WO2023155481A1 - 无线通信网络知识图谱的智能分析与应用系统及方法 - Google Patents

无线通信网络知识图谱的智能分析与应用系统及方法 Download PDF

Info

Publication number
WO2023155481A1
WO2023155481A1 PCT/CN2022/128902 CN2022128902W WO2023155481A1 WO 2023155481 A1 WO2023155481 A1 WO 2023155481A1 CN 2022128902 W CN2022128902 W CN 2022128902W WO 2023155481 A1 WO2023155481 A1 WO 2023155481A1
Authority
WO
WIPO (PCT)
Prior art keywords
wireless communication
communication network
entity
intelligent
algorithm
Prior art date
Application number
PCT/CN2022/128902
Other languages
English (en)
French (fr)
Inventor
何世文
黄永明
任鹏
詹行
王良鹏
安振宇
易云山
尤肖虎
Original Assignee
网络通信与安全紫金山实验室
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 网络通信与安全紫金山实验室 filed Critical 网络通信与安全紫金山实验室
Publication of WO2023155481A1 publication Critical patent/WO2023155481A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/02—Arrangements for optimising operational condition
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367—Ontology
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/08—Testing, supervising or monitoring using real traffic

Definitions

  • the present disclosure relates to the technical field of wireless communication network intelligent transmission and management and control, in particular to a system and method for intelligent analysis and application of wireless communication network knowledge graphs.
  • eMBB Enhanced Mobile Broadband
  • mMTC massive Machine Type of Communication
  • uRLLC Ultra-reliable and Low Latency Communications
  • a wireless communication network consists of many elements such as user terminals, access network, and core network.
  • the network structure is usually complex, including tens of thousands of performance indicators and data fields, involving different network elements, protocol stacks, etc. These indicators are usually used to characterize the performance of a wireless communication network, and are important parameters to measure the operation status of a wireless communication network. On the other hand, these indicators are closely related to user experience quality (Quality of Experience, QoE).
  • QoE Quality of Experience
  • the present disclosure provides a system and method for intelligent analysis and application of wireless communication network knowledge maps.
  • the present disclosure provides an intelligent analysis and application system for a wireless communication network knowledge map, at least including:
  • the knowledge map unit is used to construct a wireless communication network knowledge map based on endogenous factors of the wireless communication network and a map construction method, and perform in-depth analysis and map cycle optimization in combination with classified wireless communication network data;
  • the intelligent traceability unit is used to determine the abnormal diagnosis and positioning results of the wireless communication network based on the optimized wireless communication network knowledge map, as well as anomaly detection and intelligent reasoning algorithms;
  • the optimization strategy unit is used to determine a number of optimization strategies based on the strategy generation algorithm and the diagnosis and positioning results, and aim at the best execution efficiency and effect of the optimization strategy, combined with an intelligent decision-making algorithm, in several optimization strategies An optimal optimization strategy is determined in the method, and the wireless communication network is optimized.
  • the present disclosure also provides an intelligent analysis and application method of a wireless communication network knowledge map, including:
  • the knowledge map of the wireless communication network is constructed, and combined with the classified wireless communication network data for in-depth analysis and map cycle optimization;
  • a policy is used to optimize the wireless communication network.
  • the present disclosure also provides an electronic device, including a memory, a transceiver, and a processor;
  • a memory used to store computer programs; a transceiver, used to send and receive data under the control of the processor; a processor, used to read the computer programs in the memory and implement wireless communication as described in the second aspect above.
  • the present disclosure also provides a non-transitory computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the realization of the wireless communication network knowledge graph described in the second aspect above is realized. Steps of intelligent analysis and application method.
  • the present disclosure also provides a computer program product, including a computer program, and when the computer program is executed by a processor, the steps of the intelligent analysis and application method for the wireless communication network knowledge map as described in the second aspect are implemented .
  • the intelligent analysis and application system and method of the wireless communication network knowledge map provided by this disclosure, through the endogenous factors of the wireless communication network, combined with the communication expert knowledge and 3GPP communication protocol, initially constructs the wireless communication network knowledge map and performs the depth of the correlation relationship of the endogenous factors Analysis, using the analysis results to complete the optimization and update of the map, use the intelligent reasoning module to trace the source and diagnose the location after the abnormal network communication is found, use the strategy generation module to provide several feasible optimization strategies, and choose one by the strategy decision module
  • the specific strategy is delivered to the wireless communication network for execution, and the wireless communication network is cyclically optimized to effectively improve the performance of the wireless communication network and the quality of user experience.
  • FIG. 1 is a schematic diagram of an intelligent analysis and application system of a wireless communication network knowledge map provided by the present disclosure
  • FIG. 2 is a schematic flowchart of an intelligent analysis and application method of a wireless communication network knowledge map provided by the present disclosure
  • FIG. 3 is a schematic diagram of the overall flow of the intelligent analysis and application method of the wireless communication network knowledge map provided by the present disclosure
  • FIG. 4 is a schematic diagram of the business process of the intelligent analysis and application method of the wireless communication network knowledge map provided by the present disclosure
  • FIG. 5 is a partial schematic diagram of an air interface delay knowledge map of a wireless communication network knowledge map provided by the present disclosure
  • Fig. 6 is a schematic structural diagram of an electronic device provided by the present disclosure.
  • Fig. 1 is a schematic diagram of an intelligent analysis and application system of a wireless communication network knowledge map provided by the present disclosure; as shown in Fig. 1, the intelligent analysis and application system of a wireless communication network knowledge map at least includes: a knowledge map unit, an intelligent traceability unit and Tuning policy unit;
  • the knowledge map unit is used to construct a wireless communication network knowledge map based on endogenous factors of the wireless communication network and a map construction method, and perform in-depth analysis and map cycle optimization in combination with classified wireless communication network data;
  • the intelligent traceability unit is used to determine the abnormal diagnosis and positioning results of the wireless communication network based on the optimized wireless communication network knowledge map, as well as anomaly detection and intelligent reasoning algorithms;
  • the optimization strategy unit is used to determine a number of optimization strategies based on the strategy generation algorithm and the diagnosis and positioning results, and aim at the best execution efficiency and effect of the optimization strategy, combined with an intelligent decision-making algorithm, in several recommended optimization strategies
  • An optimal optimization strategy is determined in the strategy to optimize the wireless communication network.
  • the present disclosure proposes an intelligent analysis and application system and method for knowledge maps of wireless communication networks, which provides strong support for realizing the intelligence of wireless communication networks.
  • the system at least includes: a knowledge map unit, an intelligent traceability unit, and an optimization strategy unit; wherein,
  • the knowledge map unit is used to build a wireless communication network knowledge map based on the endogenous factors of the wireless communication network and the map construction method, and combine the classified wireless communication network data for in-depth analysis and map cycle optimization; among them, the wireless
  • the communication network knowledge map is constructed based on the elements in the entity library after the initialization of the endogenous factors of the wireless communication network, and the triplet composed of elements in the inter-entity relationship library.
  • the endogenous factors of the wireless communication network include The indicators and data fields stipulated in the protocol; the classified wireless communication network data can be directly classified according to the protocol, or classified according to the set rules, mainly to facilitate subsequent data processing and improve processing efficiency.
  • the knowledge graph unit includes at least a deep analysis module and a graph optimization module;
  • the in-depth analysis module is used to determine new entities, new entity types, new entity attributes, and new relationships between entities in the wireless communication network knowledge map based on the analysis algorithm of the graph model and the data model. One or more, and improve the degree of association of the corresponding relationship between entities;
  • the graph optimization module is configured to update the wireless communication network knowledge graph based on the updated entity database and/or the updated inter-entity relationship database;
  • the knowledge map unit includes at least: a deep analysis module and a map tuning module;
  • the in-depth analysis module will initialize the entity library and the relationship between entities that constitute the wireless communication network knowledge graph as input.
  • the in-depth analysis module has two main functions. One is to quantify and measure the degree of association between two entities from the perspective of the graph model, and to establish the edge connection strength; the other is to deeply analyze new entities and new entities in the wireless communication network knowledge map from the perspective of data. One or more of types, new entity attributes, and new inter-entity relationships and degrees of association.
  • the initialization of the entity database and the inter-entity relationship database constituting the wireless communication network knowledge map is determined based on the endogenous factors of the wireless communication network, and the above-mentioned endogenous factors of the wireless communication network include indicators and data specified according to the wireless communication network protocol. field.
  • Each element in the entity library includes the entity type corresponding to the element, and all attributes of the element; the inter-entity relationship library includes the specific relationship between each two elements, and the specific relationship between each two elements relationship, etc.
  • the analysis method based on the graph model and the data model determines the new relationship between entities in the wireless communication network knowledge map, analyzes the degree of association between the entities, and updates the An inter-entity relationship library, and/or determine one or more of new entities, new entity types, and new entity attributes in the wireless communication network knowledge map, and update the entity library, including:
  • this module carries out data-driven in-depth analysis of the wireless communication network knowledge map, uses the real-time data of the wireless communication network collected by the wireless data unit, and deeply analyzes the nodes in the endogenous factor knowledge map from the perspective of data ( The degree of association and dependency between entities) determines one or more of new entities, entity types, new entity attributes, and new inter-entity relationships.
  • the second main function of the in-depth analysis module is that the in-depth analysis module is based on the wireless communication network knowledge map constructed after the simple analysis, relying on the theoretical analysis method based on the graph model, to deeply explore the relationship between the endogenous factors between the nodes, such as indicators and Data fields, indicators and indicators, data fields and data fields, etc., that is, to determine the degree of association or dependence of the relationship between entities. Quantify and measure the degree of association and dependence between two nodes (entities), and establish the strength of the edge connection.
  • the graph optimization module After in-depth analysis of the wireless network knowledge graph, it is possible to obtain new entities or entity relationships in the wireless network knowledge graph, which needs to be optimized for the wireless network knowledge graph, which is mainly completed by the graph optimization module.
  • This module mainly It is used to determine the tuning content and tuning direction of the wireless communication network knowledge map.
  • updating the wireless communication network knowledge graph based on the updated entity database and/or the updated inter-entity relationship database includes:
  • parameters that need to be tuned include: One or more of entities, parameters of entity attributes, and parameters related to relationships between entities in the wireless communication network knowledge map;
  • the wireless communication network knowledge graph is updated based on the parameters that need to be tuned.
  • the map optimization module takes the relevant analysis results of the deep analysis module as input, that is, the updated entity library and/or the updated entity relationship library as input, and outputs the wireless communication network knowledge map.
  • Tuning content and tuning direction can be understood as tuning and updating of graph structures (such as errata for wrong graph structures, completion of missing graph structures, etc.), tuning and updating of relationships between entities (such as deletion of redundant associations, correction of wrong associations, or modification of the association degree of the relationship between two entities from association degree 2 to association degree 3, etc.)
  • the specific association degree value can indicate the strength of the association degree of the relationship between two entities Weak, the larger the value, the stronger the degree of association, or the weaker the degree of association, or the update of the parameters of the attribute of the entity (for example, the adjustable parameter of the attribute corresponding to the entity of a certain entity type is changed from adjustable to It is not adjustable, information such as the location of the entity needs to be modified, etc.), etc.
  • the tuning update parameters are fed back to the graph
  • the knowledge graph unit may also include: a superficial parsing module.
  • the simple analysis module mainly determines the endogenous factors of the wireless communication network composed of the indicators and data fields specified in the wireless communication network protocol, such as the 3GPP communication protocol or combined with expert knowledge in the communication field, and determines several entities based on the endogenous factors of the wireless communication network.
  • the endogenous factors mainly include the indicators and data fields specified in the wireless communication network protocol, that is, to sort out the indicators and data fields involved in the wireless communication network, and determine the initial elements of the entity library, such as
  • the indicators include radio resource control (Radio Resource Control, RRC) connection establishment success rate, handover success rate, uplink and downlink packet loss rate, etc.; data fields include subcarrier spacing, uplink and downlink time slot ratio, highest downlink modulation mode per second, etc.
  • RRC Radio Resource Control
  • Each element in the above entity library can be regarded as an entity, and the entity contains entity attributes that can characterize the characteristic parameters of the entity object, such as entity name, type, word length, communication layer it belongs to, value range, adjustability, etc.
  • the simple analysis module will also analyze various indicators (such as RRC connection establishment success rate, handover success rate, uplink and downlink packet loss rate, etc.) etc.) to determine the entity types of the elements (ie, various entities) in the above entity library, which are recorded as ⁇ entity 1 , entity 2 ,...,entity m ⁇ , there are m entity types in total, and each entity belongs to this One of m entity types. Aiming at the defined entities, the relationship between entities is analyzed in a simple way, and the relationship between several entities is established according to the analysis results of the relationship between indicators and data fields in the simple analysis module.
  • various indicators such as RRC connection establishment success rate, handover success rate, uplink and downlink packet loss rate, etc.
  • the relationship between the above-mentioned determined entities constitutes an initialized inter-entity relationship library, which includes two associated entities and the relationship between the two entities, wherein the relationship between the two entities can be recorded as ⁇ relation 1 , relation 2 ,...,relation n ⁇ , there are n types of entity relations in total, and the relationship between any two entities belongs to one of these n types of entity relations.
  • the above entity library is updated; If there are new entities, any combination of new entity types and new entity attributes, and new inter-entity relationships, update the entity library and the inter-entity relationship library at the same time .
  • the knowledge graph unit may further include: a graph building module.
  • the graph building module is mainly used to build a wireless communication network knowledge graph with topology.
  • a general triplet (head, relation, tail) of the wireless communication network is constructed, wherein, head is the head entity in the triplet, tail is the tail entity in the triplet, each Both the head entity and the tail entity in each triplet belong to one of the m types of entities established in the above-mentioned entity database, and relation is a relationship between entities, which belongs to one of the n types of relationships established in the above-mentioned inter-entity relationship database.
  • a wireless communication network knowledge graph is constructed, and the head entity and tail entity in the triplet are respectively used as nodes of the wireless communication network knowledge graph. If there is a clear pointing relationship between nodes, they are connected by a directed line segment, otherwise they are connected by an undirected line segment. By connecting all the nodes with edge connections, a wireless communication network knowledge graph with a topological structure can be obtained, which is usually a multi-relationship graph.
  • the entity types of the elements in the entity library include one or more of network-level performance evaluation indicators, user-level performance evaluation indicators, general non-tunable data parameters, and adjustable data parameters.
  • the entity relationship in the inter-entity relationship library includes one or more of causal relationship, implicit relationship and explicit relationship.
  • the entity types of the elements in the entity library determined during initialization include network-level performance evaluation indicators, user-level performance evaluation indicators, general non-tunable data parameters and adjustable data parameters.
  • the entity also includes entity attributes such as the communication layer, type, adjustability, and parameter value of the entity object.
  • the inter-entity relationship database is initialized, it is determined that the above-mentioned inter-entity relationships can generally be classified into one or more of three types: causal relationship, implicit relationship and explicit relationship, and of course there may be other inter-entity relationships.
  • the causal relationship describes the direct impact between two entities, such as the physical layer transmission rate and physical layer throughput; the implicit relationship describes the indirect impact between two entities, and there is no clear and specific expression to express the relationship between the two entities. Relationships, such as the highest modulation mode and physical layer throughput; explicit relationships describe the specific analytical expressions that can be obtained after reasoning and analysis between two entities, such as the reference signal received power and synchronization signal received power of beam 1.
  • the above is only an illustration, and does not limit the specific classification of entity types and the specific classification of the relationship between entities in the present disclosure.
  • the classification of entity types in the present disclosure may include more, and the classification of relationships between entities may also include more, all of which are dynamically determined based on endogenous factors of the wireless communication network.
  • the intelligent traceability unit includes an abnormality detection module and an intelligent reasoning module;
  • the anomaly detection module is configured to perform anomaly detection on the wireless communication network based on the optimized knowledge map of the wireless communication network and an anomaly detection algorithm to determine a fault type;
  • An intelligent reasoning module configured to determine the entity in the wireless communication network knowledge graph to which the fault type belongs based on the fault type and an intelligent reasoning algorithm.
  • the intelligent analysis and application system of wireless communication network knowledge graph also includes:
  • An intelligent traceability unit which includes: an abnormality detection module and an intelligent reasoning module; specifically,
  • the anomaly detection module is mainly based on the anomaly detection algorithm and the optimized wireless communication network knowledge map to detect the anomaly of the wireless network.
  • a network anomaly When a network anomaly is found, it completes the preliminary detection of the fault type, that is, analyzes which type of entity it is Or which part of the graph structure or which part of the relationship between entities is abnormal, including abnormal nodes (entities), edges (relationships between entities), subgraph structures (need to add or delete entities), etc., for subsequent traceability Give directions.
  • the failure types include: failures of different elements in the entity database and failures of different elements in the inter-entity relationship database.
  • the failure of different elements in the entity library that is, the failure of different entities, mainly refers to which type of entity is abnormal.
  • the failure of different elements in the inter-entity relationship library mainly refers to which part of the graph structure or the relationship between entities, or the graph structure and the relationship between entities is abnormal.
  • the intelligent reasoning module is mainly used to determine the entity to which the fault type belongs and the attribute of the entity.
  • the determining the entity in the knowledge map of the wireless communication network to which the fault type belongs based on the fault type and an intelligent reasoning algorithm includes:
  • the location includes the specific communication layer where the entity is located or the location of the graph structure, and the influencing factors are the location of the entity The extent to which the entity is influenced by other entities with which it has an inter-entity relationship.
  • the intelligent reasoning module takes the fault type obtained by the anomaly detection module as input, relies on the constructed wireless communication network knowledge map and its endogenous factor correlation, and uses artificial intelligence methods such as approximate reasoning and causal inference to complete the network performance
  • the source of the problem is to dig out which indicators or data fields of which layer caused the abnormal fluctuation of the corresponding indicators, and then determine the specific entity type of the entity.
  • the tuning policy unit includes a policy generation module and a policy decision module
  • the strategy generating module is configured to determine several wireless communication network optimization strategies based on entities in the wireless communication network knowledge map to which the fault type belongs and a strategy generation algorithm;
  • the strategy decision-making module is configured to determine an optimal optimization strategy based on an intelligent decision-making algorithm, aiming at the best execution efficiency and effect of the optimization strategy, and optimize the wireless communication network.
  • the intelligent analysis and application system of wireless communication network knowledge graph also includes:
