Wireless network planning method and device

Abstract

The present application provides a wireless network planning method and apparatus. The method comprises: obtaining multiple sets of indicator values for M wireless network performance indicators of a wireless network; M being a positive integer, and one set of indicator values in the multiple sets of indicator values being the values of each of the M wireless network performance indicators corresponding to multiple moments within a time window; determining, based on the multiple sets of indicator values and the delay state corresponding to each set of indicator values in the multiple sets of indicator values, a correlation between each of the M wireless network performance indicators and the delay state; determining N wireless network performance indicators from the M wireless network performance indicators, wherein the correlation of each of the N wireless network performance indicators meets a requirement, and N being a positive integer less than or equal to M; and determining a value range for each of the N wireless network performance indicators based on the delay state to plan the wireless network. This method implements automatic delay-oriented wireless network planning.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a wireless network planning method and device. Background Art

[0002] The high bandwidth, low latency, and high reliability of 5G networks have spurred the emergence of many latency-sensitive services. Examples include Ultra High Definition (UHD) live video streaming and Cloud Virtual Reality (Cloud VR). Traditional coverage-based wireless network planning methods for these services are inconsistent with service requirements. Therefore, to support latency experience assessment and ensure the needs of these latency-sensitive services, an automated latency-focused wireless network planning solution is urgently needed.

[0003] Application Contents

[0004] The embodiments of the present application provide a wireless network planning method and apparatus, which solve the problem of automatic wireless network planning oriented to latency indicators.

[0005] In a first aspect, the present application provides a wireless network planning method, including: obtaining multiple groups of indicator values of M wireless network performance indicators of a wireless network; wherein M is a positive integer, and one group of indicator values in the multiple groups of indicator values is the value of each indicator in the M wireless network performance indicators corresponding to multiple time moments within a time window; determining the correlation between each indicator in the M wireless network performance indicators and the delay state based on the multiple groups of indicator values and the delay state corresponding to each group of indicator values in the multiple groups of indicator values; wherein the delay state corresponding to any group of indicator values in the multiple groups of indicator values is a state that meets the target delay requirement or a state that does not meet the target delay requirement; determining N wireless network performance indicators from the M wireless network performance indicators, wherein the correlation of each wireless network performance indicator in the N wireless network performance indicators meets the requirement, and N is a positive integer less than or equal to M; determining the value range of each indicator of the N wireless network performance indicators based on the delay state to plan the wireless network.

[0006] In this method, multiple groups of indicator values of M wireless network performance indicators that affect wireless network delay are first obtained. Then, based on the multiple groups of indicator values and the delay state corresponding to each group of indicator values, the correlation between each indicator in the M wireless network performance indicators and the delay state is determined. Then, based on the correlation, N wireless network performance indicators that have a greater impact on the delay state of the wireless network are determined. Finally, the wireless network is planned based on the N wireless network performance indicators, realizing automatic delay-oriented wireless network planning and planning a wireless network that meets the target delay requirements.

[0007] In one possible implementation, the correlation between each of M wireless network performance indicators and the delay state is determined based on multiple groups of indicator values and the delay state corresponding to each group of indicator values in the multiple groups of indicator values, including: training a first network model using multiple first training sample pairs consisting of the multiple groups of indicator values and the delay states corresponding to the multiple groups of indicator values; determining the Gini impurity between each of the M wireless network performance indicators and the delay state based on the trained first network model; wherein the Gini impurity represents the correlation, and a first training sample pair includes a group of indicator values in the multiple groups of indicator values and the delay state corresponding to the group of indicator values.

[0008] In one possible implementation, the first network model is a decision tree model.

[0009] In one possible implementation, determining N wireless network performance indicators from the M wireless network performance indicators, wherein the correlation of each of the N wireless network performance indicators meets a requirement, includes: determining the N wireless network performance indicators based on Gini impurity; wherein the Gini impurity between each of the N wireless network performance indicators and the delay state is greater than or equal to a threshold. After determining the N wireless network performance indicators that have a significant impact on the delay state of the wireless network, the values of the N wireless network performance indicators are adjusted to an appropriate range to plan a wireless network that meets the delay requirements.

[0010] In another possible implementation, the threshold is preset, for example, the threshold may be preset to 0.001.

[0011] In another possible implementation, determining the value range of each indicator of N wireless network performance indicators based on the delay state includes: determining the value range of each indicator of N wireless network performance indicators based on at least one set of indicator values that meets the standard among multiple sets of indicator values, wherein the delay state corresponding to the at least one set of indicator values that meets the standard is a state that meets the target delay requirement.

[0012] In another possible implementation, the method further includes: inputting each set of indicator values from the multiple sets of indicator values into a wireless network's delay evaluation model to obtain a delay value corresponding to each set of indicator values; determining a delay state corresponding to each set of indicator values based on whether the delay value corresponding to each set of indicator values meets a target delay requirement, and if so, the delay state is a state that does not meet the target delay requirement, or if not, the delay state is a state that meets the target delay requirement. The embodiment of the present application obtains delay values by training a wireless network's delay evaluation model, thereby resolving the problem that the traditional delay value acquisition method (obtaining delay through ping) requires reliance on TCP-like communication protocols with a response mechanism, laying the foundation for wireless network delay-oriented wireless network planning.

