Optimization Method for Wireless Communication Networks Based on Knowledge Graphs and Mutual Information

Abstract

The present invention discloses a wireless communication network optimization method based on knowledge graph and mutual information, which mainly solves the problem of low accuracy in the prior art of screening factors affecting efficiency of key performance indicators (KPIs). Its implementation scheme is: 1) obtaining user measurement data in the wireless communication network and processing it to construct initial data; 2) building a wireless communication network knowledge graph around the target KPI based on the initial data; 3) calculating the edge connection strength of the cause node U and the result node V in the knowledge graph; 4) normalizing the edge connection strength to obtain the edge weight, and using the edge weight to calculate the influence efficiency between nodes; 5) sorting the influence efficiency between nodes from large to small for screening; 6) obtaining the screening results and optimizing them. The present invention improves the screening accuracy of factors that have a greater impact on the efficiency of key performance indicators (KPIs), improves the optimization effect of wireless communication networks, and can be used in wireless communication networks.

Description

Technical Field

[0001] The present invention belongs to the field of communication technology, and in particular relates to a wireless communication network optimization method, which can be used in a wireless communication network. Background Art

[0002] With the development of wireless communication networks, wireless communication network data has become diversified. The factors that affect the performance of wireless communication networks are very complex. Traditional methods are difficult to clarify the correlation between the factors. Due to the complex correlation between the factors, it is difficult to find the factors that have a greater impact on the efficiency of key performance indicators, which brings difficulties to the optimization of wireless communication networks.

[0003] As a new method for network data knowledge representation and a new tool for knowledge management, knowledge graphs can clarify the complex relationships between data and characterize the types and attributes of data and relationships between data. Therefore, knowledge graphs can be used to characterize wireless communication network data and combined with mutual information to determine the factors that have a greater impact on key performance indicators, thereby facilitating the optimization of wireless communication network performance.

[0004] The Purple Mountain Laboratory of Network Communication and Security has disclosed a "wireless communication network performance optimization method and device" in its patent document with application number CN202111000514.3. It determines the target node that affects the target indicator based on the wireless communication network performance knowledge graph, and optimizes the target indicator based on the target node. The wireless communication network performance knowledge graph is constructed using data fields and indicators related to the performance of the wireless communication network, and the edge connection strength is used to characterize the degree of association between the connected nodes. The edge connection strength between node u and node v is the sum of the probabilities that any state of the cause node u leads to all states of the result node v. The edge connection strength is normalized and used as the edge weight of the edge. The influence efficiency between nodes and the importance of nodes are calculated in combination with the edge weight and the structure diagram. This method has a low screening accuracy for factors that have a greater impact on the efficiency of key performance indicators KPIs, which limits the effect of optimizing the performance of wireless communication networks. Summary of the invention

[0005] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art and propose a wireless communication network optimization method based on knowledge graph and mutual information to improve the screening accuracy of factors that have a greater impact on the efficiency of key performance indicators (KPIs) and improve the optimization effect of wireless communication networks.

[0006] To achieve the above object, the technical solution of the present invention comprises the following steps:

[0007] (1) Obtain user measurement data in n cells, and manually remove some data fields that are obviously irrelevant to the key performance indicators (KPIs) and data fields with empty values ​​to form an initial data set;

[0008] (2) Classify the data fields in the initial data set into entities, classify the associations between the fields, and connect the entities with associations in a directed manner to form a wireless communication network knowledge graph with the target KPI as the core;

[0009] (3) Calculate the edge connection strength I(U; V) between node U and node V in the knowledge graph:

[0010]

[0011] Among them, p(u,v) is the joint probability distribution function of the values ​​of node U and node V in the knowledge graph, and p(u) and p(v) are the marginal probability distribution functions of the values ​​of node U and node V respectively;

[0012] (4) Calculate the impact efficiency between nodes:

[0013] Normalize the edge connection strength I(U; V) to get the edge weight

[0014] For any two nodes X and Y that have a connected path, calculate the product of the edge weights on their qth shortest path to obtain the influence efficiency e between the two nodes XY ;

[0015] (5) The calculated influence efficiency of each node on the target KPI node is sorted from large to small, and the first t data fields selected are adjusted to optimize the target KPI. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is an implementation flow chart of the present invention. DETAILED DESCRIPTION

[0017] The embodiments and effects of the present invention are further described in detail below with reference to the accompanying drawings:

[0018] Reference Figure 1 , the implementation steps of this example are as follows:

[0019] Step 1: Build initial data.