  • an optimization strategy unit which unit includes: a strategy generation module, a strategy decision module; specifically, the strategy generation module, based on the diagnosis and positioning results, and a strategy generation algorithm, determines a number of wireless communication network optimization strategies;
  • the strategy generation module mainly takes the diagnosis and positioning results of the intelligent reasoning module as input, and provides several possible communication networks by using the strategy generation algorithm on the premise of clarifying the location, adjustability, and influencing factors of the data fields that cause performance index fluctuations Tuning strategy, including which data field parameters to adjust and their adjustment direction. In view of the complexity of the wireless communication network, such an optimization strategy is often not unique.
  • the plurality of wireless communication network optimization strategies all include specific adjustable parameters and adjusted values of the adjustable parameters.
  • the wireless communication network optimization strategy may include specific adjustable data parameters and values that need to be adjusted for the adjustable parameters.
  • the strategy decision-making module takes several feasible communication network optimization strategies provided by the strategy generation module as input, and according to the intelligent decision-making algorithm, comprehensively considers that the execution efficiency and effect of the optimization strategy are the best, and then selects an optimal strategy and sends it to the Wireless communication network implementation to achieve network optimization.
  • the intelligent decision-making algorithm includes one or more of deep reinforcement learning DRL, Markov decision process, bionic algorithm and statistical learning algorithm.
  • the strategic decision-making module constructs an intelligent decision-making model (algorithm) based on artificial intelligence, deep reinforcement learning (Deep Reinforcement Learning, DRL), Markov decision process (Markov Decision Process, MDP), bionic algorithm and/or statistical learning, etc.
  • intelligent decision-making algorithms comprehensively considering the execution efficiency and effect of the optimization strategy, and based on the weights corresponding to the execution efficiency and execution effect, determine the corresponding weight of each optimization strategy Weighting the results, a wireless communication network optimization strategy is selected as the final decision strategy.
  • a wireless communication network optimization strategy is selected as the final decision strategy.
  • the above-mentioned intelligent analysis and application system of the wireless communication network knowledge map further includes: a wireless data unit; the wireless data unit is used to obtain the original data of the wireless communication network, and after preprocessing, Classified storage is performed according to different data types; the classified storage of different data types is realized based on a distributed system architecture.
  • the intelligent analysis and application system of the wireless communication network knowledge map also includes:
  • this unit mainly carries out wireless communication network raw data collection, and carries out preprocessing, and described preprocessing includes data format conversion, cleaning etc., and classifies and stores according to different types;
  • Data collection includes the collection of raw data of wireless network communication through methods such as soft and hard data collection.
  • Soft collection mainly refers to data collection through software such as computer programs
  • hard collection mainly refers to the collection of raw data from different wireless communication interfaces.
  • Data format conversion is mainly to convert the format of different data in the wireless communication network into a unified and recognizable format, or to convert the recognizable format of a certain unit into the recognizable or processed format of the next processing unit connected to the unit;
  • Data cleaning mainly refers to the deletion and other processing of obviously abnormal data.
  • the wireless data unit includes a data collection module, and the data collection module is constructed based on a distributed system.
  • the above-mentioned distributed system can implement distributed storage and processing of wireless communication network data, and especially provide an efficient and reliable processing and storage method for massive data storage and processing.
  • ES Hadoop distributed system infrastructure Hadoop implements a distributed file system (Hadoop Distributed File System), referred to as HDFS.
  • HDFS is highly fault-tolerant and designed to be deployed on inexpensive hardware. And it provides a high transfer rate to access application data, suitable for those applications with very large data sets.
  • HDFS relaxes the requirements of the Portable Operating System Interface (POSIX) so that data in the file system can be accessed in the form of streams.
  • POSIX Portable Operating System Interface
  • Hadoop is capable of distributed processing of large amounts of data. is processed in a reliable, efficient, and scalable manner.
  • Hadoop's reliability is mainly reflected in its assumption that computing elements and storage will fail, so it maintains multiple copies of working data to ensure that processing can be redistributed to failed nodes.
  • the efficiency of Hadoop is mainly reflected in its parallel way of working, through parallel processing to speed up the processing speed.
  • the scalability of Hadoop is mainly reflected in its ability to process PB (Petabyte, storage unit) level data.
  • PB Petabyte, storage unit
  • Hadoop relies on community servers, which are relatively low-cost and can be used by anyone.
  • the different data types include: one or more of terminal side wireless air interface data, base station side wireless air interface data, core network data, and network management data.
  • the collected data can usually be divided into four categories, but not limited to these four categories of data: the first category is wireless air interface terminal side data, mainly including L1, L2 and L3 layer data; the second category is wireless air interface base station side data, It mainly includes L1, L2 and L3 layer data; the third type is core network data, which is mainly divided into two data types: user plane and control plane; the fourth type is network management data, which mainly includes alarm, configuration and performance information.
  • the classification of wireless communication network data listed above is only an example, and the present disclosure is not limited thereto. In the prior art, all data transmitted between network elements and air interface data in a wireless communication network fall within the scope of this disclosure. within the scope of protection.
  • the alarm is mainly an alarm about the abnormal data of the wireless communication network nodes, for example, the data load of the wireless communication network nodes is large, or there are abnormal alarms about some nodes.
  • the configuration is mainly the configuration information for each node in the wireless communication network.
  • Performance information mainly refers to network-level performance information and user-level performance information in wireless communication networks.
  • the above-mentioned intelligent analysis and application system of the wireless communication network knowledge map further includes: a performance evaluation unit, which is used to combine the optimization strategy and the intelligent algorithm based on the optimized wireless communication network
  • the evaluation model determines the degree of improvement of the network-level performance and user-level performance of the wireless communication network by the optimization strategy, as well as the comprehensive capability evaluation result of the intelligent algorithm adopted by each unit, and feeds back to the corresponding unit.
  • the performance evaluation unit includes a network-level performance evaluation module and a user-level performance evaluation module;
  • the network-level performance evaluation module and the user-level performance evaluation module are used to obtain data of the wireless communication network after tuning;
  • the intelligent algorithm includes: depth analysis algorithm, abnormal detection algorithm, intelligent reasoning algorithm, strategy generation algorithm and intelligent decision-making algorithm. one or more.
  • the intelligent analysis and application system of the wireless communication network knowledge map also includes: a performance evaluation unit,
  • a performance evaluation unit which includes: a network-level performance evaluation module and a user-level performance evaluation module; specifically,
  • the performance evaluation unit has two main inputs.
  • the first input is the algorithm strategy and related output adopted by each module in the knowledge graph unit, intelligent traceability unit and optimization strategy unit, for example, including the optimization strategy selected by the policy decision-making module and issued to the wireless communication network for execution; the second The second input is the real-time data in the wireless communication network after the tuning strategy is executed.
  • This unit is mainly divided into two directions: network-level performance evaluation and user-level performance evaluation to evaluate the entire system and determine the corresponding degree of improvement. That is, it is determined whether the wireless network after tuning is improved compared with before tuning, whether the corresponding network-level performance or user-level performance is improved or not, and the specific degree of improvement or reduction is.
  • the performance evaluation unit feeds back the evaluation results to the corresponding modules in the knowledge graph unit, intelligent traceability unit, and optimization strategy unit.
  • the output results are adjusted and optimized to achieve dynamic network optimization. For example, if the feedback results indicate that the optimization strategy selected by the policy decision-making module is not effective, the module can reselect the optimization strategy by changing the intelligent decision-making algorithm.
  • the determination of the intelligent algorithm that needs to be adjusted based on the degree of improvement and the comprehensive capability evaluation result of the intelligent algorithm includes:
  • the improvement degree of network-level performance and/or user-level performance does not change or decrease, determine the comprehensive capability evaluation result of the intelligent algorithm adopted by each unit;
  • the performance evaluation unit evaluates the improvement degree of the wireless network performance from two perspectives of network-level performance and user-level performance, and if the evaluation result is no improvement, or on the contrary In the case of a decrease, determine the comprehensive capability evaluation results of the intelligent algorithms adopted by each unit.
  • the method for determining the comprehensive ability evaluation results of the intelligent algorithms adopted by each unit includes: for the intelligent algorithms adopted by all units or all modules, conduct a comprehensive evaluation through algorithm generalization ability, scalability, interpretability, etc., and determine the comprehensive ability of the algorithm.
  • a new optimal optimization scheme is determined and executed again. If the comprehensive capability evaluation results of the intelligent algorithms of all units or modules are basically the same, select other suboptimal solutions among the above-mentioned several optimization solutions for execution. If the comprehensive capability evaluation results of the intelligent algorithms of all units or modules are lower than before the optimization, continue to repeat the above steps of determining the optimal optimization solution.
  • the performance evaluation unit and the above-mentioned knowledge graph unit, intelligent traceability unit and optimization strategy unit can form an intelligent decision-making closed loop, and the data and related information of the wireless communication network can be directly input into the knowledge graph unit and the performance evaluation unit; or
  • a wireless data unit is added to preprocess the wireless communication network data and then input to the knowledge graph unit and performance evaluation unit, which also constitutes an intelligent decision-making closed loop.
  • the intelligent analysis and application system of the wireless communication network knowledge map provided by the present disclosure may include multiple implementation methods, including a knowledge map unit, an intelligent traceability unit, and an optimization strategy unit.
  • the wireless communication network executes the determined optimal optimization strategy Finally, input the optimized wireless communication network into the knowledge graph unit to update and adjust the knowledge graph of the wireless communication network; it may also include a knowledge graph unit, an intelligent traceability unit, an optimization strategy unit and a performance evaluation unit, and the wireless communication network is determined
  • the performance evaluation unit determines the improvement degree of the optimization strategy to the network-level performance and user-level performance of the wireless communication network, as well as the comprehensive capability evaluation result of the intelligent algorithm adopted by each unit, and feeds back For the corresponding unit; it may also include a knowledge graph unit, an intelligent traceability unit, an optimization strategy unit and a wireless data unit, and may also include a knowledge graph unit, an intelligent traceability unit, an optimization strategy unit, a performance evaluation unit and a wireless data unit.
  • the intelligent analysis and application system of the wireless communication network knowledge map obtaineds and stores relevant data of the wireless communication network through the wireless data unit, and combines the communication expert knowledge and the 3GPP communication protocol to initially construct the wireless communication network knowledge map and perform endogenous factors.
  • In-depth analysis of association relationships using the analysis results to complete map optimization and updating, using the intelligent reasoning module to trace the source and diagnosis and positioning after discovering network communication abnormalities, using the strategy generation module to provide several feasible optimization solutions, and the strategy decision module Select a specific policy solution and send it to the communication network for execution.
  • the performance evaluation module evaluates the execution results of the optimization strategy and the comprehensive ability of the intelligent algorithm to complete the closed loop of intelligent decision-making and effectively improve the performance of the wireless communication network and the quality of user experience.
  • Fig. 2 is a schematic flowchart of an intelligent analysis and application method of a wireless communication network knowledge map provided by the present disclosure; as shown in Fig. 2 , the method includes:
  • Step 201 based on the endogenous factors of the wireless communication network and the map construction method, construct the wireless communication network knowledge map, and combine the classified wireless communication network data to perform in-depth analysis and map cycle optimization;
  • Step 202 based on the optimized wireless communication network knowledge map, and anomaly detection and intelligent reasoning algorithms, determine the diagnosis and location results of wireless communication network anomalies;
  • Step 203 Based on the strategy generation algorithm and the diagnosis and positioning results, determine several optimization strategies, and aim at the best execution efficiency and effect of the optimization strategies, combined with the intelligent decision-making algorithm, determine the best one among the several recommended optimization strategies An optimized optimization strategy is used to optimize the wireless communication network.
  • the intelligent analysis and application method of the wireless communication network knowledge map includes:
  • Map construction and optimization are mainly stipulated by wireless communication network protocols, such as 3GPP communication protocols or combined with expert knowledge in the communication field, to determine the endogenous factors of the wireless communication network composed of indicators and data fields specified in it, based on the endogenous factors of the wireless communication network Determine several entity types and the relationship between several entities.
  • the endogenous factors mainly include the indicators and data fields specified in the wireless communication network protocol, that is, to sort out the indicators and data fields involved in the wireless communication network, and determine the initial entity library.
  • Elements such as RRC connection establishment success rate, handover success rate, uplink and downlink packet loss rate, subcarrier spacing, uplink and downlink time slot ratio, highest downlink modulation mode per second, etc.
  • the wireless communication network data is further analyzed.
  • it is usually necessary to preprocess the original wireless communication network data specifically including:
  • the soft collection mainly refers to the data collection realized by software such as computer programs
  • the hard collection mainly refers to the raw data collection of different wireless communication interfaces.
  • Data format conversion is mainly to convert the format of different data in the wireless communication network into a unified and recognizable format, or to convert the recognizable format of a certain unit into the recognizable or processed format of the next processing unit connected to the unit;
  • Data cleaning mainly refers to the deletion and other processing of obviously abnormal data.
  • Intelligent traceability based on the above-mentioned optimized wireless communication network knowledge graph, combined with anomaly detection algorithms, completes the preliminary detection of fault types, that is, analyzes which type of entity or which part of the graph structure or which part of the relationship between entities occurred abnormal.
  • intelligent reasoning algorithms it can determine the diagnosis and location results of wireless communication network abnormalities, that is, determine one or several entities or the relationship between entities that specifically have faults.
  • the strategy generation algorithm For the optimization strategy, according to the above diagnosis and location results, use the strategy generation algorithm to generate several wireless communication network optimization strategies. In view of the complexity of the wireless communication network, such an optimization strategy is often not unique. And aiming at the best execution efficiency and effect of the optimization strategy, an optimal optimization strategy is determined according to an intelligent decision-making algorithm, and the wireless communication network is optimized.
  • the intelligent analysis and application method of the knowledge graph of the wireless communication network may also include determining the impact of the optimization strategy on the wireless communication network at the network level based on the optimized wireless communication network, combining the optimization strategy and the evaluation model of the intelligent algorithm. The degree of improvement in performance and user-level performance, as well as the comprehensive capability evaluation results of the various intelligent algorithms adopted.
  • performance evaluation according to the data of the wireless communication network after tuning, evaluate the degree of improvement of the wireless communication network in the two dimensions of network-level performance evaluation and user-level performance evaluation, and the intelligent algorithm used in all the above steps Based on the comprehensive capability evaluation results, determine which intelligent algorithm needs to be adjusted in the specific step, and update the intelligent algorithm. Based on the updated intelligent algorithm, repeat the above steps to generate a new optimization strategy execution, and perform the wireless communication network again.
  • Performance evaluation to determine the degree of improvement of the entire wireless communication network in the two dimensions of network-level performance evaluation and user-level performance evaluation.
  • the intelligent analysis and application method of the wireless communication network knowledge map provided by this disclosure, through the endogenous factors of the wireless communication network, combined with the knowledge of communication experts and the 3GPP communication protocol, initially constructs the knowledge map of the wireless communication network and conducts in-depth analysis of the relationship between the endogenous factors. Use the analysis results to complete the map optimization and update, use the intelligent reasoning module to trace the source and diagnose and locate after discovering network communication abnormalities, use the strategy generation module to provide several feasible optimization solutions, and use the strategy decision module to select a specific one
  • the strategy plan is sent to the communication network for execution, and the wireless communication network is cyclically optimized to effectively improve the performance of the wireless communication network and the quality of user experience.
  • the construction of a wireless communication network knowledge map based on the endogenous factors of the wireless communication network and the map construction method, and combining the classified wireless communication network data for in-depth analysis and map cycle optimization include:
  • the entity library and the inter-entity relationship library are determined based on endogenous factors of the wireless communication network; the endogenous factors of the wireless communication network include indicators and data fields specified according to wireless communication network protocols.
  • the determination of the diagnosis and positioning result of the wireless communication network abnormality based on the optimized wireless communication network knowledge map, and anomaly detection and intelligent reasoning algorithms includes:
  • anomaly detection is performed on the wireless communication network to determine a fault type
  • An optimal tuning strategy is used to tune the wireless communication network, including:
  • an optimal optimization strategy is determined to optimize the wireless communication network.
  • the method also includes:
  • the optimization strategy determines the degree of improvement of the wireless communication network by the optimization strategy at the network level performance and user level performance, and the The comprehensive capability evaluation results of each intelligent algorithm adopted, including:
  • the intelligent algorithms include: in-depth analysis algorithms, abnormal detection algorithms, intelligent reasoning algorithms, strategy generation algorithms and intelligent decision-making algorithms one or more of .
  • the wireless communication network knowledge map is constructed based on the endogenous factors of the wireless communication network and the map construction method, and before in-depth analysis and map cycle optimization are performed in combination with the classified wireless communication network data, it also includes:
  • the raw data of the wireless communication network is obtained and preprocessed, and then classified and stored according to different data types; the classified storage of different data types is realized based on a distributed system architecture.
  • the different data types include: one or more of terminal side wireless air interface data, base station side wireless air interface data, core network data, and network management data.
  • the entity types of the elements in the entity library include one or more of network-level performance evaluation indicators, user-level performance evaluation indicators, general non-adjustable data parameters and adjustable data parameters; the relationship between entities Entity relationships in the repository include one or more of causal, implicit, and explicit relationships.
  • the analysis algorithm based on the graph model and the data model determines new relationships between entities in the knowledge map of the wireless communication network, analyzes the degree of association of the relationships between entities, and updates the relationship library between entities, and/or Or determine one or more of new entities, new entity types, and new entity attributes in the wireless communication network knowledge map, and update the entity library, including:
  • the quantitative measure of the degree of association between entities in the knowledge map of the wireless communication network is determined, and the database of the relationship between entities is updated.
  • updating the wireless communication network knowledge graph based on the updated entity database and/or the updated inter-entity relationship database includes:
  • wireless communication network One or more of entities in the knowledge graph, parameters of entity attributes, and parameters of relationships between entities;
  • the wireless communication network knowledge graph is updated based on the parameters that need to be tuned.
  • the failure types include: failures of different elements in the entity database and failures of different elements in the inter-entity relationship database.