[0013] In another possible implementation, the delay evaluation model is a second network model trained by multiple second training sample pairs, wherein one second training sample pair among the multiple second training sample pairs is composed of the indicator value of each indicator of M wireless network performance indicators in one time window of multiple time windows and the delay value corresponding to the time window.

[0014] In one possible implementation, the second network model is an LSTM model.

[0015] In a possible implementation, the length of the time window ranges from 10 to 50 seconds.

[0016] The method of the embodiment of the present application constructs a delay evaluation model based on the LSTM deep learning algorithm. The LSTM deep learning algorithm can effectively learn the time cumulative effect of wireless network performance indicators and then characterize the cumulative effect of wireless network delay, making the delay value prediction more accurate and able to output continuous delay values.

[0017] In another possible implementation, the M wireless network performance indicators are determined based on expert experience and data statistical analysis.

[0018] Specifically, the M wireless network performance indicators include at least one or more of a coverage status indicator, a retransmission error indicator, a channel scheduling indicator, a cell switching indicator, a distance indicator from a base station, and a congestion index indicator.

[0019] In another possible implementation, wireless network performance indicator dotted data corresponding to M wireless network performance indicators is collected from the CHR of a base station network management system of the wireless network; and a value for each of the M wireless network performance indicators is obtained based on the dotted data. In this embodiment of the present application, the value for each of the M wireless network performance indicators is obtained by obtaining the wireless network performance indicator dotted data from the CHR of the base station network management system. This eliminates the need for re-dotted data and modification of the data message structure, resulting in a low deployment cost and unlimited application scenarios.

[0020] In a second aspect, the present application also provides a wireless network planning device, including: an acquisition module, used to obtain multiple groups of indicator values of M wireless network performance indicators of a wireless network; wherein M is a positive integer, and one group of indicator values in the multiple groups of indicator values is the value of each indicator in the M wireless network performance indicators corresponding to multiple moments within a time window; a processing module, used to determine the correlation between each indicator in the M wireless network performance indicators and the delay state based on the multiple groups of indicator values and the delay state corresponding to each group of indicator values in the multiple groups of indicator values; wherein the delay state corresponding to any group of indicator values in the multiple groups of indicator values is a state that meets the target delay requirement or a state that does not meet the target delay requirement; a determination module, determining N wireless network performance indicators from the M wireless network performance indicators, wherein the correlation of each wireless network performance indicator in the N wireless network performance indicators meets the requirement, N is a positive integer, and N is less than or equal to M; a planning module, determining the value range of each indicator of the N wireless network performance indicators according to the delay state to plan the wireless network.

[0021] In another possible implementation, the above-mentioned processing module is specifically used to: train a first network model using multiple first training sample pairs, each of the multiple first training sample pairs including a set of indicator values from the multiple sets of indicator values and a delay state corresponding to the set of indicator values; determine the Gini impurity between each indicator in the M wireless network performance indicators and the delay state based on the trained first network model; wherein the Gini impurity represents the correlation.

[0022] In one possible implementation, the first network model is a decision tree model.

[0023] In another possible implementation, the determination module is specifically configured to: determine N wireless network performance indicators based on Gini impurity; wherein the Gini impurity between each of the N wireless network performance indicators and the delay state is greater than or equal to a threshold.

[0024] In another possible implementation, the threshold is preset, for example, the threshold may be preset to 0.001.

[0025] In another possible implementation, the planning module is specifically used to determine the value range of each indicator in N wireless network performance indicators based on at least one set of indicator values that meets the standard among multiple sets of indicator values, wherein the delay state corresponding to the at least one set of indicator values that meets the standard is a state that meets the target delay requirement.

[0026] In another possible implementation, the wireless network planning device also includes: a delay state determination module, which is used to input each group of indicator values in the multiple groups of indicator values into the delay evaluation model of the wireless network to obtain the delay value corresponding to each group of indicator values; determine the delay state corresponding to each group of indicator values according to whether the delay value corresponding to each group of indicator values meets the target delay requirement, if so, the delay state is a state that does not meet the target delay requirement, or if not, the delay state is a state that meets the target delay requirement.

[0027] In another possible implementation, the delay evaluation model is a second network model trained by multiple second training sample pairs, wherein one second training sample pair among the multiple second training sample pairs is composed of the indicator value of each indicator of M wireless network performance indicators in one time window of multiple time windows and the delay value corresponding to the above-mentioned one time window.

[0028] In one possible implementation, the second network model is an LSTM model.

[0029] In a possible implementation, the time window length ranges from 10 to 50 seconds.

[0030] In another possible implementation, the M wireless network performance indicators are determined based on expert experience and data statistical analysis.

[0031] Specifically, the M wireless network performance indicators include at least one or more of a coverage status indicator, a retransmission error indicator, a channel scheduling indicator, a cell switching indicator, a distance indicator from a base station, and a congestion index indicator.

[0032] In another possible implementation, wireless network performance indicator marking data corresponding to M wireless network performance indicators in the CHR of the wireless network base station network management system is collected; and the value of each indicator in the M wireless network performance indicators is obtained according to the marking data.