[0020] 1.1) Obtain user measurement data of three cells in a real scenario, which contains a total of 104 data fields and 10,000 measurement data, where the data fields include the user experience rate UserThp, the amount of data sent downlink by the radio link management layer DLThpVol, the downlink radio link management throughput DLThpVolLastSlot transmitted in the last time slot that makes the buffer empty, and the data transmission duration DLThpTimeRmvLastSlot after deducting the last time slot that makes the downlink buffer empty;

[0021] 1.2) Assume that the target key performance indicator KPI in the data field is the user experience rate UserThp, which is calculated using the three data fields DLThpVol, DLThpVolLastSlot, and DLThpTimeRmvLastSlot:

[0022]

[0023] 1.3) By manually removing some data fields that are obviously irrelevant to the user experience rate, such as time and verification flag, and then deleting the data with a user experience rate of 0 from the remaining 74 data fields, 4800 measurement data are left.

[0024] Step 2: Build a wireless communication network knowledge graph with the target KPI as the core.

[0025] This step takes the user experience rate as an example to build a wireless communication network knowledge graph. The specific implementation is as follows:

[0026] 2.1) Determine the entity type:

[0027] Based on the meaning of each data field in the wireless communication network and the communication network protocol, the data fields in the initial data set are divided into four categories: network-level performance evaluation indicators, user-level performance evaluation indicators, general non-adjustable data parameters, and adjustable data parameters. Among them:

[0028] Network-level performance evaluation indicators are used to evaluate the performance level of the entire wireless communication network, such as the signal-to-noise ratio of the uplink sounding reference signal UlSrsSinr, the total number of downlink physical resource blocks used DLPrbUsedNum, and the total number of physical resource blocks used for downlink data resource bearers DLDRBPrbUsedNum;

[0029] User-level performance evaluation indicators are used to evaluate the user's communication level, such as the RANK value reported by the user equipment for the last time at the reporting time, the maximum uplink transmission power PcMaxVal of the user equipment, and the power margin UePHR of the user equipment;

[0030] General non-adjustable data parameters refer to fixed parameter values ​​in the communication network, such as beam ID values, cell IDs, and user equipment IDs reported by user equipment;

[0031] Adjustable data parameters refer to parameters that can change network or user performance through adjustment, such as the distance TA between the user and the base station and the modulation mode UePuschModType of the physical uplink shared channel of the user equipment.

[0032] 2.2) Determine the relationship type:

[0033] The relationships between data fields are divided into three categories: causal relationships, implicit relationships, and explicit relationships, among which:

[0034] Causal relationship refers to the direct impact of one entity on another entity, such as the causal relationship between reference signal reception quality and signal-to-noise ratio;

[0035] Implicit relationship refers to the indirect influence of one entity on another entity, that is, there is no clear and specific expression to express the relationship between the two entities. For example, the relationship between the number of switch-out executions and the switch-out success rate is implicit.

[0036] An explicit relationship refers to a relationship where a specific analytical expression can be obtained after reasoning and analysis between two entities. For example, the amount of data sent downlink at the radio link management layer and the user experience rate are an explicit relationship.

[0037] 2.3) Construct triples:

[0038] After determining the entity type to which the data field belongs and the relationship type to which the association relationship between the data fields belongs, connect the head entity h and the tail entity t through the line segment r to construct a triple (h, r, t);

[0039] 2.4) Build a knowledge graph:

[0040] The head entity and tail entity in each triple are regarded as a node in the wireless communication network knowledge graph, and the association relationship between the head entity and the tail entity is regarded as the edge in the graph to complete the construction of the wireless communication network knowledge graph.

[0041] Step 3: Calculate the edge connection strength I(U; V) between node U and node V in the knowledge graph.

[0042] 3.1) Calculate the joint probability distribution function p(u,v) of the values ​​of nodes U and V in the knowledge graph:

[0043] p(u,v)=P{U=u,V=v};

[0044] 3.2) Calculate the marginal probability distribution function p(u) of the value of node U and the marginal probability distribution function p(v) of the value of node V in the knowledge graph:

[0045]

[0046]

[0047] 3.3) According to the results of 3.1) and 3.2), calculate the edge connection strength I(U; V) between node U and node V in the knowledge graph:

[0048]

[0049] Step 4: Calculate the impact efficiency between nodes.

[0050] 4.1) Normalize the edge connection strength I(U; V) to get the edge weight

[0051]

[0052] Among them, U is the cause node, V is the result node, for any (U,V)∈G, (U,V) indicates that there is a directed edge between node U and node V, G represents all nodes in the knowledge graph of wireless communication networks, for any result node V, there are m cause nodes U, represents the edge weight of the mth cause node;

[0053] 4.2) For any two nodes X and Y that have a connected path, calculate the product of the edge weights on their qth shortest path to obtain the influence efficiency e between the two nodes XY :

[0054]

[0055] in, It represents the product of the edge weights of all edges on the qth shortest path from node X to node Y. The qth shortest path here is a reference, which means that if there are y shortest paths, the shortest path with the largest product of the edge weights of all edges on the path is selected as the qth shortest path.

[0056] Step 5: Get the screening results and optimize them.