  • the determining the entity in the wireless communication network knowledge map to which the fault type belongs based on the fault type and an intelligent reasoning algorithm includes:
  • the location includes the specific communication layer where the entity is located or the location of the graph structure, and the influencing factors are the location of the entity The extent to which the entity is influenced by other entities with which it has an inter-entity relationship.
  • the plurality of wireless communication network optimization strategies all include specific adjustable parameters and adjusted values of the adjustable parameters.
  • the intelligent decision-making algorithm includes one or more of deep reinforcement learning DRL, Markov decision process, bionic algorithm and statistical learning algorithm.
  • determining the intelligent algorithm that needs to be adjusted based on the degree of improvement and the comprehensive capability evaluation results of the various intelligent algorithms includes:
  • the specific intelligent algorithm that needs to be adjusted is determined and updated.
  • the intelligent analysis and application method of the wireless communication network knowledge map provided by this disclosure, by acquiring and storing the relevant data of the wireless communication network, combining the knowledge of communication experts and the 3GPP communication protocol, initially constructs the knowledge map of the wireless communication network and carries out the depth of the correlation relationship of endogenous factors Analysis, using the analysis results to complete the optimization and update of the map, use the intelligent reasoning module to trace the source and diagnose the location after the abnormal network communication is found, use the strategy generation module to provide several feasible optimization strategies, and choose one by the strategy decision module The specific strategy plan is sent to the wireless communication network for execution. Finally, the performance evaluation module evaluates the execution results of the optimization strategy and the comprehensive ability of the intelligent algorithm, completing the closed-loop intelligent decision-making, and effectively improving the performance of the wireless communication network and the quality of user experience.
  • Fig. 3 is a schematic diagram of the overall process of the intelligent analysis and application method of the wireless communication network knowledge map provided by the present disclosure. As shown in Fig. 3, the overall process steps include:
  • Step 1 Raw data collection: build a wireless data unit based on the ES-Hadoop distributed system, collect, convert and clean the communication data in the wireless communication network through soft and hard data collection, and classify and store according to different data types. And perform corresponding preprocessing to convert the data into a format that can be used later.
  • These data specifically include wireless air interface data (terminal side and base station side), core network data, network management data, etc.
  • Step 2 Construction and analysis of the knowledge map: According to the wireless communication network protocol regulations, such as the 3GPP communication protocol, or combined with the knowledge of communication experts, first sort out the wireless communication network indicators and data fields as the entity nodes of the wireless communication network knowledge map, and initialize the entity library.
  • entity types can be divided into four categories: network-level performance evaluation indicators, user-level performance evaluation indicators, general non-tunable data parameters, and adjustable data parameters.
  • the entity also includes entity attributes such as the communication layer, type, adjustability, and parameter value of the entity object. If a new entity is found in the subsequent in-depth analysis, the entity library is updated according to the newly discovered entity.
  • the relationship between indicators, the relationship between indicators and data fields, and the relationship between data fields are determined.
  • the relationship between entities and entities can usually be divided into causal relationships. , implicit relationship and explicit relationship, to initialize the relationship library between entities. If a new inter-entity relationship is found in the subsequent in-depth analysis, the inter-entity relationship library is updated according to the newly discovered inter-entity relationship.
  • the causal relationship describes the direct impact between two entities, such as the physical layer transmission rate and physical layer throughput; the implicit relationship describes the indirect impact between two entities, such as the highest modulation mode and physical layer throughput; explicit The relationship describes the specific analytical expressions that can be obtained after reasoning and analysis between two entities, such as the reference signal received power and synchronization signal received power of beam 1.
  • a general triplet of the wireless communication network is constructed, namely (head entity, relation, tail entity).
  • head entity a general triplet of the wireless communication network
  • tail entity a general triplet of the wireless communication network
  • a preliminary construction Knowledge graph of wireless communication network with topology a preliminary construction Knowledge graph of wireless communication network with topology.
  • the constructed wireless communication network knowledge map is updated to realize dynamic optimization of the map.
  • Step 3 Use intelligent reasoning to trace the source: based on the optimized wireless communication network knowledge map, combined with anomaly detection algorithms, anomaly detection is performed on the wireless communication network.
  • the wireless communication network performance or user experience quality declines, that is, some indicators in the knowledge graph fluctuate beyond the preset threshold range, according to the constructed wireless communication network knowledge
  • Artificial intelligence algorithms such as cluster analysis, graph neural network, approximate reasoning, causal inference, etc., realize the retrospective reasoning that causes the fluctuation of such indicators, and specifically locate which indicators or data fields are located in which communication layer according to the entity attributes, so that Complete the process of tracing the source of network performance changes.
  • Step 4 Strategy generation and decision-making: After the intelligent reasoning process, strategy generation and strategy decision-making are performed. specifically,
  • Strategic decision-making Construct intelligent decision-making algorithms based on artificial intelligence, bionic algorithms, statistical learning and other theories, such as deep reinforcement learning DRL and Markov decision process MDP, use the generated various network optimization strategies as input, and use intelligent decision-making algorithms , considering the execution efficiency and effect of the optimization strategy comprehensively, a wireless communication network optimization strategy is selected as the final decision-making strategy.
  • Step 5 Policy distribution and performance evaluation: After intelligent decision-making, implement policy distribution and performance evaluation. specifically,
  • the network optimization strategy selected by the policy decision module is issued to the wireless communication network, and the corresponding adjustable parameters are adjusted according to the strategy to complete the network optimization process.
  • Performance evaluation For a specific application scenario, establish an intelligent evaluation model for wireless communication networks. After implementing the final network optimization strategy, evaluate whether the corresponding performance indicators have been improved or reached the expected goals, and whether other indicators that characterize the performance of wireless communication networks It will cause obvious impact, and then evaluate the effectiveness, real-time and robustness of the network optimization strategy from two aspects of network-level performance and user-level performance, and at the same time evaluate the comprehensive ability of each module using intelligent algorithms and feed back the corresponding evaluation results to The strategy decision module, strategy generation module, intelligent reasoning module, and in-depth analysis module are used for each module to adjust and optimize its own methods and output results according to the specific situation, forming a closed-loop feedback.
  • Figure 4 is a schematic diagram of the business process of the intelligent analysis and application method of the wireless communication network knowledge graph provided by the present disclosure. The process roughly includes:
  • Step 1 Collect data.
  • the communication data in the comprehensive test network are collected, converted and cleaned by means of soft and hard data collection, and are classified and stored according to different data types, and format conversion is performed.
  • the collected data required in this use case mainly includes wireless air interface data (such as terminal signal-to-noise ratio, throughput, etc.) and core network data (N1/N2/N3, etc.).
  • Step 2 Build a knowledge map.
  • wireless communication network protocol regulations such as 3GPP communication protocols or combined with expert knowledge in the communication field, determine the endogenous factors of the wireless communication network composed of indicators and data fields specified therein, sort out the indicators and data fields specified in the wireless communication network, and clarify Attributes such as communication layer, type, adjustability, and parameter value are used as entity nodes of the knowledge graph. Clarify the relationship between entities and entities, construct a triplet, namely (head entity, relationship, tail entity), and construct a wireless communication network knowledge graph with a topological structure, as shown in Figure 5, which shows a knowledge graph of air interface delay Partial schematic.
  • the entity whose entity type is determined as general non-modular data parameters includes: physical uplink shared channel (Physical Uplink Share Channel, PUSCH) block error rate, uplink (Uplink, UL) total response times per second, uplink negative response per second
  • PUSCH Physical Uplink Share Channel
  • PUSCH Physical Uplink Share Channel
  • Uplink Uplink
  • UL uplink
  • NR New Radio
  • CCE Control Channel Elements
  • ILLA Inner Loop Link Adaptation
  • Entities that determine the entity type as adjustable data parameters include: Outer Loop Link Adaptation offset (OLLA offset), uplink non-retransmission average modulation and coding strategy (Modulation and Coding Scheme, MCS), Uplink binary phase shift keying (Binary Phase Shift Keying, BPSK) modulation ratio per second, uplink quadrature phase shift keying (Quadrature Phase Shift Keying, QPSK) modulation ratio per second, uplink 16 quadrature amplitude modulation per second ( 16 Quadrature Amplitude Modulation, 16QAM) modulation ratio, uplink 64 quadrature amplitude modulation (64 Quadrature Amplitude Modulation, 64QAM) modulation ratio per second, uplink 256 quadrature amplitude modulation (256 Quadrature Amplitude Modulation, 256QAM) modulation ratio per second .
  • the key indicators of the wireless communication network in this embodiment are: network delay and bit error rate at each stage of the air interface and core network.
  • the data field involves air interface and
  • Step 3 In-depth analysis and map tuning.
  • In-depth analysis of the constructed wireless communication network knowledge graph is carried out.
  • the theoretical analysis algorithm is used to deeply mine the knowledge graph structure (such as nodes, edges) and the relationship between endogenous factors (relationships between entities);
  • a data-driven approach is used to conduct in-depth analysis of the constructed wireless communication network knowledge graph. From the above two aspects, in-depth analysis of the relationship between delay and endogenous factors between each node, such as signal-to-noise ratio, modulation and coding selection, network switching, etc. Based on the in-depth analysis results, the wireless communication network knowledge map is dynamically optimized.
  • Step four detection and reasoning.
  • the key KPI indicators in wireless network performance fluctuate or are abnormal, such as delay and bit error rate increase suddenly, based on the constructed wireless communication network knowledge map and the relationship between nodes (entity relationship), based on graph neural network Algorithms such as network, approximate reasoning, and causal inference perform retrospective reasoning to locate the indicators or data field problems of which layers.
  • the bit error rate is higher than 10 -5
  • the air interface delay is higher than 1 ms
  • Step 5 Recommended tuning.
  • the strategy generation and decision-making algorithm based on association rules and deep learning, based on the constructed wireless communication network knowledge map and the relationship between endogenous factors between nodes (entities), adopts theoretical analysis algorithms such as artificial intelligence, bionic algorithms, and statistical learning to establish
  • the network parameter adjustment model is based on the iterative optimization of the algorithm based on the collected data, and the adjustment parameters are solved. Considering the execution efficiency and effect of the optimization strategy, the optimal wireless communication network optimization strategy is selected as the final decision-making strategy.
  • Step 6 Send the network optimization decision-making strategy selected by the strategy decision-making module to the comprehensive test network, and adjust the corresponding adjustable parameters according to the strategy to complete the network optimization process.
  • the delivered tuning strategy includes media access control layer (Media Access Control, MAC) and physical layer parameters, such as modulation and coding strategy MCS, beamforming matrix, and the like.
  • Fig. 6 is a schematic structural diagram of an electronic device provided by the present disclosure; as shown in Fig. 6, the electronic device includes a memory 620, a transceiver 610 and a processor 600; wherein, the processor 600 and the memory 620 can also be physically separated layout.
  • the memory 620 is used to store computer programs; the transceiver 610 is used to send and receive data under the control of the processor 600 .
  • the transceiver 610 is used to receive and transmit data under the control of the processor 600 .
  • the bus architecture may include any number of interconnected buses and bridges, specifically one or more processors represented by the processor 600 and various circuits of the memory represented by the memory 620 are linked together.
  • the bus architecture can also link together various other circuits such as peripherals, voltage regulators, and power management circuits, etc., which are well known in the art and therefore will not be further described in this application.
  • the bus interface provides the interface.
  • Transceiver 610 may be a plurality of elements, including a transmitter and a receiver, providing a unit for communicating with various other devices over transmission media, including wireless channels, wired channels, optical cables, and other transmission media.
  • the processor 600 is responsible for managing the bus architecture and general processing, and the memory 620 can store data used by the processor 600 when performing operations.
  • the processor 600 may be a central processing unit (Central Processing Unit, CPU), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field-programmable gate array (Field-Programmable Gate Array, FPGA) or a complex programmable logic device (Complex Programmable Logic Device, CPLD), the processor can also adopt a multi-core architecture.
  • CPU Central Processing Unit
  • ASIC Application Specific Integrated Circuit
  • FPGA Field-Programmable Gate Array
  • CPLD Complex Programmable Logic Device
  • the processor 600 is used to execute any of the methods provided in the embodiments of the present application according to the obtained executable instructions by calling the computer program stored in the memory 620, for example:
  • the knowledge map of the wireless communication network is constructed, and combined with the classified wireless communication network data for in-depth analysis and map cycle optimization;
  • the wireless communication network knowledge map is constructed based on the endogenous factors of the wireless communication network and the map construction method, and combined with the classified wireless communication network data for in-depth analysis and map cycle optimization, including:
  • the entity library and the inter-entity relationship library are determined based on endogenous factors of the wireless communication network; the endogenous factors of the wireless communication network include indicators and data fields specified according to wireless communication network protocols.
  • the determination of the diagnosis and positioning result of the wireless communication network abnormality based on the optimized wireless communication network knowledge map, and anomaly detection and intelligent reasoning algorithms includes:
  • anomaly detection is performed on the wireless communication network to determine a fault type
  • An optimal tuning strategy is used to tune the wireless communication network, including:
  • an optimal optimization strategy is determined to optimize the wireless communication network.
  • the method also includes:
  • the optimization strategy determines the degree of improvement of the wireless communication network by the optimization strategy at the network level performance and user level performance, and the The comprehensive capability evaluation results of each intelligent algorithm adopted, including:
  • the intelligent algorithms include: in-depth analysis algorithms, abnormal detection algorithms, intelligent reasoning algorithms, strategy generation algorithms and intelligent decision-making algorithms one or more of .
  • the wireless communication network knowledge map is constructed based on the endogenous factors of the wireless communication network and the map construction method, and before in-depth analysis and map cycle optimization are performed in combination with the classified wireless communication network data, it also includes:
  • the raw data of the wireless communication network is obtained and preprocessed, and then classified and stored according to different data types; the classified storage of different data types is realized based on a distributed system architecture.
  • the different data types include: one or more of terminal side wireless air interface data, base station side wireless air interface data, core network data, and network management data.
  • the entity types of the elements in the entity library include one or more of network-level performance evaluation indicators, user-level performance evaluation indicators, general non-adjustable data parameters and adjustable data parameters; the relationship between entities Entity relationships in the repository include one or more of causal, implicit, and explicit relationships.
  • the analysis algorithm based on the graph model and the data model determines new relationships between entities in the knowledge map of the wireless communication network, analyzes the degree of association of the relationships between entities, and updates the relationship library between entities, and/or Or determine one or more of new entities, new entity types, and new entity attributes in the wireless communication network knowledge map, and update the entity library, including:
  • updating the wireless communication network knowledge graph based on the updated entity database and/or the updated inter-entity relationship database includes:
  • wireless communication network One or more of entities in the knowledge graph, parameters of entity attributes, and parameters of relationships between entities;
  • the wireless communication network knowledge graph is updated based on the parameters that need to be tuned.
  • the failure types include: failures of different elements in the entity database and failures of different elements in the inter-entity relationship database.
  • the determining the entity in the wireless communication network knowledge map to which the fault type belongs based on the fault type and an intelligent reasoning algorithm includes:
  • the location includes the specific communication layer where the entity is located or the location of the graph structure, and the influencing factors are the location of the entity The extent to which the entity is influenced by other entities with which it has an inter-entity relationship.
  • the intelligent decision-making algorithm includes one or more of deep reinforcement learning DRL, Markov decision process, bionic algorithm and statistical learning algorithm.
  • determining the intelligent algorithm that needs to be adjusted based on the degree of improvement and the comprehensive capability evaluation results of the various intelligent algorithms includes:
  • the present disclosure also provides a computer program product
  • the computer program product includes a computer program stored on a non-transitory computer-readable storage medium
  • the computer program includes program instructions, and when the program instructions are executed by a computer When executing, the computer can execute the steps of the intelligent analysis and application method of the wireless communication network knowledge map provided by the above embodiments, for example including:
  • the knowledge map of the wireless communication network is constructed, and combined with the classified wireless communication network data for in-depth analysis and map cycle optimization;
  • a policy is used to optimize the wireless communication network.
  • the steps also include:
  • the wireless communication network knowledge map is constructed based on the endogenous factors of the wireless communication network and the map construction method, and before in-depth analysis and map cycle optimization are performed in combination with the classified wireless communication network data, it also includes:
  • the raw data of the wireless communication network is obtained and preprocessed, and then classified and stored according to different data types; the classified storage of different data types is realized based on a distributed system architecture.
  • the embodiments of the present application also provide a processor-readable storage medium, the processor-readable storage medium stores a computer program, and the computer program is used to make the processor execute the above-mentioned embodiments.
  • the steps of the intelligent analysis and application method of the wireless communication network knowledge map include, for example:
  • the knowledge map of the wireless communication network is constructed, and combined with the classified wireless communication network data for in-depth analysis and map cycle optimization;
  • a policy is used to optimize the wireless communication network.
  • the steps also include:
  • the wireless communication network knowledge map is constructed based on the endogenous factors of the wireless communication network and the map construction method, and before in-depth analysis and map cycle optimization are performed in combination with the classified wireless communication network data, it also includes:
  • the raw data of the wireless communication network is obtained and preprocessed, and then classified and stored according to different data types; the classified storage of different data types is realized based on a distributed system architecture.
  • the processor-readable storage medium can be any available medium or data storage device that can be accessed by a processor, including but not limited to magnetic storage (e.g., floppy disk, hard disk, magnetic tape, magneto-optical disk (MO), etc.), optical storage (e.g., CD, DVD, BD, HVD, etc.), and semiconductor memory (such as ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)), etc.
  • magnetic storage e.g., floppy disk, hard disk, magnetic tape, magneto-optical disk (MO), etc.
  • optical storage e.g., CD, DVD, BD, HVD, etc.
  • semiconductor memory such as ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)
  • the device embodiments described above are only illustrative, and the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in One place, or it can be distributed to multiple network elements. Part or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. It can be understood and implemented by those skilled in the art without any creative effort.
  • each implementation can be implemented by means of software plus a necessary general-purpose hardware platform, and of course also by hardware.
  • the essence of the above technical solution or the part that contributes to the prior art can be embodied in the form of software products, and the computer software products can be stored in computer-readable storage media, such as ROM/RAM, magnetic discs, optical discs, etc., including several instructions to make a computer device (which may be a personal computer, server, or network device, etc.) execute the methods described in various embodiments or some parts of the embodiments.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Signal Processing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Computational Linguistics (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Animal Behavior & Ethology (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