[0033] In a third aspect, the present application also provides a wireless network delay evaluation method, including: obtaining multiple groups of indicator values of M wireless network performance indicators of the wireless network; wherein M is a positive integer, and one group of indicator values in the multiple groups of indicator values is the value of each indicator in the M wireless network performance indicators corresponding to multiple time moments within a time window; inputting each group of indicator values in the multiple groups of indicator values into a delay evaluation model of the wireless network to obtain the delay value corresponding to each group of indicator values; wherein the delay evaluation model is a second network model trained by multiple second training sample pairs, wherein one second training sample pair in the multiple second training sample pairs is composed of the indicator value of each indicator in the M wireless network performance indicators in a time window within multiple time windows and the delay value corresponding to the time window.

[0034] The wireless network delay value evaluation method provided in this application solves the problem of difficulty in obtaining the delay value of the wireless network by constructing a delay evaluation model and using wireless network performance indicators as input features.

[0035] In one possible implementation, the second network model is an LSTM model.

[0036] In a possible implementation, the length of the time window ranges from 10 to 50 seconds.

[0037] In another possible implementation, the M wireless network performance indicators are determined based on expert experience and data statistical analysis.

[0038] Specifically, the M wireless network performance indicators include at least one or more of a coverage status indicator, a retransmission error indicator, a channel scheduling indicator, a cell switching indicator, a distance indicator from a base station, and a congestion index indicator.

[0039] In another possible implementation, wireless network performance indicator marking data corresponding to M wireless network performance indicators in the CHR of the wireless network base station network management system is collected; and the value of each indicator in the M wireless network performance indicators is obtained according to the marking data.

[0040] In a fourth aspect, the present application also provides a wireless network delay evaluation device, including: an acquisition module, used to obtain multiple groups of indicator values of M wireless network performance indicators of the wireless network; wherein M is a positive integer, and one group of indicator values in the multiple groups of indicator values is the value of each indicator in the M wireless network performance indicators corresponding to multiple time moments within a time window; a delay evaluation model generation module, used to generate a delay evaluation model, wherein the delay evaluation model is a second network model trained by multiple second training sample pairs, wherein one second training sample pair in the multiple second training sample pairs is composed of the indicator value of each indicator in the M wireless network performance indicators in a time window in multiple time windows and the delay value corresponding to the time window; an input module, used to input each group of indicator values in the multiple groups of indicator values into the delay evaluation model of the wireless network to obtain the delay value corresponding to each group of indicator values.

[0041] In one possible implementation, the second network model is an LSTM model.

[0042] In a possible implementation, the length of the time window ranges from 10 to 50 seconds.

[0043] In another possible implementation, the M wireless network performance indicators are determined based on expert experience and data statistical analysis.

[0044] Specifically, the M wireless network performance indicators include at least one or more of a coverage status indicator, a retransmission error indicator, a channel scheduling indicator, a cell switching indicator, a distance indicator from a base station, and a congestion index indicator.

[0045] In another possible implementation, wireless network performance indicator marking data corresponding to M wireless network performance indicators in the CHR of the wireless network base station network management system is collected; and the value of each indicator in the M wireless network performance indicators is obtained according to the marking data.

[0046] In the fifth aspect, the present application also provides a computing device, including a memory and a processor, wherein the memory stores executable code, and the processor executes the executable code to implement the method provided in the first aspect and / or the method provided in the third aspect of the present application. It is easy to understand that the computing device can be a wireless network planning device or a wireless network delay evaluation device.

[0047] In a sixth aspect, the present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed in a computer, it enables the computer to execute the method provided in the first aspect and / or the method provided in the third aspect of the present application.

[0048] In a seventh aspect, the present application also provides a computer program or computer program product, which includes instructions that, when executed, cause a computer to execute the method provided in the first aspect and / or the method provided in the third aspect of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 A schematic diagram of the architecture of a wireless network planning system provided in an embodiment of the present application;

[0050] Figure 2 The present invention provides a flowchart of a wireless network planning method according to an embodiment of the present invention;

[0051] Figure 3 A schematic diagram of determining M wireless network performance indicators based on expert experience provided in an embodiment of the present application;

[0052] Figure 4 A schematic diagram of the structure of the decision tree model provided in the embodiment of the present application;

[0053] Figure 5 A flowchart of a wireless network delay evaluation method provided in an embodiment of the present application;

[0054] Figure 6 A schematic diagram of a learning process for a wireless network delay assessment model provided in an embodiment of the present application;

[0055] Figure 7Schematic diagram of the loss function during the delay evaluation model training process provided in an embodiment of the present application;

[0056] Figure 8 A schematic diagram of the delay evaluation model structure provided in an embodiment of the present application;

[0057] Figure 9 This is a structural diagram of the LSTM model;

[0058] Figure 10 Schematic diagram of Sigmoid function;

[0059] Figure 11 A schematic diagram of a sequence of delay values output by a delay evaluation model provided in an embodiment of the present application;

[0060] Figure 12 A schematic diagram of the structure of a wireless network planning device provided in an embodiment of the present application;

[0061] Figure 13 A schematic diagram of the structure of a wireless network delay evaluation device provided in an embodiment of the present application;

[0062] Figure 14 A schematic diagram of the structure of the computing device provided in this application. DETAILED DESCRIPTION

[0063] The technical solution of the present application is further described in detail below through the accompanying drawings and examples.