[0057] 5.1) Taking the user experience rate UserThp as node Y, calculate the influence efficiency of any node on the user experience rate;

[0058] 5.2) Sort the impact efficiency of each node on the user experience rate from large to small, filter out the first 40 data fields, and select the adjustable data parameters of the data fields from the 40 data fields;

[0059] 5.3) Adjust the adjustable data parameters according to the parameter value range specified by the wireless communication network protocol, so as to optimize the user experience rate.

[0060] The above description is only a specific example of the present invention and does not constitute any limitation to the present invention. It is obvious that for professionals in this field, after understanding the content and principles of the present invention, they may make various modifications and changes in form and details without departing from the principles and structures of the present invention. However, these modifications and changes based on the ideas of the present invention are still within the scope of protection of the claims of the present invention.

Claims

1. A wireless communication network optimization method based on knowledge graph and mutual information, characterized in that: The steps include: (1) Obtain user measurement data in n cells, and manually remove some data fields that are obviously irrelevant to the key performance indicators (KPIs) and data fields with empty values ​​to form an initial data set; (2) Classify the data fields in the initial data set into entities, classify the associations between the fields, and connect the entities with associations in a directed manner to form a wireless communication network knowledge graph with the target KPI as the core; (3) Calculate the edge connection strength I(U; V) between node U and node V in the knowledge graph: Among them, p(u,v) is the joint probability distribution function of the values ​​of node U and node V in the knowledge graph, and p(u) and p(v) are the marginal probability distribution functions of the values ​​of node U and node V respectively; (4) Calculate the impact efficiency between nodes: Normalize the edge connection strength I(U; V) to get the edge weight For any two nodes X and Y that have a connected path, calculate the product of the edge weights on their qth shortest path to obtain the influence efficiency e between the two nodes XY ; (5) The calculated influence efficiency of each node on the target KPI node is sorted from large to small, and the first t data fields selected are adjusted to optimize the target KPI.

2. The method according to claim 1, characterized in that In step (2), the data fields in the initial data set are classified into four categories based on the meaning of each data field in the wireless communication network and the communication network protocol: network-level performance evaluation indicators, user-level performance evaluation indicators, general non-adjustable data parameters, and adjustable data parameters. Network-level performance evaluation indicators are used to evaluate the performance level of the entire wireless communication network; User-level performance evaluation indicators are used to evaluate the user's communication level; General non-adjustable data parameters refer to fixed parameter values ​​in the communication network; Adjustable data parameters refer to parameters that can be adjusted to change network or user performance.

3. The method according to claim 1, characterized in that In step (2), the associations between fields are classified into three categories: causal relationships, implicit relationships, and explicit relationships, where: Causation refers to the direct effect of one entity on another; Implicit relationship refers to the indirect influence of one entity on another entity, that is, there is no clear and specific expression to express the relationship between the two; An explicit relationship means that after reasoning and analysis between two entities, a specific analytical expression can be obtained.

4. The method according to claim 1, characterized in that: In step (2), directed connections are made between entities with associated relationships, which can be implemented as follows: After determining the entity type to which the data field belongs and the relationship type to which the association between the data fields belongs, connect the head entity h and the tail entity t through the line segment r to construct a triple (h, r, t); The head entity and tail entity in each triple are regarded as a node in the wireless communication network knowledge graph, and the association relationship between the head entity and the tail entity is regarded as an edge in the graph.

5. The method according to claim 1, characterized in that The joint probability distribution function p(u,v) of the values ​​of nodes U and V in the knowledge graph involved in the edge connection strength I(U;V) in step (3) is expressed as follows: p(u,v)=P{U=u,V=v}.

6. The method according to claim 1, characterized in that The marginal probability distribution function p(u) of the node U value and the marginal probability distribution function p(v) of the node V value in the knowledge graph involved in the edge connection strength i(U; V) in step (3) are expressed as follows:

7. The method according to claim 1, characterized in that The normalized edge weights obtained in step (4) It is expressed as follows: Among them, U is the cause node, V is the result node, for any (U,V)∈G, (U,V) indicates that there is a directed edge between node U and node V, G represents all nodes in the knowledge graph of wireless communication networks, for any result node V, there are m cause nodes U, represents the edge weight of the mth cause node.

8. The method according to claim 1, characterized in that The influence efficiency e between the two nodes is calculated in step (4): XY , the expression is as follows: in, It represents the product of the edge weights of all edges on the qth shortest path from node X to node Y. The qth shortest path is a reference, that is, if there are y shortest paths, the shortest path with the largest product of the edge weights of all edges on the path is selected as the qth shortest path.

9. The method according to claim 1, characterized in that: In step (5), the first t data fields selected are adjusted by selecting adjustable data parameters from the first 40 data fields selected and adjusting them according to the parameter value range specified by the wireless communication network protocol.

Images