本公开提供一种无线通信网络知识图谱的智能分析与应用系统及方法，该系统至少包括知识图谱单元、智能溯源单元和调优策略单元。知识图谱单元用于无线通信网络知识图谱构建，并结合无线通信网络数据实现图谱深度解析和循环调优；智能溯源单元用于对网络异常检测以及追因溯源；调优策略单元用于生成若干调优策略，并确定最优调优策略，对所述无线通信网络进行调优。本公开提供的上述系统，形成智能闭环反馈，对无线通信网络进行循环调优，有效提高无线通信网络性能和用户体验质量。

Description

无线通信网络知识图谱的智能分析与应用系统及方法
相关申请的交叉引用
本申请要求于2022年02月17日提交的申请号为202210144211.7，发明名称为“无线通信网络知识图谱的智能分析与应用架构及方法”的中国专利申请的优先权，其通过引用方式全部并入本文。
技术领域
本公开涉及无线通信网络智能传输与管控技术领域，尤其涉及一种无线通信网络知识图谱的智能分析与应用系统及方法。
背景技术
随着无线移动通信技术的不断发展，人们对带宽、数据传输速率、时延、可靠性等指标提出了越来越高的要求。尤其是如今迈进了5G时代，引入了增强移动带宽(Enhanced Mobile Broadband，eMBB)、海量机器类通信(massive Machine Type of Communication，mMTC)和超可靠低时延通信(Ultra-reliable and Low Latency Communications，uRLLC)三大典型应用场景，使得无线通信网络的结构越来越复杂，功能越来越多样化。此外，无线通信网络的终端类型及行为、数据业务需求、系统资源具有较高的动态性、较强的时效性和相互耦合性等显著特征。
一个无线通信网络由用户终端、接入网和核心网等诸多要素构成，网络结构通常比较复杂，包含了数以万计的性能指标和数据字段，涉及到不同的网元、协议栈等。这些指标通常用来表征无线通信网络的性能，是衡量一个无线通信网络运行状态的重要参数。另一方面，这些指标又与用户体验质量(Quality of Experience，QoE)紧密相关。鉴于无线网络结构的复杂性以及业务类型的多样性，这些无线通信网络数据之间内生因素关联关系通常比较复杂。因此，影响无线通信网络性能和用户体验质量的因素十分繁多，有效地厘清各指标与数据字段之间的关联关系，才能更好地去理解整个无线通信网络的结构与运行机制，进而在网络性能恶化或者用户体验质量下降时，精准地定位出引起这些变化的原因，从而针对具体原因给出相应的调优方案。
发明内容
针对现有技术存在的问题，本公开提供一种无线通信网络知识图谱的智能分析与应用系统及方法。
第一方面，本公开提供一种无线通信网络知识图谱的智能分析与应用系统，至少包括：
知识图谱单元、智能溯源单元和调优策略单元；
所述知识图谱单元，用于基于无线通信网络的内生因素以及图谱构建方法，构建无线通信网络知识图谱，并结合分类后的无线通信网络数据进行深度解析和图谱循环调优；
所述智能溯源单元，用于基于调优后的无线通信网络知识图谱，以及异常检测和智能推理算法，确定无线通信网络异常的诊断定位结果；
所述调优策略单元，用于基于策略生成算法和所述诊断定位结果，确定若干调优策略，并以调优策略执行效率和效果最好为目标，结合智能决策算法，在若干调优策略中确定一种最优的调优策略，对所述无线通信网络进行调优。
第二方面，本公开还提供一种无线通信网络知识图谱的智能分析与应用方法，包括：
基于无线通信网络的内生因素以及图谱构建方法，构建无线通信网络知识图谱，并结合分类后的无线通信网络数据进行深度解析和图谱循环调优；
基于调优后的无线通信网络知识图谱，以及异常检测和智能推理算法，确定无线通信网络异常的诊断定位结果；
基于策略生成算法和所述诊断定位结果，确定若干调优策略，并以调优策略执行效率和效果最好为目标，结合智能决策算法，在若干调优策略中确定一种最优的调优策略，对所述无线通信网络进行调优。
第三方面，本公开还提供一种电子设备，包括存储器，收发机，处理器；
存储器，用于存储计算机程序；收发机，用于在所述处理器的控制下收发数据；处理器，用于读取所述存储器中的计算机程序并实现如上所述第二方面所述无线通信网络知识图谱的智能分析与应用方法的步骤。
第四方面，本公开还提供一种非暂态计算机可读存储介质，其上存储有计算机程序，该计算机程序被处理器执行时实现如上所述第二方面所述的无线通信网络知识图谱的智能分析与应用方法的步骤。
第五方面，本公开还提供一种计算机程序产品，包括计算机程序，所述计算机程序被处理器执行时实现如上所述第二方面所述的无线通信网络知识图谱的智能分析与应用方法的步骤。
本公开提供的无线通信网络知识图谱的智能分析与应用系统及方法，通过无线通信网络的内生因素，结合通信专家知识及3GPP通信协议初步构建无线通信网络知识图谱并进行内生因素关联关系深度解析，利用解析结果完成图谱调优更新，在发现网络通信异常后利用智能推理模块进行追因溯源和诊断定位，利用策略生成模块提供若干种可行的调优策略，并由策略决策模块选择一种具体的策略下发给无线通信网络执行，对所述无线通信网络进行循环调优，有效提高无线通信网络性能和用户体验质量。
附图说明
为了更清楚地说明本公开或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作一简单地介绍，显而易见地，下面描述中的附图是本公开的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1是本公开提供的无线通信网络知识图谱的智能分析与应用系统示意图；
图2是本公开提供的无线通信网络知识图谱的智能分析与应用方法的流程示意图；
图3是本公开提供的无线通信网络知识图谱的智能分析与应用方法的整体流程示意图；
图4是本公开提供的无线通信网络知识图谱的智能分析与应用方法的业务流程示意图；
图5是本公开提供的无线通信网络知识图谱的空口时延知识图谱局部示意图；
图6是本公开提供的一种电子设备的结构示意图。
具体实施方式
为使本公开的目的、技术方案和优点更加清楚，下面将结合本公开中的附图，对本公开中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本公开一部分实施例，而不是全部的实施例。基于本公开中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本公开保护的范围。
下面结合图1-图6描述本公开的无线通信网络知识图谱的智能分析与应用系统及方法。
图1是本公开提供的无线通信网络知识图谱的智能分析与应用系统示意图；如图1所示，该无线通信网络知识图谱的智能分析与应用系统，至少包括：知识图谱单元、智能溯源单元和调优策略单元；
所述知识图谱单元，用于基于无线通信网络的内生因素以及图谱构建方法，构建无线通信网络知识图谱，并结合分类后的无线通信网络数据进行深度解析和图谱循环调优；
所述智能溯源单元，用于基于调优后的无线通信网络知识图谱，以及异常检测和智能推理算法，确定无线通信网络异常的诊断定位结果；
所述调优策略单元，用于基于策略生成算法和所述诊断定位结果，确定若干调优策略，并以调优策略执行效率和效果最好为目标，结合智能决策算法，在若干推荐调优策略中确定一种最优的调优策略，对所述无线通信网络进行调优。
具体地，鉴于无线网络结构的复杂性以及业务类型的多样性，这些无线通信网络数据之间内生因素关联关系通常比较复杂。因此，影响无线通信网络性能和用户体验质量的因素十分繁多，只有有效地厘清各指标与数据字段之间的关联关系，才能更好地去理解整个无线通信网络的结构与运行机制，为提升无线通信网络性能，改善用户体验质量，本公开提出了一 种无线通信网络知识图谱的智能分析与应用系统及方法，为实现无线通信网络的智能化提供了有力的支持。
所述系统至少包括：知识图谱单元、智能溯源单元和调优策略单元；其中，
(1)知识图谱单元，用于基于无线通信网络的内生因素以及图谱构建方法，构建无线通信网络知识图谱，并结合分类后的无线通信网络数据进行深度解析和图谱循环调优；其中，无线通信网络知识图谱是基于无线通信网络的内生因素初始化后的实体库中元素，以及实体间关系库中元素组成的三元组构建的，所述无线通信网络的内生因素包括根据无线通信网络协议规定的指标和数据字段；所述分类后的无线通信网络数据可以是根据协议直接分类的，或者认为设定的规则进行分类的，主要是便于后续数据处理，提高处理的效率。
可选地，在本公开的实施例中，所述知识图谱单元至少包括深度解析模块和图谱调优模块；
其中，所述深度解析模块，用于基于图模型和数据模型的分析算法，确定无线通信网络知识图谱中新的实体、新的实体类型、新的实体属性和各实体间的新的关系中的一个或多个，并完善各实体间对应关系的关联程度；
所述图谱调优模块，用于基于更新后的所述实体库和/或更新后的所述实体间关系库，更新所述无线通信网络知识图谱；
具体地，知识图谱单元至少包括：深度解析模块和图谱调优模块；
深度解析模块将初始化构成无线通信网络知识图谱的实体库和实体间关系库作为输入，深度解析模块主要有两个功能。其一是，从图模型角度，对两个实体间关系的关联程度进行量化和度量，确立边连接强度；其二是，从数据角度深入分析无线通信网络知识图谱中新的实体、新的实体类型、新的实体属性和新的实体间关系及关联程度中的一个或者多个。其中，所述初始化构成无线通信网络知识图谱的实体库以及实体间关系库是基于无线通信网络的内生因素确定的，上述无线通信网络的内生因素包括根据无线通信网络协议规定的指标和数据字段。所述实体库中每个元素包括该元素对应的实体类型，该元素具有的所有属性；所述实体间关系库包含每两个元素之间的具体关系，以及每两个元素之间所述具体关系的关联程度等。
可选地，在本公开实施例中，所述基于图模型和数据模型的分析方法，确定无线通信网络知识图谱中各实体间新的关系，并分析各实体间关系的关联程度，更新所述实体间关系库，和/或确定无线通信网络知识图谱中新的实体、新的实体类型和新的实体属性中的一个或多个，更新所述实体库，包括：
基于图模型算法，确定无线通信网络知识图谱中各实体间关系的关联程度的量化度量，更新所述实体间关系库；
基于获取的分类后的无线通信网络数据，确定无线通信网络知识图谱中新的实体、新的实体类型、新的实体属性和各实体间新的关系中的一个或多个，更新所述实体库和/或所述 实体间关系库。
具体地，深度解析模块主要功能之一：该模块开展数据驱动的无线通信网络知识图谱深度解析，利用无线数据单元采集的无线通信网络实时数据，从数据角度深入分析内生因素知识图谱中节点(实体)间的关联程度及依赖关系，确定新的实体、实体类型、新的实体属性和新的实体间关系中的一个或多个。深度解析模块主要功能之二是，深度解析模块基于浅显解析后所构建的无线通信网络知识图谱，依托基于图模型的理论分析方法，深入挖掘各节点之间的内生因素关联关系，如指标与数据字段、指标与指标、数据字段与数据字段等，即确定各实体间关系的关联程度或依赖关系。对两个节点(实体)间的关联程度和依赖关系进行量化和度量，确立边连接强度。
对无线网络知识图谱进行深度解析之后，可能获得无线网络知识图谱中新增的实体，或者实体关系等，则需要该无线网络知识图谱进行优化，则主要由图谱调优模块来完成，该模块主要用于确定无线通信网络知识图谱的调优内容和调优方向。
可选地，在本公开实施例中，基于所述更新后的所述实体库和/或更新后的所述实体间关系库，更新所述无线通信网络知识图谱，包括：
基于无线通信网络协议规定和理论分析算法，以及所述更新后的所述实体库和/或更新后的所述实体间关系库，确定需要调优的参数，所述需要调优的参数包括：无线通信网络知识图谱中实体、实体属性的参数和实体间关系相关的参数中的一个或多个；
基于所述需要调优的参数，更新所述无线通信网络知识图谱。
具体地，图谱调优模块以深度解析模块的相关分析结果作为输入，即所述更新后的所述实体库和/或更新后的所述实体间关系库作为输入，输出无线通信网络知识图谱的调优内容和调优方向，包括实体的增加或者删除，可理解为图结构的调优更新(例如错误图结构的勘误、遗漏图结构的补全等)、实体间关系的调优更新(例如多余关联关系的删除、错误关联关系的更正或将某两个实体间关系的关联程度从关联程度2修改为关联程度3等)，具体的关联程度数值可以表示两实体间关系的关联程度的强弱，数值越大可以表示关联程度越强，也可以表示关联程度越弱，或者实体的属性的参数的更新(例如某个实体类型的实体对应的属性中参数的可调性从可调修改为不可调，实体所在位置等信息需要修改等)等。同时，将调优更新参数反馈给图谱构建模块以完成动态调整，实现知识图谱的循环调优。
可选地，该知识图谱单元还可以包括：浅显解析模块。浅显解析模块主要通过无线通信网络协议规定，比如3GPP通信协议或者结合通信领域专家知识，确定其中规定的指标和数据字段等构成的无线通信网络内生因素，基于无线通信网络内生因素确定若干实体类型及若干实体间关系，所述内生因素主要包括根据无线通信网络协议规定的指标和数据字段，也就是对无线通信网络所涉及的指标和数据字段进行梳理，确定实体库的初始元素，比如各指标包括无线资源控制(Radio Resource Control，RRC)连接建立成功率、切换成功率、上下行丢包率等；数据字段包括子载波间隔、上下行时隙配比、每秒下行最高调制方式等。上述实 体库中每个元素均可以作为一个实体，实体包含可以表征实体对象特征参数的实体属性，例如实体名称、类型、字长、所属通信层、取值范围、可调性等。此外浅显解析模块还会对各指标(例如RRC连接建立成功率、切换成功率、上下行丢包率等)和数据字段(例如子载波间隔、上下行时隙配比、每秒下行最高调制方式等)进行分类，确定上述实体库中元素(即各种实体)的实体类型，记为{entity 1,entity 2,...,entity m}，共有m种实体类型，每一个实体均属于这m种实体类型之一。针对定义的实体，对实体间的关联关系进行浅显解析，依据浅显解析模块对各指标和数据字段之间关联关系的解析结果，确立若干实体与实体之间的关系。上述确定的若干实体与实体之间的关系构成初始化的实体间关系库，所述实体间关系库包括关联的两实体，以及两实体间关系，其中，两实体间关系可记为{relation 1,relation 2,...,relation n}，共n种实体关系，任意两个实体之间的关系均属于这n种实体关系之一。并且根据内生因素的特性，在对无线通信网络数据的进一步分析过程中，若出现新的实体，新的实体类型，和新的实体属性中的任意组合，则更新上述实体库；若出现新的实体间关系，则更新上述实体间关系库；若同时出现新的实体，新的实体类型和新的实体属性的任意组合，和新的实体间关系，则同时更新实体库和实体间关系库。
可选地，该知识图谱单元还可以包括：图谱构建模块。图谱构建模块主要用于构建一个具有拓扑结构的无线通信网络知识图谱。依据初始化的实体库和实体间关系库，构建无线通信网络通用三元组(head，relation，tail)，其中，head为三元组中的头实体，tail为三元组中的尾实体，每个三元组中的头实体和尾实体均属于上述实体库中确立的m种实体类型之一，relation为实体间的关系，属于上述实体间关系库中确立的n种关系之一。然后，构建无线通信网络知识图谱，以三元组中的头实体和尾实体分别作为无线通信网络知识图谱的节点。节点与节点之间如果有明确的指向关系，则用一条有向线段来连接，否则用一条无向线段连接。连接所有存在边连接关系的节点，即可得到具有拓扑结构的无线通信网络知识图谱，通常是一种多关系图。
可选地，在本公开实施例中，所述实体库中元素的实体类型包括网络级性能评估指标，用户级性能评估指标，通用非调型数据参数和可调型数据参数中的一个或多个；所述实体间关系库中的实体关系包括因果关系，隐式关系和显式关系中的一个或多个。
具体地，为了更清晰地说明本公开中实体库，初始化时确定上述实体库中元素的实体类型包括网络级性能评估指标，用户级性能评估指标，通用非调型数据参数和可调型数据参数四类中的一个或者多个，当然可能存在其他的实体类型。同时，实体也包含了实体对象所属通信层、类型、可调性、参数值等实体属性。而且实体间关系库初始化时，确定上述实体间关系通常可以分为因果关系、隐式关系和显式关系三类中的一个或多个，当然可能还有其他的实体间关系。其中，因果关系描述了两个实体间的直接影响，例如物理层传输速率与物理层吞吐量；隐式关系描述了两个实体间的间接影响，没有明确具体的表达式可表示两者间的 关系，例如最高调制方式与物理层吞吐量；显式关系描述了两个实体间经过推理分析后可得到具体的解析表达式，例如波束1的参考信号接收功率与同步信号接收功率。以上仅仅示例性说明，不对本公开中实体类型的具体分类和实体间关系的具体分类做限定。实际本公开中对实体类型的分类可以包括更多，实体间关系的分类也可以包括更多，均是基于无线通信网络的内生因素动态确定的。
可选地，在本公开实施例中，所述智能溯源单元包括异常检测模块和智能推理模块；
其中，所述异常检测模块，用于基于所述调优后的无线通信网络知识图谱，结合异常检测算法，对无线通信网络进行异常检测，确定故障类型；
智能推理模块，用于基于所述故障类型以及智能推理算法，确定所述故障类型所属的无线通信网络知识图谱中的实体。
具体地，无线通信网络知识图谱的智能分析与应用系统还包括：
(2)智能溯源单元，该单元包括：异常检测模块、智能推理模块；具体地，
异常检测模块主要是基于异常检测算法和调优后的无线通信网络知识图谱对无线网络进行异常检测，在发现网络异常的情况下，完成对故障类型的初步检测，即分析出是哪一类实体或哪一部分图谱结构或者哪一部分的实体间关系发生了异常，其中包括出现异常的节点(实体)、边(实体间关系)、子图结构(需要增加或者删除实体)等，为后续追因溯源指明方向。
可选地，在本公开实施例中，所述故障类型包括：实体库中不同元素的故障和实体间关系库中不同元素的故障。所述实体库中不同元素的故障，即不同实体的故障，主要指哪一类实体发生了异常。所述实体间关系库中不同元素的故障，主要指哪一部分图谱结构或实体间关系，或图谱结构和实体间关系发生了异常。
智能推理模块主要用于确定故障类型所属的实体以及实体属性。
可选地，在本公开实施例中，所述基于所述故障类型以及智能推理算法，确定所述故障类型所属的无线通信网络知识图谱中的实体，包括：
确定所述故障类型所属的实体，所述实体是实体库中特定的元素；
基于所述实体的属性，确定所述实体所在位置、可调性以及影响因素，所述所在位置包括所述实体位于的具体通信层或所处图谱结构的位置，所述影响因素是所述实体和与其存在实体间关系的其他实体对所述实体的影响程度。
具体地，智能推理模块以异常检测模块得出的故障类型为输入，依托构建的无线通信网络知识图谱及其内生因素关联关系，借助于近似推理、因果推断等人工智能方法，完成影响网络性能问题的追因溯源，即深入挖掘出是哪一层的哪些指标或数据字段造成了相应指标的异常波动，也就确定了实体的具体实体类型。上述故障类型可能有一个或者多个，对应的实体也可能有一个或者多个。依据该实体所属的实体属性甄别哪些数据字段是可调参数，同时推理出引起这些指标或数据字段变化的具体原因(例如：无线信道衰落、网络拥塞等)，并 根据实体间关系，确定该实体存在关联的其他实体，用同样的方法推理出关联实体的哪些数据字段是可调参数，以及引起这些指标或者数据字段变化的具体原因。
可选地，在本公开实施例中，所述调优策略单元包括策略生成模块和策略决策模块；
其中，所述策略生成模块，用于基于所述故障类型所属的无线通信网络知识图谱中的实体，以及策略生成算法，确定若干无线通信网络调优策略；
所述策略决策模块，用于基于智能决策算法，以所述调优策略的执行效率和效果最好为目标，确定一个最优的调优策略，对所述无线通信网络进行优化。
具体地，无线通信网络知识图谱的智能分析与应用系统还包括：
(3)调优策略单元，该单元包括：策略生成模块、策略决策模块；具体地，所述策略生成模块，基于所述诊断定位结果，以及策略生成算法，确定若干无线通信网络调优策略；
策略生成模块主要以智能推理模块的诊断定位结果作为输入，在明确了引起性能指标波动的数据字段所属位置、可调性、影响因素的前提下，利用策略生成算法，提供若干种可能的通信网络调优策略，包括调节哪些数据字段参数，以及它们的调节方向。鉴于无线通信网络的复杂性，这样的调优策略往往是不唯一的。
可选地，在本公开实施例中，所述若干无线通信网络调优策略均包括具体的可调参数，以及可调参数调整的数值。
即所述无线通信网络调优策略中可以包括具体的可调型数据参数，以及可调型参数需要调整的数值。