[0064] A wireless network planning method and device of an embodiment of the present application can be applied to various wireless communication network systems, such as: global system of mobile communication (GSM) system, code division multiple access (CDMA) system, wideband code division multiple access (WCDMA) system, general packet radio service (GPRS), long term evolution (LTE) system, LTE frequency division duplex (FDD) system, LTE time division duplex (TDD), universal mobile telecommunication system (UMTS) or world-wide interoperability for microwave access (WiMAX) communication system or 5G communication system or future communication system, etc.

[0065] Figure 1 The following is an architectural diagram of a wireless network planning system provided by an embodiment of the present application. The architectural diagram shows a radio access network (RAN) system 1 of a wireless communication network system and a wireless network planning system 2 in communication therewith, wherein the radio access network system 1 includes a number of base stations (e.g., base station 11, base station 12, ..., base station N in the figure) and a network management system 10 for managing the base stations. The wireless network planning system 2 interacts with the network management system 10. The wireless network planning system 2 includes a delay evaluation device 21 and a wireless network planning device 22. The delay evaluation device 21 is used to obtain wireless network performance indicator data from the network management system 10 and determine the delay value of the wireless network based on the wireless network performance indicator data; the planning device 22 is used to determine the wireless network performance indicators that have a greater impact on the wireless network delay status based on the wireless network performance indicator data, and further determine planning suggestions for the wireless network to meet the target delay requirements based on these wireless network performance indicators.

[0066] Specifically, the delay evaluation device 21 is used to collect M wireless network performance indicator data from the CHR (Call History Record) from the network management system 10, where M is a positive integer, determine the delay value of the wireless network based on the M wireless network performance indicator data, and output the delay value and send the delay value and the M wireless network performance indicator data to the wireless network planning device 22. The wireless network planning device 22 determines the delay state corresponding to the delay value, and determines N wireless network performance indicators (N is a positive integer less than or equal to M) that have a greater impact on the delay state of the wireless network based on multiple groups of M wireless network performance indicator data and delay states, and then outputs planning suggestions based on the N wireless network performance indicators to enable the wireless network to meet the target delay requirements.

[0067] It can be understood that the delay evaluation device 21 and the wireless network planning device 22 can be separately set up (for example, two different physical entities or logical modules) or combined together (for example, the same physical entity or logical module), and this application does not limit this.

[0068] When the delay evaluation device 21 and the wireless network planning device 22 are separately provided, the delay evaluation device 21 can be provided within the network management system 10. For example, the delay evaluation device 21 can be deployed as a plug-in within the network management system 10, facilitating the delay evaluation device 21 to obtain wireless network performance indicator data from the network management system 10. The delay evaluation device 21 transmits the wireless network delay value derived based on the wireless network performance indicator to the wireless network planning device 22, which is provided outside the network management system 10. The wireless network planning device 22 can obtain the wireless network performance indicator data from the network management system 10 or the delay evaluation device 21, and determine planning recommendations based on multiple sets of wireless network performance indicator data and the wireless network delay values to ensure that the wireless network meets the target delay requirement.

[0069] When the delay evaluation device 21 and the wireless network planning device 22 are integrated, the wireless network planning system 2 can be installed within the network management system 10, allowing the delay evaluation device 21 and / or the wireless network planning device 22 to obtain wireless network performance indicator data from the network management system 10. The wireless network planning system 2 can also be installed external to the network management system 10, eliminating the need to modify the network management system 10 and achieving zero modification to the existing network structure.

[0070] Of course, the wireless network planning system 2 may also not include the delay evaluation device 21. The wireless network planning device 22 is directly connected to the network management system 10 for interaction, obtains multiple sets of wireless network performance indicator data and the delay value corresponding to each set of wireless network performance indicator data (for example, the delay value can be measured or obtained through ping values, etc.), determines the delay state based on the delay value, and determines the wireless network performance indicators that have a greater impact on the wireless network delay state based on the multiple sets of wireless network performance indicator data and the delay state corresponding to each set of wireless network performance indicator data, and further determines planning recommendations for the wireless network to meet the target delay requirements based on these wireless network performance indicators.

[0071] Figure 2 A flowchart of a wireless network planning method provided by an embodiment of the present application is shown. It is understood that this method can be executed by the aforementioned wireless network planning system. It is also understood that the wireless network planning system can be any device, equipment, platform, or device cluster with computing and processing capabilities. The specific implementation of each step is described below.

[0072] Step 201: Acquire multiple groups of indicator values of M wireless network performance indicators of a wireless network; wherein M is a positive integer, and one group of indicator values in the multiple groups of indicator values is a value of each indicator in the M wireless network performance indicators corresponding to multiple moments in a time window.

[0073] M wireless network performance indicators that may affect wireless network delay may be determined based on expert experience and data statistical analysis methods.

