor-quality entities include inclusion relationships and causal relationships. Furthermore, a poor-quality knowledge graph is constructed based on the poor-quality entities and the relationships between them.
[0103] In some embodiments, the third acquisition module 703 is configured to acquire one or more quality-poor cause entities that have a causal relationship with the first quality-poor experience entity as second quality-poor cause entities based on the quality-poor knowledge graph. It also acquires a third quality-poor cause entity based on the quality-poor knowledge graph. The relationship between the third quality-poor cause entities is a causal relationship, or the relationship between the third quality-poor cause entity and the second quality-poor cause entity is a causal relationship. The first quality-poor cause entity is either the second quality-poor cause entity or the third quality-poor cause entity. The set of first quality-poor cause entities includes both the second and third quality-poor cause entities.
[0104] In some embodiments, such as Figure 8As shown, the determination module 704 includes a causal relationship chain construction unit 7041, a potential root cause set acquisition unit 7042, a probability acquisition unit 7043, and a root cause determination unit 7044. The causal relationship chain construction unit 7041 is configured to construct a causal relationship chain for the first poor quality experience entity based on the relationships between third poor quality cause entities, the relationships between the third poor quality cause entity and the second poor quality cause entity, and the relationships between the second poor quality cause entity and the first poor quality experience entity. The poor quality entity at the very end of each causal relationship chain for the first poor quality experience entity is the first poor quality experience entity. The potential root cause set acquisition unit 7042 is configured to acquire a set of potential root causes for the first poor quality experience entity based on the causal relationship chain of the first poor quality experience entity. The potential root cause of the first poor quality experience entity is the third poor quality cause entity at the very beginning of each causal relationship chain of the first poor quality experience entity. The probability acquisition unit 7043 is configured to acquire the probability of each potential root cause appearing in the set of potential root causes for the first poor quality experience entity. The root cause determination unit 7044 is configured to determine the root cause of the first poor quality experience entity based on the probability of occurrence of each potential root cause in the set of potential root causes of the first poor quality experience entity.
[0105] In some embodiments, such as Figure 9 As shown, the determining module 704 further includes a Bayesian network construction unit 7045. The Bayesian network construction unit 7045 is configured to construct a Bayesian network for the first poor quality experience entity based on the causal relationship chain of all first poor quality experience entities. The root node of the Bayesian network corresponds to the first poor quality experience entity, the leaf nodes correspond to the potential root causes of the first poor quality experience entity, and the intermediate nodes correspond to the first poor quality cause entities between the first poor quality experience entity and its potential root causes. The probability acquisition unit 7043 is configured to acquire the initial value of the probability of each node in the Bayesian network of the first poor quality experience entity based on the historical data of the poor quality analysis manual and poor quality cases. Furthermore, it updates the probability of each node in the Bayesian network of the first poor quality experience entity based on real-time data related to the first poor quality experience entity. The data related to the first poor quality experience entity includes real-time data of the acquisition base station operating parameters, real-time performance management data, real-time configuration management data, and real-time measurement report data related to the first poor quality experience entity. The root cause determination unit is configured to determine the probability of the occurrence of the potential root cause of each first-quality-poor-experience entity based on the probability of each node in the updated Bayesian network of the first-quality-poor-experience entity.
[0106] In some embodiments, the probability acquisition unit 7043 is configured to acquire, based on the quality defect analysis manual and historical data of quality defect cases, the prior probabilities of the leaf nodes in the Bayesian network of the first quality defect experience entity and the conditional probability table corresponding to the intermediate nodes in the Bayesian network of the first quality defect experience entity. The prior probability values of the leaf nodes in the Bayesian network of the first quality defect experience entity are used as the initial values of the probability of each leaf node in the Bayesian network of the first quality defect experience entity. Furthermore, the probability values in the conditional probability table corresponding to the intermediate nodes in the Bayesian network of the first quality defect experience entity are used as the initial values of the probability of each intermediate node in the Bayesian network of the first quality defect experience entity.
[0107] In some embodiments, the root cause determination unit 7044 is configured to: set a first probability threshold; and determine the root cause of the first poor quality experience entity from among the potential root causes of the first poor quality experience entity based on the first probability threshold. The sum of probabilities of the root causes of the first poor quality experience entity is not less than the first probability threshold.