策略决策模块将策略生成模块提供的若干种可行的通信网络调优策略作为输入，依据智能决策算法，综合考虑调优策略的执行效率和效果最佳，从中选择一种最优的策略下发给无线通信网络执行，实现网络调优。
可选地，在本公开实施例中，所述智能决策算法包括深度强化学习DRL、马尔可夫决策过程、拟生算法和统计学习算法中的一个或多个。
策略决策模块构建基于人工智能、深度强化学习(Deep Reinforcement Learning，DRL)、马尔可夫决策过程(Markov Decision Process，MDP)、拟生算法和/或统计学习等理论的智能决策模型(算法)，以策略生成的多种网络调优策略为输入，利用智能决策算法，综合考虑调优策略的执行效率和效果，并基于执行效率和执行效果分别对应的权重，确定每个调优策略的对应的加权结果，选择一种无线通信网络调优策略作为最终决策策略。考虑上述各算法的优势，确定最合适的某一个或者某几个的结合作为智能决策算法。
可选地，在本公开实施例中，上述无线通信网络知识图谱的智能分析与应用系统还包括：无线数据单元；所述无线数据单元用于获取无线通信网络原始数据，并进行预处理后，按照不同数据类型进行分类存储；所述不同数据类型的分类存储，是基于分布式系统架构实现的。
具体地，上述无线通信网络知识图谱的智能分析与应用系统还包括：
(4)无线数据单元，该单元主要执行无线通信网络原始数据采集、并进行预处理，所述预处理包括数据格式转换、清洗等，并按照不同的类型进行分类存储；
数据采集包括通过软硬数据采集等方法对无线网络通信原始数据进行采集，软采集主要是指通过计算机程序等软件实现的数据采集，硬采集主要是对不同无线通信接口原始数据采集。
数据格式转换主要是对无线通信网络中不同数据的格式转换成统一可识别的格式，或者将某个单元可识别的格式转换成与该单元连接的下一个处理单元可识别或处理的格式；
数据清洗主要指对明显异常的数据进行删除等处理。
可选地，在本公开实施例中，所述无线数据单元包括数据采集模块，所述数据采集模块是基于分布式系统构建的。
具体地，上述分布式系统可以实现对无线通信网络数据的分布式存储和处理，尤其对于大量数据存储和处理，提供一种高效，可靠的处理和存储方式。比如，ES Hadoop分布式系统基础结构，Hadoop实现了一个分布式文件系统(Hadoop Distributed File System)，简称HDFS。HDFS有着高容错性的特点，并且设计用来部署在低廉的硬件上。而且它提供高传输率来访问应用程序的数据，适合那些有着超大数据集的应用程序。HDFS放宽了可移植操作系统接口(Portable Operating System Interface，POSIX)的要求这样可以流的形式访问文件系统中的数据。Hadoop能够对大量数据进行分布式处理。是以一种可靠、高效、可伸缩的方式进行处理的。Hadoop的可靠主要体现在其假设计算元素和存储会失败，因此维护多个工作数据副本，确保能够针对失败的节点重新分布处理。Hadoop的高效是主要体现在其并行的方式工作，通过并行处理加快处理速度。Hadoop的可伸缩性，主要体现在其能够处理PB(Petabyte，存储单位)级数据。此外，Hadoop依赖于社区服务器，成本比较低，任何人都可以使用。
可选地，在本公开实施例中，所述不同数据类型包括：终端侧无线空口数据，基站侧无线空口数据，核心网数据和网管数据中的一个或多个。
采集的数据通常可以分为四大类，但不局限于这四类数据：第一类为无线空口终端侧数据，主要包含L1、L2和L3层数据；第二类为无线空口基站侧数据，主要包含L1、L2和L3层数据；第三类为核心网数据，主要分为用户面和控制面两种数据类型；第四类为网管数据，主要包含告警、配置和性能信息。应理解，以上列举的对无线通信网络数据的分类仅为示例性说明，本公开并未限定于此，现有技术中无线通信网络中所有网元间传输的数据以及空口数据均落入本公开的保护范围内。
告警主要是对无线通信网络节点数据存在异常的告警，比如无线通信网络节点数据负载较大，或者某些节点存在异常的告警。
配置主要是对无线通信网络中各节点的配置信息。
性能信息主要指无线通信网络中网络级性能信息和用户级性能信息。
可选地，在本公开实施例中，上述无线通信网络知识图谱的智能分析与应用系统还包括：性能评估单元，该单元用于基于调优后的无线通信网络，结合调优策略和智能算法的评估模型，确定所述调优策略对无线通信网络在网络级性能和用户级性能的改善度，以及每个单元所采用的智能算法的综合能力评估结果，并反馈给对应单元。
可选地，在本公开实施例中，所述性能评估单元包括网络级性能评估模块和用户级性能评估模块；
其中，所述网络级性能评估模块和用户级性能评估模块用于获取调优后的所述无线通信网络的数据；
确定在网络级性能评估和用户级性能评估两个维度所述调优策略对无线通信网络性能的改善度；
基于所述改善度以及所述智能算法的综合能力评估结果，确定需要调整的智能算法；所述智能算法包括：深度解析算法，异常检测算法，智能推理算法，策略生成算法和智能决策算法中的一个或多个。
具体地，无线通信网络知识图谱的智能分析与应用系统还包括：性能评估单元，
(5)性能评估单元，该单元包括：网络级性能评估模块、用户级性能评估模块；具体地，
性能评估单元主要有两个输入。第一个输入是知识图谱单元、智能溯源单元和调优策略单元中各模块所采用的算法策略及相关输出，例如包括策略决策模块所选择下发给无线通信网络执行的调优策略等；第二个输入是调优策略执行以后无线通信网络中的实时数据。该单元主要分为网络级性能评估和用户级性能评估两个方向来对整个系统进行评估，确定对应的改善度。即确定调优后的无线网络是否和调优前有所改善，是提高了还是未提高对应的网络级性能或者用户级性能，具体提高或者降低的程度是多少。首先，需要评估所选调优策略的有效性、实时性和鲁棒性，包括调优方案执行之后整个通信网络的性能以及用户体验质量是否得到提高和改善(例如网络时延、上下行丢包率是否降低，RRC连接建立成功率是否提高等)。其次，评估各个模块中采用的智能算法，具体包括算法泛化能力、可扩展性、可解释性等。随着无线通信网络技术的发展，评估体系也需要不断更新，评估对象和评估方法也将不断扩充和完善。
另一方面，性能评估单元将评估结果反馈给知识图谱单元、智能溯源单元和调优策略单元中的相应模块，各模块根据评估反馈结果，结合相应的性能评估准则，对自身模块采用的方法和输出结果进行调整优化，实现网络动态调优，例如反馈结果表明策略决策模块选择的调优策略效果不佳，则该模块可以通过改变智能决策算法对调优策略进行重新选择。
可选地，在本公开实施例中，所述基于所述改善度以及所述智能算法的综合能力评估结果，确定需要调整的智能算法，包括：
若所述最优的调优策略执行后，网络级性能和/或用户级性能的改善度未改变或降低的 情况下，确定各单元采用的智能算法的综合能力评估结果；
基于各单元采用的智能算法的综合能力评估结果，结合相应的评估准则，确定需要调整的智能算法并更新。
具体地，无线通信网络执行了最优的调优方案之后，性能评估单元从网络级性能和用户级性能两个角度分别对无线网络性能的改善度进行评估，若评估结果为没有提高，或者反而降低的情况下，确定各单元采用的智能算法的综合能力评估结果。
确定各单元采用的智能算法的综合能力评估结果的方法包括：对所有单元或所有模块采用的智能算法，通过算法泛化能力、可扩展性、可解释性等进行综合评估，确定该算法的综合能力评估结果，并结合评估准则，确定需要调整的智能算法并更新。上述需要调整的智能算法可能有一个或者多个，即对应着一个或者多个单元或者模块需要调整智能算法。
基于对应模块或者单元更新后的智能算法，确定新的最优的调优方案，再次执行。若所有单元或者模块的智能算法的综合能力评估结果基本相同，则在上述若干调优方案中选择其他次优的方案进行执行。若所有单元或者模块的智能算法的综合能力评估结果比调优之前降低了，则继续重复上述确定最优的调优方案的步骤。
需要说明的是，性能评估单元与上述的知识图谱单元、智能溯源单元和调优策略单元可以构成一个智能决策闭环，无线通信网络的数据和相关信息可直接输入知识图谱单元和性能评估单元；或者为了保证无线通信网络数据的有效性，提高本系统的实施效率，增加无线数据单元对无线通信网络数据进行预处理后再输入知识图谱单元和性能评估单元，同样构成一个智能决策闭环。即本公开提供的无线通信网络知识图谱的智能分析与应用系统可以包括多种实现方式，可以包括知识图谱单元、智能溯源单元和调优策略单元，无线通信网络执行完确定的最优调优策略后，将调优后的无线通信网络输入知识图谱单元，对无线通信网络知识图谱进行更新调整；也可以包括知识图谱单元、智能溯源单元、调优策略单元和性能评估单元，无线通信网络完成确定的最优调优策略后，性能评估单元确定所述调优策略对无线通信网络在网络级性能和用户级性能的改善度，以及每个单元所采用的智能算法的综合能力评估结果，并反馈给对应单元；也可以包括知识图谱单元、智能溯源单元、调优策略单元和无线数据单元，也可以包括知识图谱单元、智能溯源单元、调优策略单元、性能评估单元和无线数据单元。
本公开提供的无线通信网络知识图谱的智能分析与应用系统，通过无线数据单元获取并存储无线通信网络的相关数据，结合通信专家知识及3GPP通信协议初步构建无线通信网络知识图谱并进行内生因素关联关系深度解析，利用解析结果完成图谱调优更新，在发现网络通信异常后利用智能推理模块进行追因溯源和诊断定位，利用策略生成模块提供若干种可行的调优方案，并由策略决策模块选择一种具体的策略方案下发给通信网络执行，最后通过性能评估模块对调优策略执行结果及智能算法综合能力进行评估，完成智能决策闭环，有效提高无线通信网络性能和用户体验质量。
图2是本公开提供的无线通信网络知识图谱的智能分析与应用方法的流程示意图；如图2所示，该方法包括：
步骤201、基于无线通信网络的内生因素以及图谱构建方法，构建无线通信网络知识图谱，并结合分类后的无线通信网络数据进行深度解析和图谱循环调优；
步骤202、基于调优后的无线通信网络知识图谱，以及异常检测和智能推理算法，确定无线通信网络异常的诊断定位结果；
步骤203、基于策略生成算法和所述诊断定位结果，确定若干调优策略，并以调优策略执行效率和效果最好为目标，结合智能决策算法，在若干推荐调优策略中确定一种最优的调优策略，对所述无线通信网络进行调优。
具体地，本公开提供的无线通信网络知识图谱的智能分析与应用方法包括：
图谱构建及调优，主要通过无线通信网络协议规定，比如3GPP通信协议或者结合通信领域专家知识，确定其中规定的指标和数据字段等构成的无线通信网络内生因素，基于无线通信网络内生因素确定若干实体类型及若干实体间关系，所述内生因素主要包括根据无线通信网络协议规定的指标和数据字段，也就是对无线通信网络所涉及的指标和数据字段进行梳理，确定实体库的初始元素，比如RRC连接建立成功率、切换成功率、上下行丢包率、子载波间隔、上下行时隙配比、每秒下行最高调制方式等。并根据各指标和数据字段之间的关联关系进行解析，确定实体间关系库的初始元素，并根据内生因素的特性，在对无线通信网络数据的进一步分析过程中，若出现新的实体，新的实体类型，和新的实体属性的任意组合，则更新上述实体库；若出现新的实体间关系，则更新上述实体间关系库；若同时出现新的实体，新的实体类型和新的实体属性的任意组合，和新的实体间关系，则同时更新实体库和实体间关系库。进而确定无线通信网络知识图谱需要调优的参数。
其中，上述根据内生因素的特性，再对无线通信网络数据进一步分析，为了保障无线通信网络数据的有效性，通常需要对原始的无线通信网络数据进行预处理，具体包括：
数据采集，通过软硬数据采集等方法对无线通信网络原始数据进行采集，软采集主要是指通过计算机程序等软件实现的数据采集，硬采集主要是对不同无线通信接口原始数据采集。
数据格式转换主要是对无线通信网络中不同数据的格式转换成统一可识别的格式，或者将某个单元可识别的格式转换成与该单元连接的下一个处理单元可识别或处理的格式；
数据清洗主要指对明显异常的数据进行删除等处理。
智能溯源，对上述调优后的无线通信网络知识图谱，结合异常检测算法，完成对故障类型的初步检测，即分析出是哪一类实体或哪一部分图谱结构或者哪一部分的实体间关系发生了异常。并结合智能推理算法，确定无线通信网络异常的诊断定位结果，即确定具体存在故障的某一个或者某几个实体或者实体间关系。
调优策略，根据上述诊断定位结果，利用策略生成算法，生成若干无线通信网络调优策 略，鉴于无线通信网络的复杂性，这样的调优策略往往是不唯一的。并以所述调优策略的执行效率和效果最好为目标，依据智能决策算法，确定一个最优的调优策略，对所述无线通信网络进行调优。
可选地，该无线通信网络知识图谱的智能分析与应用方法还可以包括基于调优后的无线通信网络，结合调优策略和智能算法的评估模型，确定调优策略对无线通信网络在网络级性能和用户级性能的改善度，以及所采用的各个智能算法的综合能力评估结果。
具体包括：性能评估，根据调优后的无线通信网络的数据，评估所述无线通信网络在在网络级性能评估和用户级性能评估两个维度的改善度，以及上述所有步骤采用的智能算法的综合能力评估结果，确定具体需要调整哪个步骤中采用的智能算法，并对该智能算法进行更新，基于更新后的智能算法，重复上述步骤产生新的调优策略执行，并再次对无线通信网络进行性能评估，确定整个无线通信网络在网络级性能评估和用户级性能评估两个维度的改善度。
本公开提供的无线通信网络知识图谱的智能分析与应用方法，通过无线通信网络的内生因素，结合通信专家知识及3GPP通信协议初步构建无线通信网络知识图谱并进行内生因素关联关系深度解析，利用解析结果完成图谱调优更新，在发现网络通信异常后利用智能推理模块进行追因溯源和诊断定位，利用策略生成模块提供若干种可行的调优方案，并由策略决策模块选择一种具体的策略方案下发给通信网络执行，对无线通信网络进行循环调优，有效提高无线通信网络性能和用户体验质量。
可选地，所述基于无线通信网络的内生因素以及图谱构建方法，构建无线通信网络知识图谱，并结合分类后的无线通信网络数据进行深度解析和图谱循环调优包括：
基于图模型和数据模型的分析算法，确定无线通信网络知识图谱中各实体间新的关系，并分析各实体间关系的关联程度，更新所述实体间关系库，和/或确定无线通信网络知识图谱中新的实体、新的实体类型和新的实体属性中的一个或多个，更新所述实体库；
基于更新后的所述实体库和/或更新后的所述实体间关系库，更新所述无线通信网络知识图谱；
所述实体库和所述实体间关系库是基于无线通信网络的内生因素确定的；所述无线通信网络的内生因素包括根据无线通信网络协议规定的指标和数据字段。
可选地，所述基于调优后的无线通信网络知识图谱，以及异常检测和智能推理算法，确定无线通信网络异常的诊断定位结果，包括：
基于所述调优后的无线通信网络知识图谱，以及异常检测算法，对无线通信网络进行异常检测，确定故障类型；
基于所述故障类型以及智能推理算法，确定所述故障类型所属的无线通信网络知识图谱中的实体。
可选地，所述基于策略生成算法和所述诊断定位结果，确定若干调优策略，并以调优策 略执行效率和效果最好为目标，结合智能决策算法，在若干调优策略中确定一种最优的调优策略，对所述无线通信网络进行调优，包括：
基于所述故障类型所属的无线通信网络知识图谱中的实体，以及策略生成算法，确定若干无线通信网络调优策略；
基于智能决策算法，以所述调优策略的执行效率和效果最好为目标，确定一个最优的调优策略，对所述无线通信网络进行调优。
可选地，所述方法还包括：
基于调优后的无线通信网络，结合调优策略和智能算法的评估模型，确定所述调优策略对无线通信网络在网络级性能和用户级性能的改善度，以及所采用的各个智能算法的综合能力评估结果。
可选地，所述基于调优后的无线通信网络，结合调优策略和智能算法的评估模型，确定所述调优策略对无线通信网络在网络级性能和用户级性能的改善度，以及所采用的各个智能算法的综合能力评估结果，包括：
获取调优后的所述无线通信网络的数据；
确定在网络级性能评估和用户级性能评估两个维度所述调优策略对无线通信网络性能的改善度；
基于所述改善度以及所述各个智能算法的综合能力评估结果，确定需要调整的智能算法；所述智能算法包括：深度解析算法，异常检测算法，智能推理算法，策略生成算法和智能决策算法中的一个或多个。
可选地，所述基于无线通信网络的内生因素以及图谱构建方法，构建无线通信网络知识图谱，并结合分类后的无线通信网络数据进行深度解析和图谱循环调优之前，还包括：
获取无线通信网络原始数据，并进行预处理后，按照不同数据类型进行分类存储；所述不同数据类型的分类存储是基于分布式系统架构实现的。
可选地，所述不同数据类型包括：终端侧无线空口数据，基站侧无线空口数据，核心网数据和网管数据中的一个或多个。
可选地，所述实体库中元素的实体类型包括网络级性能评估指标，用户级性能评估指标，通用非调型数据参数和可调型数据参数中的一个或多个；所述实体间关系库中实体关系包括因果关系，隐式关系和显式关系中的一个或多个。
可选地，所述基于图模型和数据模型的分析算法，确定无线通信网络知识图谱中各实体间新的关系，并分析各实体间关系的关联程度，更新所述实体间关系库，和/或确定无线通信网络知识图谱中新的实体、新的实体类型和新的实体属性中的一个或多个，更新所述实体库，包括：
基于图模型算法，确定无线通信网络知识图谱中各实体间关系的关联程度的量化度量，更新所述实体间关系库。
基于获取的分类后的无线通信网络数据，确定无线通信网络知识图谱中新的实体、新的实体类型、新的实体属性和各实体间新的关系中的一个或多个，更新所述实体库和/或所述实体间关系库。
可选地，所述基于更新后的所述实体库和/或更新后的所述实体间关系库，更新所述无线通信网络知识图谱，包括：
基于无线通信网络协议规定和理论分析算法，以及所述更新后的所述实体库和/或所述实体间关系库，确定需要调优的参数，所述需要调优的参数包括：无线通信网络知识图谱中实体、实体属性的参数和实体间关系的参数中的一个或多个；
基于所述需要调优的参数，更新所述无线通信网络知识图谱。
可选地，所述故障类型包括：实体库中不同元素的故障和实体间关系库中不同元素的故障。
可选地，所述基于所述故障类型以及智能推理算法，确定所述故障类型所属的无线通信网络知识图谱中的实体，包括：
确定所述故障类型所属的实体，所述实体是实体库中特定的元素；
基于所述实体的属性，确定所述实体所在位置、可调性以及影响因素，所述所在位置包括所述实体位于的具体通信层或所处图谱结构的位置，所述影响因素是所述实体和与其存在实体间关系的其他实体对所述实体的影响程度。
可选地，所述若干无线通信网络调优策略均包括具体的可调参数，以及可调参数调整的数值。
可选地，所述智能决策算法包括深度强化学习DRL、马尔可夫决策过程、拟生算法和统计学习算法中的一个或多个。
可选地，所述基于所述改善度以及所述各个智能算法的综合能力评估结果，确定需要调整的智能算法，包括：
若所述最优的调优策略执行后，网络级性能和/或用户级性能的改善未改变或者降低的情况下，确定各个智能算法的综合能力评估结果；
基于各个智能算法的综合能力评估结果，结合相应的评估准则，确定需要调整的特定智能算法并更新。
本公开提供的无线通信网络知识图谱的智能分析与应用方法，通过获取并存储无线通信网络的相关数据，结合通信专家知识及3GPP通信协议初步构建无线通信网络知识图谱并进行内生因素关联关系深度解析，利用解析结果完成图谱调优更新，在发现网络通信异常后利用智能推理模块进行追因溯源和诊断定位，利用策略生成模块提供若干种可行的调优策略，并由策略决策模块选择一种具体的策略方案下发给无线通信网络执行，最后通过性能评估模块对调优策略执行结果和智能算法综合能力进行评估，完成智能决策闭环，有效提高无线通信网络性能和用户体验质量。
下面以具体例子对本公开提供的无线通信网络知识图谱的智能分析与应用方法进行说明。
图3是本公开提供的无线通信网络知识图谱的智能分析与应用方法的整体流程示意图，如图3所示，整体流程步骤包括：
步骤一、原始数据采集：搭建基于ES-Hadoop分布式系统的无线数据单元，通过软硬数据采集等方式对无线通信网络中的通信数据进行采集、转换和清洗，按照不同的数据类型分类存储，并进行相应的预处理，将数据转化成后续可用的格式，这些数据具体包括无线空口数据(终端侧和基站侧)、核心网数据、网管数据等。