[0074] For example, based on expert experience, seven categories of wireless network performance indicators are determined, and their specific items are described in the following table:

[0075]

[0076]

[0077] Based on expert experience and data statistical analysis, the spectrum efficiency index and data channel status index can be used to construct a high-order congestion index (Ind cong ) is used to characterize the congestion state of the wireless network. As a possible implementation method, the corresponding M wireless network performance indicators are determined according to the above-mentioned categories of wireless network performance indicators. The M wireless network performance indicators include at least one or more of the following: coverage status indicator, retransmission error indicator, channel scheduling indicator, cell switching indicator, distance indicator from the base station, congestion index indicator (such as Figure 3 shown).

[0078] Furthermore, the wireless network planning device collects wireless network performance indicator marking data corresponding to M wireless network performance indicators from the CHR of the base station network management system of the wireless network, and obtains multiple groups of indicator values of the M wireless network performance indicators from the marking data, wherein a group of indicator values is the value of each indicator of the M wireless network performance indicators corresponding to multiple moments within a time window.

[0079] For example, a sliding time window method can be used to obtain multiple sets of data from the wireless network performance indicator data. See Table 1. The length of a time window can be 10s. The data in Table 1 represents the value of each indicator in the M wireless network performance indicators corresponding to multiple moments in the time window.

[0080]

[0081]

[0082] Table 1

[0083] Step S202: Determine the correlation between each indicator in the M wireless network performance indicators and the delay state based on the multiple groups of indicator values and the delay state corresponding to each group of indicator values in the multiple groups of indicator values; wherein the delay state corresponding to any one group of indicator values in the multiple groups of indicator values is a state that meets the target delay requirement or a state that does not meet the target delay requirement.

[0084] In step 201, multiple sets of M wireless network performance indicators are obtained. This step also obtains the delay status corresponding to each of the multiple sets of indicator values. It is understood that the delay status includes a state where the target delay requirement is met or a state where the target delay requirement is not met. When the wireless network delay value is less than or equal to the target delay value, the current delay status is determined to be a state where the target delay requirement is met. When the wireless network delay value is greater than the target delay value, the current delay status is determined to be a state where the delay requirement is not met. The delay value can be measured, obtained by obtaining a network ping value, or calculated and predicted. For example, a delay evaluation model can be constructed, and each set of indicator values from the multiple sets of indicator values can be input into the delay evaluation model to obtain the delay value corresponding to each set of indicator values. The specific construction and training of the delay evaluation model are described below.

[0085] It should be noted that each set of multiple sets of M wireless network performance indicators represents the values of each of the M wireless network performance indicators at multiple moments within a time window. For example, when the time window length is 10 seconds, each set of indicator values represents the values of each of the M wireless network performance indicators at multiple moments within that 10-second period. This means that each set of indicator values is a multidimensional array. For details, see the data in Table 1. The latency status corresponding to each set of indicator values represents the latency status at the moment within the time window to which each set of indicator values belongs.

[0086] It is understandable that the target delay value can be preset. For example, the target delay value can be the value of the SLA delay guarantee indicator. The SL delay guarantee indicator is generated based on the collected user requirements for wireless network delay. For example, if the user demand is not to have a network delay greater than 40ms, then the SLA delay guarantee indicator is 40ms.

[0087] Then, based on the multiple sets of indicator values obtained above and the delay states corresponding to each of the multiple sets of indicator values, the correlation between each of the M wireless network performance indicators and the delay state is determined. There are various methods for calculating the correlation. For example, the Gini impurity can be used to represent the correlation. The Gini impurity can be calculated using a network model. For example, a first network model is trained using multiple first training sample pairs consisting of the multiple sets of indicator values and the delay states corresponding to the multiple sets of indicator values. The Gini impurity between each of the M wireless network performance indicators and the delay state is then determined based on the trained first network model.

[0088] Optionally, the first training model may be a decision tree model.

[0089] Specifically, for calculation convenience, a label may be added to the delay state. A label 0 is added to the state that meets the target delay requirement, and a label 1 is added to the state that does not meet the target delay requirement.

[0090] Furthermore, a decision tree model for delay analysis is trained using data sample pairs composed of multiple sets of indicator values and their corresponding delay status labels. The trained decision tree model is then used to calculate the Gini impurity of the input features of each of the M wireless network performance indicators (it should be explained here that the input features of each of the M wireless network performance indicators represent an abstraction of the indicator value of each of the M wireless network performance indicators, and a wireless network performance indicator value is a specific instance of the input features of the wireless network performance indicator) relative to the delay status label. The Gini impurity represents the correlation between the delay status label and the input features of each of the M wireless network performance indicators. This is shown in Table 2. The horizontal axis represents the M wireless network performance indicators, and the vertical axis represents the time window. The fill value in the table represents the Gini impurity between a wireless network performance indicator in the M wireless network performance indicators and the delay status at a specific moment. The last row shows the average Gini impurity between each of the M wireless network performance indicators and the delay status within the time window.