[0108] Embodiments of the present invention provide a computer device, such as... Figure 10 As shown, the computer device 1000 includes a memory 1001 and a processor 1002. The memory 1001 stores a computer program. When the processor 1002 runs the computer program stored in the memory 1001, the processor 1002 executes the above-mentioned method for locating the root cause of poor call quality.
[0109] The specific solution and beneficial effects of the computer device provided by the embodiments of the present invention can be found in the relevant description of the call quality poor root cause localization method provided by the embodiments of the present invention, which will not be repeated here.
[0110] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the processor performs the above-described method for locating the root cause of poor call quality.
[0111] The specific scheme and beneficial effects of the computer-readable storage medium provided by the embodiments of the present invention can be found in the relevant description of the call quality poor root cause localization method provided by the embodiments of the present invention, which will not be repeated here.
[0112] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.
Claims
1. A method for call quality problem root cause positioning, characterized in that, include: Obtain historical data from the quality defect analysis manual and quality defect case studies; A quality defect knowledge graph is obtained based on the quality defect analysis manual and historical data of the quality defect cases. The poor quality entities in the poor quality knowledge graph include poor quality experience entities and poor quality cause entities; Based on the quality defect knowledge graph, one or more quality defect cause entities that have a causal relationship with the first quality defect experience entity are obtained as the second quality defect cause entities; A third quality-poor cause entity is obtained based on the quality-poor knowledge graph; the relationship between the third quality-poor cause entities is a causal relationship, or the relationship between the third quality-poor cause entity and the second quality-poor cause entity is a causal relationship; the first quality-poor cause entity is either the second quality-poor cause entity or the third quality-poor cause entity; the set of the first quality-poor cause entities includes both the second quality-poor cause entity and the third quality-poor cause entity; the set of the first quality-poor cause entities includes multiple first quality-poor cause entities; the first quality-poor experience entity is any one of the quality-poor experience entities. Based on the relationship between the third poor quality cause entities, the relationship between the third poor quality cause entity and the second poor quality cause entity, and the relationship between the second poor quality cause entity and the first poor quality experience entity, a causal relationship chain for the first poor quality experience entity is constructed. The lowest quality entity in the causal chain of each first poor quality experience entity is the first poor quality experience entity; The set of potential root causes of the first poor quality experience entity is obtained based on the causal relationship chain of the first poor quality experience entity; the potential root cause of the first poor quality experience entity is the third poor quality cause entity at the very beginning of each causal relationship chain of the first poor quality experience entity. Construct a Bayesian network for the first poor quality experience entity based on the causal relationship chain of all the first poor quality experience entities; the root node of the Bayesian network corresponds to the first poor quality experience entity, the leaf node of the Bayesian network corresponds to the potential root cause of the first poor quality experience entity, and the intermediate node of the Bayesian network corresponds to the first poor quality cause entity between the first poor quality experience entity and the potential root cause of the first poor quality experience entity. Based on the poor quality analysis manual and the historical data of the poor quality cases, obtain the initial values of the probability of each leaf node and each intermediate node in the Bayesian network of the first poor quality experience entity; The probability values of each leaf node and each intermediate node in the Bayesian network of the first poor quality experience entity are updated based on real-time data related to the first poor quality experience entity; the real-time data related to the first poor quality experience entity includes real-time data of the acquisition base station operating parameters, real-time data of performance management, real-time data of configuration management, and real-time data of measurement reports related to the first poor quality experience entity. The probability of the potential root cause of each of the first poor quality experience entities is determined based on the probability of each leaf node and each intermediate node in the updated Bayesian network of the first poor quality experience entity. as well as The root cause of the first poor quality experience entity is determined based on the probability of occurrence of each potential root cause in the set of potential root causes of the first poor quality experience entity.
2. The call quality problem root cause positioning method according to claim 1, characterized in that, The step of obtaining a quality defect knowledge graph based on the quality defect analysis manual and historical data of the quality defect cases includes: Based on the aforementioned quality defect analysis manual and historical data from the quality defect cases, we obtain the quality defect entities and the relationships between them; the relationships between the quality defect entities include inclusion relationships and causal relationships; and... The poor quality knowledge graph is constructed based on the poor quality entities and the relationships between them.