步骤二、知识图谱构建及解析：通过无线通信网络协议规定，比如3GPP通信协议，或者结合通信专家知识，首先梳理无线通信网络指标和数据字段，作为无线通信网络知识图谱的实体节点，并初始化实体库。通常实体类型可以分为网络级性能评估指标、用户级性能评估指标，通用非调型数据参数和可调型数据参数四类。同时，实体也包含了实体对象所属通信层、类型、可调性、参数值等实体属性。后续深度解析若发现新的实体，则根据新发现的实体，对该实体库进行更新。同理根据上述协议规定或者通信专家知识，确定指标之间关系，指标和数据字段之间关系，以及数据字段之间关系，基于上述关系进而确定实体与实体之间的关系通常可以分为因果关系、隐式关系和显式关系三类，初始化实体间关系库。后续深度解析若发现新的实体间关系，则根据新发现的实体间关系，对该实体间关系库进行更新。其中，因果关系描述了两个实体间的直接影响，例如物理层传输速率与物理层吞吐量；隐式关系描述了两个实体间的间接影响，例如最高调制方式与物理层吞吐量；显式关系描述了两个实体间经过推理分析后可得到具体的解析表达式，例如波束1的参考信号接收功率与同步信号接收功率。
依据确立的实体及实体间的关系，构建无线通信网络通用三元组，即(头实体，关系，尾实体)。将三元组作为无线通信网络知识图谱的基本单元，以头实体和尾实体作为知识图谱中的节点，以有向线段或无向线段来表征两个节点(实体)间的关联关系，初步构建具有拓扑结构的无线通信网络知识图谱。
一方面，根据初步构建的无线通信网络知识图谱，利用基于图模型的节点稀疏表示算法、余弦相似度计算与节点关联分析模型，深入挖掘各节点之间的内生因素关联关系，完成节点(实体)间关联程度的量化度量。另一方面，基于无线数据单元采集存储的无线通信网络运行数据，开展数据驱动的无线通信网络知识图谱深度解析，利用采集的无线通信网络实时数据，深入分析无线通信网络知识图谱中节点(实体)间的关联程度及依赖关系。根据上述两方面深度解析结果，对构建的无线通信网络知识图谱进行更新，实现图谱的动态调优。
步骤三、利用智能推理进行追因溯源：基于调优后的无线通信网络知识图谱，结合异常检测算法，对无线通信网络进行异常检测。当无线通信网络性能或用户体验质量下降时，即知识图谱中部分指标发生波动，超出预设的门限范围之外，依据构建的无线通信网络知识图 谱及其内生因素关联关系，利用基于图团体检测的聚类分析、图神经网络、近似推理、因果推断等人工智能算法，实现引起该类指标波动原因的回溯推理，根据实体属性具体定位到位于哪一通信层的哪些指标或者数据字段，从而完成网络性能变化的追因溯源过程。
步骤四、策略生成与决策：在经过智能推理过程后，进行策略生成和策略决策。具体地，
策略生成：根据无线通信网络问题诊断定位结果，得到了引起网络性能或用户体验下降的关键性能指标，并且推理出了影响该指标的具体数据字段及其属性(可调性、调节范围等)，然后利用策略生成算法，依据构建的无线通信网络知识图谱及各节点(实体)间内生因素关联关系，提供若干个可能的网络调优策略，包括调节哪些可调参数以及它们的调节方向。
策略决策：构建基于人工智能、拟生算法、统计学习等理论的智能决策算法，比如深度强化学习DRL和马尔可夫决策过程MDP，以生成的多种网络调优策略为输入，利用智能决策算法，综合考虑调优策略的执行效率和效果，选择一种无线通信网络调优策略作为最终决策策略。
步骤五、策略下发与性能评估：在智能决策之后，执行策略下发和性能评估。具体地，
策略下发：将策略决策模块选择的网络调优策略下发给无线通信网络，并按照策略对相应的可调参数进行调节，完成网络调优过程。
性能评估：针对特定的应用场景，建立一个无线通信网络智能评估模型，在执行完最终网络调优策略后，评估对应性能指标是否得到改善或达到预期目标，是否对其他表征无线通信网络性能的指标造成明显影响，进而从网络级性能和用户级性能两个方面评估网络调优策略的有效性、实时性和鲁棒性，同时评估各模块采用智能算法的综合能力并将对应的评估结果反馈给策略决策模块、策略生成模块、智能推理模块、以及深度解析模块，以供各模块视具体情况对自身采用的方法和输出结果进行调整优化，形成闭环反馈。
图4是本公开提供的无线通信网络知识图谱的智能分析与应用方法的业务流程示意图，如图4所示，该业务流程以超可靠低时延通信uRLLC场景下某个特定应用为例，该流程大致包括：
步骤一、采集数据。根据实施例需要，通过软硬数据采集等方式对综合试验网中通信数据进行采集、转换和清洗，按照不同的数据类型分类存储，并做格式转换。其中，本用例所需采集数据主要包括无线空口数据(如终端的信噪比、吞吐量等)和核心网数据(N1/N2/N3等)。
步骤二、构建知识图谱。通过无线通信网络协议规定，比如3GPP通信协议或者结合通信领域专家知识，确定其中规定的指标和数据字段等构成的无线通信网络内生因素，梳理所述无线通信网络规定的指标和数据字段，明确所属通信层、类型、可调性、参数值等属性，作为知识图谱的实体节点。明确实体及实体间的关系，构建三元组，即(头实体，关系，尾实体)，构建具有拓扑结构的无线通信网络知识图谱，如图5所示，展示了一种空口时延知识图谱局部示意图。其中，比如确定实体类型为通用非调型数据参数的实体包括：物理上行 共享信道(Physical Uplink Share Channel，PUSCH)误块率，上行(Uplink，UL)每秒总应答次数，上行每秒负应答次数，新空口(New Radio，NR)控制信道单元(Control Channel Elements，CCE)个数，内环链路自适应(Inner Loop Link Adaptation，ILLA)对应的实体类型为通用非调型数据参数。确定实体类型为可调型数据参数的实体包括：外环链路自适应偏置(Outer Loop Link Adaptation offset，OLLA offset)，上行非重传平均调制与编码策略(Modulation and Coding Scheme，MCS)，每秒上行二进制相移键控(Binary Phase Shift Keying，BPSK)调制占比，每秒上行正交相移键控(Quadrature Phase Shift Keying，QPSK)调制占比，每秒上行16正交幅度调制(16 Quadrature Amplitude Modulation，16QAM)调制占比，每秒上行64正交幅度调制(64 Quadrature Amplitude Modulation，64QAM)调制占比，每秒上行256正交幅度调制(256 Quadrature Amplitude Modulation，256QAM)调制占比。结合场景需要，本实施例无线通信网络关键指标为：空口及核心网各阶段网络时延、误码率。数据字段涉及空口和核心网相关数据。
步骤三：深度解析与图谱调优。对构建的无线通信网络知识图谱进行深度解析，一方面从图模型的角度利用理论分析算法对知识图谱结构(例如节点、边)及其内生因素关联关系(实体间关系)进行深入挖掘；另一方面，以采集存储的综合试验网实时数据为基础，利用数据驱动的方式对构建的无线通信网络知识图谱进行深度解析。从上述两个方面，深入分析时延与各节点之间的内生因素关联关系，如与信噪比、调制编码选择、网络切换等。基于深度解析结果，对无线通信网络知识图谱进行动态调优。
步骤四、检测与推理。在无线网络性能中的关键KPI指标发生波动或异常，如时延、误码率突然增大等问题时，依据构建的无线通信网络知识图谱及节点间关联关系(实体间关系)，基于图神经网络、近似推理、因果推断等算法进行回溯推理，定位出哪些层的指标或者数据字段问题。本实施例中，误码率高于10 -5，空口时延高于1ms即开始回溯推理。
步骤五、推荐调优。基于关联规则和深度学习的策略生成与决策算法，依据构建的无线通信网络知识图谱及各节点(实体)间内生因素关联关系，采用人工智能、拟生算法、统计学习等理论分析算法，建立网络参数调节模型，基于采集数据进行算法迭代优化，求解出调节参数，综合考虑调优策略的执行效率和效果，选择最优无线通信网络调优策略作为最终决策策略。
步骤六、将策略决策模块选择的网络调优决策策略下发给综合试验网，并按照策略对相应的可调参数进行调节，完成网络调优过程。针对uRLLC场景，构建无线通信网络智能评估模型，在执行完最终网络调优策略后，评估对应KPI指标是否得到改善或达到预期目标(例如空口时延是否降低，误码率是否减少等)，是否对其他表征无线通信网络性能的指标造成明显影响，并将评估结果反馈给深度解析、智能推理、策略生成、策略决策模块以供进一步调整优化。在本实施例中，下发的调优策略包括介质访问控制层(Media Access Control，MAC)和物理层参数，如调制与编码策略MCS、波束赋形矩阵等。
图6是本公开提供的一种电子设备的结构示意图；如图6所示，该电子设备，包括存储器620，收发机610和处理器600；其中，处理器600与存储器620也可以物理上分开布置。
存储器620，用于存储计算机程序；收发机610，用于在处理器600的控制下收发数据。
具体地，收发机610用于在处理器600的控制下接收和发送数据。
其中，在图6中，总线架构可以包括任意数量的互联的总线和桥，具体由处理器600代表的一个或多个处理器和存储器620代表的存储器的各种电路链接在一起。总线架构还可以将诸如外围设备、稳压器和功率管理电路等之类的各种其他电路链接在一起，这些都是本领域所公知的，因此，本申请不再对其进行进一步描述。总线接口提供接口。收发机610可以是多个元件，即包括发送机和接收机，提供用于在传输介质上与各种其他装置通信的单元，这些传输介质包括无线信道、有线信道、光缆等传输介质。
处理器600负责管理总线架构和通常的处理，存储器620可以存储处理器600在执行操作时所使用的数据。
处理器600可以是中央处理器(Central Processing Unit，CPU)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现场可编程门阵列(Field-Programmable Gate Array，FPGA)或复杂可编程逻辑器件(Complex Programmable Logic Device，CPLD)，处理器也可以采用多核架构。
处理器600通过调用存储器620存储的计算机程序，用于按照获得的可执行指令执行本申请实施例提供的任一所述方法，例如：
基于无线通信网络的内生因素以及图谱构建方法，构建无线通信网络知识图谱，并结合分类后的无线通信网络数据进行深度解析和图谱循环调优；
基于调优后的无线通信网络知识图谱，以及异常检测和智能推理算法，确定无线通信网络异常的诊断定位结果；
基于策略生成算法和所述诊断定位结果，确定若干调优策略，并以调优策略执行效率和效果最好为目标，结合智能决策算法，在若干调优策略中确定一种最优的调优策略，对所述无线通信网络进行调优；
基于调优后的无线通信网络，结合调优策略和智能算法的评估模型，确定所述调优策略对无线通信网络在网络级性能和用户级性能的改善度，以及所采用的各个智能算法的综合能力评估结果。
可选地，所述基于无线通信网络的内生因素以及图谱构建方法，构建无线通信网络知识图谱，并结合分类后的无线通信网络数据进行深度解析和图谱循环调优，包括：
基于图模型和数据模型的分析算法，确定无线通信网络知识图谱中各实体间新的关系，并分析各实体间关系的关联程度，更新所述实体间关系库，和/或确定无线通信网络知识图谱中新的实体、新的实体类型和新的实体属性中的一个或多个，更新所述实体库；
基于更新后的所述实体库和/或更新后的所述实体间关系库，更新所述无线通信网络知 识图谱；
所述实体库和所述实体间关系库是基于无线通信网络的内生因素确定的；所述无线通信网络的内生因素包括根据无线通信网络协议规定的指标和数据字段。
可选地，所述基于调优后的无线通信网络知识图谱，以及异常检测和智能推理算法，确定无线通信网络异常的诊断定位结果，包括：
基于所述调优后的无线通信网络知识图谱，以及异常检测算法，对无线通信网络进行异常检测，确定故障类型；
基于所述故障类型以及智能推理算法，确定所述故障类型所属的无线通信网络知识图谱中的实体。
可选地，所述基于策略生成算法和所述诊断定位结果，确定若干调优策略，并以调优策略执行效率和效果最好为目标，结合智能决策算法，在若干调优策略中确定一种最优的调优策略，对所述无线通信网络进行调优，包括：
基于故障类型所属的无线通信网络知识图谱中的实体，以及策略生成算法，确定若干无线通信网络调优策略；
基于智能决策算法，以所述调优策略的执行效率和效果最好为目标，确定一个最优的调优策略，对所述无线通信网络进行调优。
可选地，所述方法还包括：
基于调优后的无线通信网络，结合调优策略和智能算法的评估模型，确定所述调优策略对无线通信网络在网络级性能和用户级性能的改善度，以及所采用的智能算法的综合能力评估结果。
可选地，所述基于调优后的无线通信网络，结合调优策略和智能算法的评估模型，确定所述调优策略对无线通信网络在网络级性能和用户级性能的改善度，以及所采用的各个智能算法的综合能力评估结果，包括：
获取调优后的所述无线通信网络的数据；
确定在网络级性能评估和用户级性能评估两个维度所述调优策略对无线通信网络性能的改善度；
基于所述改善度以及所述各个智能算法的综合能力评估结果，确定需要调整的智能算法；所述智能算法包括：深度解析算法，异常检测算法，智能推理算法，策略生成算法和智能决策算法中的一个或多个。
可选地，所述基于无线通信网络的内生因素以及图谱构建方法，构建无线通信网络知识图谱，并结合分类后的无线通信网络数据进行深度解析和图谱循环调优之前，还包括：
获取无线通信网络原始数据，并进行预处理后，按照不同数据类型进行分类存储；所述不同数据类型的分类存储是基于分布式系统架构实现的。
可选地，所述不同数据类型包括：终端侧无线空口数据，基站侧无线空口数据，核心网 数据和网管数据中的一个或多个。
可选地，所述实体库中元素的实体类型包括网络级性能评估指标，用户级性能评估指标，通用非调型数据参数和可调型数据参数中的一个或多个；所述实体间关系库中实体关系包括因果关系，隐式关系和显式关系中的一个或多个。
可选地，所述基于图模型和数据模型的分析算法，确定无线通信网络知识图谱中各实体间新的关系，并分析各实体间关系的关联程度，更新所述实体间关系库，和/或确定无线通信网络知识图谱中新的实体、新的实体类型和新的实体属性中的一个或多个，更新所述实体库，包括：
基于图模型算法，确定无线通信网络知识图谱中各实体间关系的关联程度的量化度量，更新所述实体间关系库；
基于获取的分类后的无线通信网络数据，确定无线通信网络知识图谱中新的实体、新的实体类型、新的实体属性和各实体间新的关系中的一个或多个，更新所述实体库和/或所述实体间关系库。
可选地，所述基于更新后的所述实体库和/或更新后的所述实体间关系库，更新所述无线通信网络知识图谱，包括：
基于无线通信网络协议规定和理论分析算法，以及所述更新后的所述实体库和/或所述实体间关系库，确定需要调优的参数，所述需要调优的参数包括：无线通信网络知识图谱中实体、实体属性的参数和实体间关系的参数中的一个或多个；
基于所述需要调优的参数，更新所述无线通信网络知识图谱。
可选地，所述故障类型包括：实体库中不同元素的故障和实体间关系库中不同元素的故障。
可选地，所述基于所述故障类型以及智能推理算法，确定所述故障类型所属的无线通信网络知识图谱中的实体，包括：
确定所述故障类型所属的实体，所述实体是实体库中特定的元素；
基于所述实体的属性，确定所述实体所在位置、可调性以及影响因素，所述所在位置包括所述实体位于的具体通信层或所处图谱结构的位置，所述影响因素是所述实体和与其存在实体间关系的其他实体对所述实体的影响程度。
可选地，所述智能决策算法包括深度强化学习DRL、马尔可夫决策过程、拟生算法和统计学习算法中的一个或多个。
可选地，所述基于所述改善度以及所述各个智能算法的综合能力评估结果，确定需要调整的智能算法，包括：
若所述最优的调优策略执行后，网络级性能和/或用户级性能的改善度未改变或者降低的情况下，确定各个智能算法的综合能力评估结果；
基于各个智能算法的综合能力评估结果，结合相应的评估准则，确定需要调整的特定智 能算法并更新。
在此需要说明的是，本申请实施例提供的上述电子设备，能够实现上述方法实施例所实现的所有方法步骤，且能够达到相同的技术效果，在此不再对本实施例中与方法实施例相同的部分及有益效果进行具体赘述。
另一方面，本公开还提供一种计算机程序产品，所述计算机程序产品包括存储在非暂态计算机可读存储介质上的计算机程序，所述计算机程序包括程序指令，当所述程序指令被计算机执行时，计算机能够执行上述各实施例所提供的无线通信网络知识图谱的智能分析与应用方法的步骤，例如包括：
基于无线通信网络的内生因素以及图谱构建方法，构建无线通信网络知识图谱，并结合分类后的无线通信网络数据进行深度解析和图谱循环调优；
基于调优后的无线通信网络知识图谱，以及异常检测和智能推理算法，确定无线通信网络异常的诊断定位结果；
基于策略生成算法和所述诊断定位结果，确定若干调优策略，并以调优策略执行效率和效果最好为目标，结合智能决策算法，在若干调优策略中确定一种最优的调优策略，对所述无线通信网络进行调优。
可选地，所述步骤还包括：
基于调优后的无线通信网络，结合调优策略和智能算法的评估模型，确定所述调优策略对无线通信网络在网络级性能和用户级性能的改善度，以及所采用的智能算法的综合能力评估结果。
可选地，所述基于无线通信网络的内生因素以及图谱构建方法，构建无线通信网络知识图谱，并结合分类后的无线通信网络数据进行深度解析和图谱循环调优之前，还包括：
获取无线通信网络原始数据，并进行预处理后，按照不同数据类型进行分类存储；所述不同数据类型的分类存储是基于分布式系统架构实现的。
另一方面，本申请实施例还提供一种处理器可读存储介质，所述处理器可读存储介质存储有计算机程序，所述计算机程序用于使所述处理器执行上述各实施例提供的无线通信网络知识图谱的智能分析与应用方法的步骤，例如包括：
基于无线通信网络的内生因素以及图谱构建方法，构建无线通信网络知识图谱，并结合分类后的无线通信网络数据进行深度解析和图谱循环调优；
基于调优后的无线通信网络知识图谱，以及异常检测和智能推理算法，确定无线通信网络异常的诊断定位结果；
基于策略生成算法和所述诊断定位结果，确定若干调优策略，并以调优策略执行效率和效果最好为目标，结合智能决策算法，在若干调优策略中确定一种最优的调优策略，对所述无线通信网络进行调优。
可选地，所述步骤还包括：
基于调优后的无线通信网络，结合调优策略和智能算法的评估模型，确定所述调优策略对无线通信网络在网络级性能和用户级性能的改善度，以及所采用的智能算法的综合能力评估结果。
可选地，所述基于无线通信网络的内生因素以及图谱构建方法，构建无线通信网络知识图谱，并结合分类后的无线通信网络数据进行深度解析和图谱循环调优之前，还包括：
获取无线通信网络原始数据，并进行预处理后，按照不同数据类型进行分类存储；所述不同数据类型的分类存储是基于分布式系统架构实现的。
所述处理器可读存储介质可以是处理器能够存取的任何可用介质或数据存储设备，包括但不限于磁性存储器(例如软盘、硬盘、磁带、磁光盘(MO)等)、光学存储器(例如CD、DVD、BD、HVD等)、以及半导体存储器(例如ROM、EPROM、EEPROM、非易失性存储器(NAND FLASH)、固态硬盘(SSD))等。
以上所描述的装置实施例仅仅是示意性的，其中所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。本领域普通技术人员在不付出创造性的劳动的情况下，即可以理解并实施。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到各实施方式可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件。基于这样的理解，上述技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品可以存储在计算机可读存储介质中，如ROM/RAM、磁碟、光盘等，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行各个实施例或者实施例的某些部分所述的方法。
最后应说明的是：以上实施例仅用以说明本公开的技术方案，而非对其限制；尽管参照前述实施例对本公开进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本公开各实施例技术方案的精神和范围。