[0091] In other words, the Gini impurity in Table 2 has two meanings:

[0092] One of the meanings is: within a time window (for example, the time window length is 10s, i.e., 10s), the correlation between the input features of the wireless network performance indicators at different times and the delay state label within the time window;

[0093] Another meaning is: based on the Gini impurity at different moments in the time window, the Gini impurity of the entire time window is obtained (the method for determining the Gini impurity of the entire time window here is not unique. The embodiment of the present application uses the average value of the Gini impurity at each moment in the time window as the Gini impurity of the time window).

[0094]

[0095] Table 2

[0096] The specific structure of the decision tree model can be found in Figure 4, the decision tree model includes a root node, leaf nodes, and internal nodes; wherein, the root node is the input feature, for example, the input feature is the input feature of each indicator in the M wireless network performance within the time window (Input_feature=time1_ServingSS-RSRP,…,time1_hanover_3,time2_Serving SS-RSRP,…,time2_hanover_3,…,time10_Serving SS-RSRP,…,time10_hanover_3), and the input instance is the sample instance corresponding to the input feature, that is, the indicator value of each indicator in the M wireless network performance indicators. The leaf node is the output result corresponding to the input instance, that is, the state label 0 that meets the target delay requirement or the state label 1 that does not meet the target delay requirement. The internal node is the judgment condition for the left and right selection of the branch, and the judgment condition is the value state of a certain dimension feature. For example, Figure 4 In the tree structure shown, assuming that the judgment condition of the internal node is time1_hanover_3<1, the output result of the left leaf node is 0.

[0097] The training process of the decision tree model:

[0098] Based on a large amount of sample data (the first training sample pairs), the model uses the value range of each dimension's features (for example, a wireless network performance indicator at each moment) to set reasonable judgment conditions (how to judge reasonable: using these conditions to reduce uncertainty, which is also referred to as Gini impurity. In layman's terms, for a specific wireless network performance indicator, choosing the appropriate split point to maximize the final classification accuracy) to ensure that as many samples as possible are correctly classified. In this process, the internal nodes (decision conditions) formed, together with the root node and corresponding leaf nodes, form a decision tree model.

[0099] When a decision tree model with a classification accuracy that meets the requirements is obtained through training, the parameters of its internal nodes are also obtained, namely the Gini impurity of the wireless network performance index and the delay state.

[0100] S203. Determine N wireless network performance indicators from the M wireless network performance indicators, wherein the correlation corresponding to each wireless network performance indicator in the N wireless network performance indicators meets the requirement, N is a positive integer, and N is less than or equal to M.

[0101] It should be noted that the meaning of "the correlation corresponding to each of the N wireless network performance indicators meeting the requirement" is that the correlation between each of the N wireless network performance indicators and the delay state is greater than a threshold. From the previous step S202, we obtain the Gini impurity that represents the correlation. The N wireless network performance indicators can be determined based on the Gini impurity. In one example, wireless network performance indicators with a Gini impurity greater than a threshold are selected. Optionally, the threshold can be 0.001. This means that wireless network performance indicators with a Gini impurity greater than 0.001 are selected. The selected wireless network performance indicators are the N wireless network performance indicators.

[0102] In another example, a hard threshold algorithm (for example, a threshold of 0.001) is used to exclude wireless network performance indicator input features with relatively low Gini impurity. Then, the remaining wireless network performance indicator input features are screened, and the first N wireless network performance indicator input features are screened in ascending order. For example, if N is 3, the first three wireless network performance indicator input features (wireless network performance indicators in ascending order) are screened. If the number of features is less than three, all are screened. The wireless network performance indicators corresponding to the screened features are the N wireless network performance indicators.

[0103] Step S204: Determine a value range of each of the N wireless network performance indicators according to the delay status to plan the wireless network.

[0104] The joint distribution of N wireless network statistics under the condition that the delay label is 0 (the common value range of the indicator values of the first three wireless network performance indicators in the ascending order above) is counted to obtain wireless network planning suggestions.

[0105] Specifically, among the multiple groups of indicator values of M wireless network performance indicators of the wireless network, the value range W1 of the wireless network performance indicator ranked first in ascending order, the value range W2 of the wireless network performance indicator ranked second in ascending order, and the value range W3 of the wireless network performance indicator ranked third in ascending order are counted. The value ranges W1&W2&W3 of the first three wireless network performance indicators in ascending order are output as planning suggestions for the wireless network, so that the wireless network planning device plans the wireless network according to at least the value ranges W1&W2&W3 of the first three wireless network performance indicators in ascending order to obtain a wireless network that meets the target delay requirements.

[0106] In one example, the wireless network planning device can output planning suggestions for the wireless network (for example, the value range W1&W2&W3 of the first three wireless network performance indicators in ascending order), delay status and / or delay distribution. For the specific method of obtaining the delay status, please refer to the above step 202, and for the method of obtaining the delay distribution, please refer to the description below.

[0107] The present application also discloses a method for evaluating wireless network delay, such as Figure 5 As shown, the following steps are included:

[0108] S801. Obtain multiple groups of indicator values of M wireless network performance indicators of the wireless network; where M is a positive integer, and one group of indicator values in the multiple groups of indicator values is a value of each indicator in the M wireless network performance indicators corresponding to multiple moments in a time window.