3. The method for locating the root cause of poor call quality according to claim 2, characterized in that, The step of obtaining the initial values of the probabilities of each leaf node and each intermediate node in the Bayesian network of the first quality-poor experience entity based on the quality-poor analysis manual and historical data of the quality-poor cases includes: Based on the quality defect analysis manual and the historical data of the quality defect cases, obtain the prior probabilities of the leaf nodes in the Bayesian network of the first quality defect experience entity and the conditional probability table corresponding to the intermediate nodes in the Bayesian network of the first quality defect experience entity. The prior probability values of the leaf nodes in the Bayesian network of the first poor quality experience entity are used as the initial probability values of each leaf node in the Bayesian network of the first poor quality experience entity; and The probability values in the conditional probability table corresponding to the intermediate nodes in the Bayesian network of the first poor quality experience entity are used as the initial values of the probability of each intermediate node in the Bayesian network of the first poor quality experience entity. 4.The call quality problem root cause positioning method according to claim 2, characterized in that, The step of determining the root cause of the first poor quality experience entity based on the probability of occurrence of each potential root cause in the set of potential root causes includes: Set a first probability threshold; and The root cause of the first poor quality experience entity is determined from the potential root causes of the first poor quality experience entity based on the first probability threshold; the sum of the probabilities of the root causes of the first poor quality experience entity is not less than the first probability threshold.
5. A call quality root cause positioning system, comprising: include: The first acquisition module is set to acquire historical data of the poor quality analysis manual and poor quality cases; The second acquisition module is configured to acquire a quality defect knowledge graph based on the quality defect analysis manual and the historical data of the quality defect cases; the quality defect entities in the quality defect knowledge graph include quality defect experience entities and quality defect reason entities. The third acquisition module is configured to acquire one or more quality-poor cause entities that have a causal relationship with the first quality-poor experience entity as second quality-poor cause entities based on the quality-poor knowledge graph; acquire a third quality-poor cause entity based on the quality-poor knowledge graph; the relationship between the third quality-poor cause entities is a causal relationship, or the relationship between the third quality-poor cause entity and the second quality-poor cause entity is a causal relationship; the first quality-poor cause entity is either the second quality-poor cause entity or the third quality-poor cause entity; the set of the first quality-poor cause entities includes the second quality-poor cause entity and the third quality-poor cause entity; the set of the first quality-poor cause entities includes multiple first quality-poor cause entities; the first quality-poor experience entity is any one of the quality-poor experience entities; The determination module is configured to: construct a causal relationship chain for the first poor quality experience entity based on the relationship between the third poor quality cause entity, the relationship between the third poor quality cause entity and the second poor quality cause entity, and the relationship between the second poor quality cause entity and the first poor quality experience entity; The lowest quality entity in the causal chain of each first poor quality experience entity is the first poor quality experience entity; the set of potential root causes of the first poor quality experience entities is obtained based on their causal chains; the potential root cause of each first poor quality experience entity is the highest quality cause entity in each causal chain of the first poor quality experience entity; a Bayesian network of the first poor quality experience entities is constructed based on all their causal chains; the root node of the Bayesian network corresponds to the first poor quality experience entity, the leaf nodes of the Bayesian network correspond to the potential root causes of the first poor quality experience entity, and the intermediate nodes of the Bayesian network correspond to the potential root causes of the first poor quality experience entity. The process involves defining the first poor quality experience entity and its potential root causes as the first poor quality experience entity; obtaining initial probabilities for each leaf node and each intermediate node in the Bayesian network of the first poor quality experience entity based on the poor quality analysis manual and historical data from the poor quality cases; updating the probabilities for each leaf node and each intermediate node in the Bayesian network of the first poor quality experience entity based on real-time data related to the first poor quality experience entity; the real-time data related to the first poor quality experience entity includes real-time data on base station operating parameters, performance management, configuration management, and measurement reports related to the first poor quality experience entity; and determining the probability of each potential root cause of the first poor quality experience entity occurring based on the updated probabilities of each leaf node and each intermediate node in the Bayesian network of the first poor quality experience entity. Furthermore, the root cause of the first poor quality experience entity is determined based on the probability of occurrence of each potential root cause in the set of potential root causes of the first poor quality experience entity.
6. A computer device, comprising: It includes a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the call quality poor root cause localization method according to any one of claims 1 to 4.
7. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by a processor, the processor performs the call quality poor root cause localization method according to any one of claims 1 to 4.
Citation Information
Patent Citations
Logic enhancement method and device of knowledge graph
CN108461151A
Heat supply system intelligent diagnosis method and system based on knowledge graph and Bayesian
CN115511119A