Claims (32)

  1. 一种无线通信网络知识图谱的智能分析与应用系统，其特征在于，所述系统至少包括：
    知识图谱单元、智能溯源单元和调优策略单元；
    所述知识图谱单元，用于基于无线通信网络的内生因素以及图谱构建方法，构建无线通信网络知识图谱，并结合分类后的无线通信网络数据进行深度解析和图谱循环调优；
    所述智能溯源单元，用于基于调优后的无线通信网络知识图谱，以及异常检测和智能推理算法，确定无线通信网络异常的诊断定位结果；
    所述调优策略单元，用于基于策略生成算法和所述诊断定位结果，确定若干调优策略，并以调优策略执行效率和效果最好为目标，结合智能决策算法，在若干调优策略中确定一种最优的调优策略，对所述无线通信网络进行调优。
  2. 根据权利要求1所述的无线通信网络知识图谱的智能分析与应用系统，其特征在于，所述知识图谱单元至少包括深度解析模块和图谱调优模块；
    其中，所述深度解析模块，用于基于图模型和数据模型的分析算法，确定无线通信网络知识图谱中各实体间新的关系，并分析各实体间关系的关联程度，更新实体间关系库，和/或确定无线通信网络知识图谱中新的实体、新的实体类型和新的实体属性中的一个或者多个，更新实体库；
    所述图谱调优模块，用于基于更新后的所述实体库和/或更新后的所述实体间关系库，更新所述无线通信网络知识图谱；
    所述实体库和所述实体间关系库是基于无线通信网络的内生因素确定的；所述无线通信网络的内生因素包括根据无线通信网络协议规定的指标和数据字段。
  3. 根据权利要求1所述的无线通信网络知识图谱的智能分析与应用系统，其特征在于，所述智能溯源单元包括异常检测模块和智能推理模块；
    其中，所述异常检测模块，用于基于所述调优后的无线通信网络知识图谱，以及异常检测算法，对无线通信网络进行异常检测，确定故障类型；
    智能推理模块，用于基于所述故障类型以及智能推理算法，确定所述故障类型所属的无线通信网络知识图谱中的实体。
  4. 根据权利要求1所述的无线通信网络知识图谱的智能分析与应用系统，其特征在于，所述调优策略单元包括策略生成模块和策略决策模块；
    其中，所述策略生成模块，用于基于故障类型所属的无线通信网络知识图谱中的实体，以及策略生成算法，确定若干无线通信网络调优策略；
    所述策略决策模块，用于基于智能决策算法，以所述调优策略的执行效率和效果最好为目标，确定一个最优的调优策略，对所述无线通信网络进行调优。
  5. 根据权利要求1所述的无线通信网络知识图谱的智能分析与应用系统，其特征在于， 所述系统还包括性能评估单元；
    所述性能评估单元，用于基于调优后的无线通信网络，结合调优策略和智能算法的评估模型，确定所述调优策略对无线通信网络在网络级性能和用户级性能的改善度，以及每个单元所采用的智能算法的综合能力评估结果，并反馈给对应单元。
  6. 根据权利要求5所述的无线通信网络知识图谱的智能分析与应用系统，其特征在于，所述性能评估单元包括网络级性能评估模块和用户级性能评估模块；
    其中，所述网络级性能评估模块和用户级性能评估模块用于获取调优后的所述无线通信网络的数据；
    确定在网络级性能评估和用户级性能评估两个维度所述调优策略对无线通信网络性能的改善度；
    基于所述改善度以及所述智能算法的综合能力评估结果，确定需要调整的智能算法；所述智能算法包括：深度解析算法，异常检测算法，智能推理算法，策略生成算法和智能决策算法中的一个或多个。
  7. 根据权利要求1或5所述的无线通信网络知识图谱的智能分析与应用系统，其特征在于，所述系统还包括无线数据单元；所述无线数据单元用于获取无线通信网络原始数据，并进行预处理后，按照不同数据类型进行分类存储；所述不同数据类型的分类存储，是基于分布式系统架构实现的。
  8. 根据权利要求7所述的无线通信网络知识图谱的智能分析与应用系统，其特征在于，所述不同数据类型包括：终端侧无线空口数据，基站侧无线空口数据，核心网数据和网管数据中的一个或者多个。
  9. 根据权利要求2所述的无线通信网络知识图谱的智能分析与应用系统，其特征在于，所述实体库中元素的实体类型包括网络级性能评估指标，用户级性能评估指标，通用非调型数据参数和可调型数据参数中的一个或多个；所述实体间关系库中实体关系包括因果关系，隐式关系和显式关系中的一个或多个。
  10. 根据权利要求2所述的无线通信网络知识图谱的智能分析与应用系统，其特征在于，所述基于图模型和数据模型的分析算法，确定无线通信网络知识图谱中各实体间新的关系，并分析各实体间关系的关联程度，更新实体间关系库，和/或确定无线通信网络知识图谱中新的实体、新的实体类型和新的实体属性中的一个或者多个，更新实体库，包括：
    基于图模型算法，确定无线通信网络知识图谱中各实体间关系的关联程度的量化度量，更新所述实体间关系库；
    基于获取的分类后的无线通信网络数据，确定无线通信网络知识图谱中新的实体、新的实体类型、新的实体属性和各实体间新的关系中的一个或多个，更新所述实体库和/或所述实体间关系库。
  11. 根据权利要求2所述的无线通信网络知识图谱的智能分析与应用系统，其特征在 于，所述基于更新后的所述实体库和/或更新后的所述实体间关系库，更新所述无线通信网络知识图谱，包括：
    基于无线通信网络协议规定和理论分析算法，以及所述更新后的所述实体库和/或更新后的所述实体间关系库，确定需要调优的参数，所述需要调优的参数包括：无线通信网络知识图谱中实体、实体属性的参数和实体间关系的参数中的一个或者多个；
    基于所述需要调优的参数，更新所述无线通信网络知识图谱。
  12. 根据权利要求3或4所述的无线通信网络知识图谱的智能分析与应用系统，其特征在于，所述故障类型包括：实体库中不同元素的故障和实体间关系库中不同元素的故障。
  13. 根据权利要求3所述的无线通信网络知识图谱的智能分析与应用系统，其特征在于，所述基于所述故障类型以及智能推理算法，确定所述故障类型所属的无线通信网络知识图谱中的实体，包括：
    确定所述故障类型所属的实体，所述实体是实体库中特定的元素；
    基于所述实体的属性，确定所述实体所在位置、可调性以及影响因素，所述所在位置包括所述实体位于的具体通信层或所处图谱结构的位置，所述影响因素是所述实体和与其存在实体间关系的其他实体对所述实体的影响程度。
  14. 根据权利要求1或4或6所述的无线通信网络知识图谱的智能分析与应用系统，其特征在于，所述智能决策算法包括深度强化学习DRL、马尔可夫决策过程、拟生算法和统计学习算法中的一个或多个。
  15. 根据权利要求6所述的无线通信网络知识图谱的智能分析与应用系统，其特征在于，所述基于所述改善度以及所述智能算法的综合能力评估结果，确定需要调整的智能算法，包括：
    若所述最优的调优策略执行后，网络级性能和/或用户级性能的改善度未改变或降低的情况下，确定各单元采用的智能算法的综合能力评估结果；
    基于各单元采用智能算法的综合能力评估结果，结合相应的评估准则，确定需要调整的智能算法并更新。
  16. 一种无线通信网络知识图谱的智能分析与应用方法，其特征在于，包括：
    基于无线通信网络的内生因素以及图谱构建方法，构建无线通信网络知识图谱，并结合分类后的无线通信网络数据进行深度解析和图谱循环调优；
    基于调优后的无线通信网络知识图谱，以及异常检测和智能推理算法，确定无线通信网络异常的诊断定位结果；
    基于策略生成算法和所述诊断定位结果，确定若干调优策略，并以调优策略执行效率和效果最好为目标，结合智能决策算法，在若干调优策略中确定一种最优的调优策略，对所述无线通信网络进行调优。
  17. 根据权利要求16所述的无线通信网络知识图谱的智能分析与应用方法，其特征在 于，所述基于无线通信网络的内生因素以及图谱构建方法，构建无线通信网络知识图谱，并结合分类后的无线通信网络数据进行深度解析和图谱循环调优，包括：
    基于图模型和数据模型的分析算法，确定无线通信网络知识图谱中各实体间新的关系，并分析各实体间关系的关联程度，更新实体间关系库，和/或确定无线通信网络知识图谱中新的实体、新的实体类型和新的实体属性中的一个或者多个，更新实体库；
    基于更新后的所述实体库和/或更新后的所述实体间关系库，更新所述无线通信网络知识图谱；
    所述实体库和所述实体间关系库是基于无线通信网络的内生因素确定的；所述无线通信网络的内生因素包括根据无线通信网络协议规定的指标和数据字段。
  18. 根据权利要求16所述的无线通信网络知识图谱的智能分析与应用方法，其特征在于，所述基于调优后的无线通信网络知识图谱，以及异常检测和智能推理算法，确定无线通信网络异常的诊断定位结果，包括：
    基于所述调优后的无线通信网络知识图谱，以及异常检测算法，对无线通信网络进行异常检测，确定故障类型；
    基于所述故障类型以及智能推理算法，确定所述故障类型所属的无线通信网络知识图谱中的实体。
  19. 根据权利要求16所述的无线通信网络知识图谱的智能分析与应用方法，其特征在于，所述基于策略生成算法和所述诊断定位结果，确定若干调优策略，并以调优策略执行效率和效果最好为目标，结合智能决策算法，在若干调优策略中确定一种最优的调优策略，对所述无线通信网络进行调优，包括：
    基于故障类型所属的无线通信网络知识图谱中的实体，以及策略生成算法，确定若干无线通信网络调优策略；
    基于智能决策算法，以所述调优策略的执行效率和效果最好为目标，确定一个最优的调优策略，对所述无线通信网络进行调优。
  20. 根据权利要求16所述的无线通信网络知识图谱的智能分析与应用方法，其特征在于，所述方法还包括：
    基于调优后的无线通信网络，结合调优策略和智能算法的评估模型，确定所述调优策略对无线通信网络在网络级性能和用户级性能的改善度，以及所采用的各个智能算法的综合能力评估结果。
  21. 根据权利要求20所述的无线通信网络知识图谱的智能分析与应用方法，其特征在于，所述基于调优后的无线通信网络，结合调优策略和智能算法的评估模型，确定所述调优策略对无线通信网络在网络级性能和用户级性能的改善度，以及所采用的各个智能算法的综合能力评估结果，包括：
    获取调优后的所述无线通信网络的数据；
    确定在网络级性能评估和用户级性能评估两个维度所述调优策略对无线通信网络性能的改善度；
    基于所述改善度以及所述各个智能算法的综合能力评估结果，确定需要调整的智能算法；所述智能算法包括：深度解析算法，异常检测算法，智能推理算法，策略生成算法和智能决策算法中的一个或多个。
  22. 根据权利要求16或20所述的无线通信网络知识图谱的智能分析与应用方法，其特征在于，所述基于无线通信网络的内生因素以及图谱构建方法，构建无线通信网络知识图谱，并结合分类后的无线通信网络数据进行深度解析和图谱循环调优之前，还包括：
    获取无线通信网络原始数据，并进行预处理后，按照不同数据类型进行分类存储；所述不同数据类型的分类存储是基于分布式系统架构实现的。
  23. 根据权利要求22所述的无线通信网络知识图谱的智能分析与应用方法，其特征在于，所述不同数据类型包括：终端侧无线空口数据，基站侧无线空口数据，核心网数据和网管数据中的一个或多个。
  24. 根据权利要求17所述的无线通信网络知识图谱的智能分析与应用方法，其特征在于，所述实体库中元素的实体类型包括网络级性能评估指标，用户级性能评估指标，通用非调型数据参数和可调型数据参数中的一个或多个；所述实体间关系库中实体关系包括因果关系，隐式关系和显式关系中的一个或多个。
  25. 根据权利要求17所述的无线通信网络知识图谱的智能分析与应用方法，其特征在于，所述基于图模型和数据模型的分析算法，确定无线通信网络知识图谱中各实体间新的关系，并分析各实体间关系的关联程度，更新所述实体间关系库，和/或确定无线通信网络知识图谱中新的实体、新的实体类型和新的实体属性中的一个或多个，更新所述实体库，包括：
    基于图模型算法，确定无线通信网络知识图谱中各实体间关系的关联程度的量化度量，更新所述实体间关系库；
    基于获取的分类后的无线通信网络数据，确定无线通信网络知识图谱中新的实体、新的实体类型、新的实体属性和各实体间新的关系中的一个或多个，更新所述实体库和/或所述实体间关系库。
  26. 根据权利要求17所述的无线通信网络知识图谱的智能分析与应用方法，其特征在于，所述基于更新后的所述实体库和/或更新后的所述实体间关系库，更新所述无线通信网络知识图谱，包括：
    基于无线通信网络协议规定和理论分析算法，以及所述更新后的所述实体库和/或所述实体间关系库，确定需要调优的参数，所述需要调优的参数包括：无线通信网络知识图谱中实体、实体属性的参数和实体间关系的参数中的一个或多个；
    基于所述需要调优的参数，更新所述无线通信网络知识图谱。
  27. 根据权利要求18或19所述的无线通信网络知识图谱的智能分析与应用方法，其 特征在于，所述故障类型包括：实体库中不同元素的故障和实体间关系库中不同元素的故障。
  28. 根据权利要求18所述的无线通信网络知识图谱的智能分析与应用方法，其特征在于，所述基于所述故障类型以及智能推理算法，确定所述故障类型所属的无线通信网络知识图谱中的实体，包括：
    确定所述故障类型所属的实体，所述实体是实体库中特定的元素；
    基于所述实体的属性，确定所述实体所在位置、可调性以及影响因素，所述所在位置包括所述实体位于的具体通信层或所处图谱结构的位置，所述影响因素是所述实体和与其存在实体间关系的其他实体对所述实体的影响程度。
  29. 根据权利要求16或19或21所述的无线通信网络知识图谱的智能分析与应用方法，其特征在于，所述智能决策算法包括深度强化学习DRL、马尔可夫决策过程、拟生算法和统计学习算法中的一个或多个。
  30. 根据权利要求21所述的无线通信网络知识图谱的智能分析与应用方法，其特征在于，所述基于所述改善度以及所述各个智能算法的综合能力评估结果，确定需要的调整的智能算法，包括：
    若所述最优的调优策略执行后，网络级性能和/或用户级性能的改善度未改变或者降低的情况下，确定各个智能算法的综合能力评估结果；
    基于各个智能算法的综合能力评估结果，结合相应的评估准则，确定需要调整的特定智能算法并更新。
  31. 一种电子设备，包括存储器，收发机，处理器；
    存储器，用于存储计算机程序；收发机，用于在所述处理器的控制下收发数据；处理器，用于执行所述存储器中的计算机程序并实现如权利要求16至30任一项所述的无线通信网络知识图谱的智能分析与应用方法的步骤。
  32. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质存储有计算机程序，所述计算机程序用于使计算机执行权利要求16至30任一项所述的无线通信网络知识图谱的智能分析与应用方法的步骤。
PCT/CN2022/128902 2022-02-17 2022-11-01 无线通信网络知识图谱的智能分析与应用系统及方法 WO2023155481A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210144211.7A CN114205852B (zh) 2022-02-17 2022-02-17 无线通信网络知识图谱的智能分析与应用系统及方法
CN202210144211.7 2022-02-17