[0109] Specifically, the method for obtaining multiple groups of indicator values of the M wireless network performance indicators of the wireless network refers to the above step 201, which will not be repeated here.

[0110] Step 802: Input each group of indicator values from the multiple groups of indicator values into a delay evaluation model of the wireless network to obtain a delay value corresponding to each group of indicator values; wherein the delay evaluation model is a second network model trained by multiple second training sample pairs, wherein one second training sample pair from the multiple second training sample pairs is composed of an indicator value of each indicator of M wireless network performance indicators in one time window from the multiple time windows and a delay value corresponding to the time window.

[0111] The purpose of learning the delay evaluation model is to obtain the mapping relationship between multiple groups of indicator values of M wireless network performance indicators and delay values of the wireless network using the second network model based on a large amount of existing historical observation data.

[0112] Optionally, the second network model may be an LSTM network model, and the LSTM network obtained by learning the delay evaluation model is recorded as function g(.). The specific steps of learning the delay evaluation model are as follows: Figure 6 shown.

[0113] Step 901: Perform data cleaning and feature normalization on the historically observed wireless network performance indicator data and delay data. In this example, data cleaning includes the following two steps:

[0114] (1) Eliminate abnormal data values, such as values greater than a preset range;

[0115] (2) Eliminate input feature samples with missing data.

[0116] It can be understood that the feature normalization processing refers to normalization processing using a rescaling normalization method, the wireless network performance indicator data includes the indicator value of the wireless network performance indicator, and the delay data includes the delay value.

[0117] Step 902: Determine the input feature time window length n, i.e., the time window of wireless network performance indicator data required to evaluate network latency. For example, assuming the time window length n = 10, this means using the wireless network performance indicator values from seconds 1-10 within the time window to predict the latency value at the 10th second. In this embodiment, the time window length ranges from 10 to 50 seconds.

[0118] Step 903: Use the ping delay data in the historical observation data as the training label, and the mean square error function as the loss function. In the embodiment of the present application, since we consider evaluating the delay of the wireless network, the delay data in the real historical observation data is used as the training label in the learning of the delay evaluation model. On the other hand, in this embodiment, since we use the LSTM model (common models include RNN (Recurrent Neural Networks, recurrent neural network) model and HMM (Hidden Markov Model)) to build a regression model, the mean square error is used as the loss function in the training process of the delay evaluation model (common regression model loss functions include mean absolute error (MAE) and root mean square error (RMSE)). The loss function is used to measure the status of model training. The loss function is specifically: Loss = (label Delay -g(feature)) 2 .

[0119] Step 904: Use the historically observed wireless network performance indicator data and the corresponding delay data as data sample pairs to perform deep learning network model training. When the loss function converges, a trained delay evaluation model is obtained. A data sample pair consists of the indicator value of each indicator in the M wireless network performance indicators within a time window and the delay value corresponding to the time window. For example, the time window length n = 10s, the data sample pair consists of 10 consecutive seconds of wireless network performance indicator data and the delay data corresponding to the 10th second of the consecutive 10 seconds, or the data sample pair consists of 10 consecutive seconds of wireless network performance indicator data and the delay data corresponding to the 5th second of the consecutive 10 seconds. Figure 7 As shown, Figure 7 The horizontal axis represents the number of iterations of the algorithm model, and the vertical axis represents the loss function, which is the mean square error between the model evaluation delay and the delay in the real historical observation data. Figure 7 As shown in the figure, when the number of iterations is less than 150, the loss function of the model is in a fluctuating state. When the number of iterations is greater than 150, the loss function of the model converges stably to around 0.06, indicating that the model has reached the optimal solution, that is, the model has been trained.

[0120] Figure 8This is a schematic diagram of the delay evaluation model structure provided in the embodiment of the present application, such as Figure 8 As shown in FIG, the delay evaluation model includes at least four layers, namely, an input layer, a long short-term memory layer, an attention layer, and an output layer.

[0121] LSTM (Long Short Term Memory) is a well-known long short-term memory network, which is primarily used for predicting time series problems. Its algorithm uses a "gate control" unit to control the impact of input features at different times on the output results.

[0122] Existing deep learning time series prediction algorithms (especially RNNs) suffer from long-term dependency issues. This occurs because after many stages of computation, features from earlier time slices are overwritten. For example, if we want to predict the singular or plural form of the verb preceding "full" in the sentence "The cat, which already ate a bunch of food, was full," the word "full" clearly depends on the singular or plural form of the second word, "cat," not the preceding word, "food." However, the RNN structure loses its ability to learn such long-term information as time slices increase.

[0123] The LSTM model uses the "gate control" unit to save the input features at different times and transmit them to the output end, thus solving the long-term dependency problem of the model.

[0124] Figure 9 This is a structural diagram of the LSTM model. t-1 、X t ,…X t+n It represents the features corresponding to the t-1, t, ... t+n moments in the input feature set (i.e., the input features of the LSTM network are a combination of features at multiple moments, not just a single moment). The corresponding output is h t-1 、h t 、h t+n (That is, the input features at each moment will correspond to an output result, and this result can be output or not).