Publications (1)

Publication Number Publication Date
WO2023155481A1 true WO2023155481A1 (zh) 2023-08-24

Family

ID=80645570

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/128902 WO2023155481A1 (zh) 2022-02-17 2022-11-01 无线通信网络知识图谱的智能分析与应用系统及方法

Country Status (2)

Country Link
CN (1) CN114205852B (zh)
WO (1) WO2023155481A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117421642A (zh) * 2023-12-18 2024-01-19 电子科技大学 基于深度学习的智能检波器数据存储方法及相关设备

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114205852B (zh) * 2022-02-17 2022-05-03 网络通信与安全紫金山实验室 无线通信网络知识图谱的智能分析与应用系统及方法

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111191851A (zh) * 2020-01-03 2020-05-22 中国科学院信息工程研究所 一种基于知识图谱的数据中心能效优化方法
CN112335278A (zh) * 2018-06-22 2021-02-05 上海诺基亚贝尔股份有限公司 用于网络优化的方法、装置和计算机可读介质
CN112714032A (zh) * 2021-03-29 2021-04-27 网络通信与安全紫金山实验室 无线网络协议知识图谱构建分析方法、系统、设备及介质
CN113259972A (zh) * 2021-06-08 2021-08-13 网络通信与安全紫金山实验室 基于无线通信网络数据仓库构建方法、系统、设备及介质
CN113453257A (zh) * 2021-08-30 2021-09-28 网络通信与安全紫金山实验室 无线通信网络性能优化方法及装置
CN114205852A (zh) * 2022-02-17 2022-03-18 网络通信与安全紫金山实验室 无线通信网络知识图谱的智能分析与应用架构及方法

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109978052B (zh) * 2019-03-25 2020-12-11 北京快电科技有限公司 一种用户侧能源设备智慧检修方法
CN111010311B (zh) * 2019-11-25 2022-07-08 江苏艾佳家居用品有限公司 一种基于知识图谱的智能网络故障诊断方法
CN112446144A (zh) * 2020-11-17 2021-03-05 哈工大机器人（合肥）国际创新研究院 一种大型回转机组故障诊断方法及装置

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112335278A (zh) * 2018-06-22 2021-02-05 上海诺基亚贝尔股份有限公司 用于网络优化的方法、装置和计算机可读介质
CN111191851A (zh) * 2020-01-03 2020-05-22 中国科学院信息工程研究所 一种基于知识图谱的数据中心能效优化方法
CN112714032A (zh) * 2021-03-29 2021-04-27 网络通信与安全紫金山实验室 无线网络协议知识图谱构建分析方法、系统、设备及介质
CN113259972A (zh) * 2021-06-08 2021-08-13 网络通信与安全紫金山实验室 基于无线通信网络数据仓库构建方法、系统、设备及介质
CN113453257A (zh) * 2021-08-30 2021-09-28 网络通信与安全紫金山实验室 无线通信网络性能优化方法及装置
CN114205852A (zh) * 2022-02-17 2022-03-18 网络通信与安全紫金山实验室 无线通信网络知识图谱的智能分析与应用架构及方法

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117421642A (zh) * 2023-12-18 2024-01-19 电子科技大学 基于深度学习的智能检波器数据存储方法及相关设备
CN117421642B (zh) * 2023-12-18 2024-03-12 电子科技大学 基于深度学习的智能检波器数据存储方法及相关设备

Also Published As

Publication number Publication date
CN114205852A (zh) 2022-03-18
CN114205852B (zh) 2022-05-03

Similar Documents

Publication Publication Date Title
WO2023155481A1 (zh) 无线通信网络知识图谱的智能分析与应用系统及方法
Zhang et al. New approach of multi-path reliable transmission for marginal wireless sensor network
Dong et al. Dynamic packet length control in wireless sensor networks
CN103596191B (zh) 一种无线传感器网络智能配置系统及其方法
Medeiros et al. A survey on data analysis on large-Scale wireless networks: online stream processing, trends, and challenges
Sun et al. Accelerating convergence of federated learning in MEC with dynamic community
US20230281513A1 (en) Data model training method and apparatus
WO2021063254A1 (zh) 一种信息处理方法、相关设备及计算机存储介质
Hu et al. Clustered data sharing for Non-IID federated learning over wireless networks
Kontos et al. A topology inference algorithm for wireless sensor networks
US12015531B2 (en) Method for generating a quality of experience (QoE) index by way of ensemble of expectation scores
WO2023041992A1 (en) Systems and methods for performing root cause analysis
Muñoz et al. A method for identifying faulty cells using a classification tree-based UE diagnosis in LTE
CN114698016A (zh) 一种质差小区网络故障检测方法及装置
Lu et al. Maximizing multicast lifetime in unreliable wireless ad hoc network
He et al. Age optimal sampling for unreliable channels under unknown channel statistics
Li et al. The variable-weight MADM algorithm for wireless network
Reali et al. Probabilistic codebook-based fault localization in data networks
Li et al. Learning to Transmit with Provable Guarantees in Wireless Federated Learning
Tay Decentralized detection in resource-limited sensor network architectures
US12052735B1 (en) Method for offloading decision and resource allocation based on integration of communication, sensing and computing
US20230319662A1 (en) Method and apparatus for programmable and customized intelligence for traffic steering in 5g networks using open ran architectures
Zhu et al. Edge-Assisted Video Transmission with Adaptive Key Frame Selection: A Hierarchical DRL Approach
Wang et al. Research on Fault Identification Method of Power System Communication Network Based on Deep Learning
EP4258730A1 (en) Method and apparatus for programmable and customized intelligence for traffic steering in 5g networks using open ran architectures

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22926785

Country of ref document: EP

Kind code of ref document: A1