[0125] Figure 9 It shows that LSTM is composed of multiple LSTM basic units connected in series, where each basic unit shares the same network. Figure 9 It can be seen that LSTM consists of two memory lines, namely long-term memory C t and short-term memory t Among them, long-term memory C tRepresents the state learned by the network model from the previous time series. Its state update formula is as follows:

[0126]

[0127] The above formula consists of two parts, among which f t ×C t-1 Part represents the long-term memory part, where f t Represents the "forget gate", which is used to indicate whether to accept or forget the memory C at time t-1 t-1 In the embodiment, it means that when predicting the delay value at time t, the memory related to the wireless network performance indicator at time t-1 needs to be acquired or forgotten. The implementation principle of the forget gate is as follows:

[0128] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0129] σ=sigmoid(·)

[0130] W in the formula f represents the weight parameter, b f Represents the bias. The input of the “forget gate” is the input feature x at this moment. t and the predicted delay h at the previous moment t-1 In the LSTM model, the value of the sigmoid function tends to two extremes (0 or 1), such as Figure 10 As shown, it plays the role of a control unit. In the LSTM model, it means that the model will be based on the input data feature x at the current moment. t And the predicted output delay result h at the previous moment t-1 Decide together, for the long-term memory C of the previous moment t trade-offs.

[0131] Part of it represents the new memory learned at the current moment, and updates these memories to the long-term memory C t The principle is as follows, where i t The implementation method and f t Same, represents a gate control unit, indicating whether the learned memory needs to be Updated to long-term memory, therefore, it is also called the learning gate.

[0132] i t =σ(W i ·[h t-1 ,x t ]+bi )

[0133] In addition, learned memory It can be expressed as the following formula:

[0134]

[0135] In this algorithm, this part indicates whether to record the learned relationship between the wireless network performance indicator characteristics and latency into the long-term memory.

[0136] Finally, the output of the LSTM unit, that is, the predicted delay at each moment in this algorithm, can be expressed as:

[0137] h t =o t ×tanh(C t )

[0138] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0139] The output is composed of the output gate (the implementation principle is similar to the forget gate, because it controls the output result, it is called the output gate) and the memory C at this moment t Jointly determined, that is, the output at this moment is determined by the output feature x at the current moment t And the output feature h at the previous moment t-1 Joint decision.

[0140] In summary, the LSTM model can learn the impact of wireless performance indicators at different times on latency based on data characteristics (i.e., long-term memory). It can also train the impact of wireless network performance indicators at the current moment on latency. Therefore, the LSTM model can effectively learn the cumulative effects of different wireless network performance indicators over time.

[0141] The purpose of the delay evaluation model in the embodiment of the present application is to predict the delay value at the next moment, which is a typical time series prediction problem. Therefore, the delay evaluation model can be constructed using the LSTM model.

[0142] The LSTM model is applied to the embodiments of this application to construct and train a latency assessment model. The input data for the LSTM layer is: the index values of M wireless network performance indicators within a time window, that is, the index values of M wireless network performance indicators within n seconds. For example, if n is 10, then the index values of M wireless network performance indicators at the 1st second, the index values of M wireless network performance indicators at the 2nd second, ..., and the index values of M wireless network performance indicators at the 10th second. The data format is a 10*M data matrix, where M represents the number of features per second, that is, the number of wireless network performance indicators. The output data of the LSTM model is the prediction results for each of the 10 seconds, and the data format is a 10*1 data matrix. The attention layer is equivalent to a feature-enhanced network. Our final prediction result is a weighted sum of the 10 seconds of data from the LSTM layer. The function of the attention layer is to re-weight the 10 data points, increasing the weight of factors affecting the final result. The final output of the latency assessment model is the latency value at the 10th second.

[0143] Multiple sets of index values of M wireless network performance indicators in continuous time windows are input into the delay evaluation model, and the delay evaluation model outputs a delay value sequence of the wireless network (such as Figure 11 The latency distribution of wireless networks is calculated by statistically analyzing the latency distribution. The latency distribution includes the mean latency and the latency tail index. (The latency tail index is the probability of a latency value exceeding a fixed threshold. For example, if the proportion of latency data with a latency value greater than 40ms in the total latency data is 5%, the tail index is 5%).

[0144] A wireless network planning method in an embodiment of the present application constructs a decision tree model and a latency assessment model, using existing CHR wireless network performance indicator data as input features. This method plans the wireless network based on SLA guarantee indicators without modifying existing network nodes and data packets, thus achieving low-cost wireless network planning. Furthermore, given the universality of the algorithm's input wireless network performance indicator data, this solution has a wide range of application scenarios. Furthermore, this solution also outputs latency status and distribution, which can be used to analyze the statistical characteristics of latency in wireless networks, improving the solution's scalability.

[0145] It should be understood that the size of the serial numbers of the above steps does not mean the order of execution. The order of execution of each step should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present application. For example, step 902 can be executed before step 901, that is, the length of the input feature time window is determined first, and then data cleaning is performed.

[0146] Combined with the above Figures 1 to 11 , describes the wireless network planning method and wireless network delay evaluation method in detail, and will be combined with Figures 12 to 14 , a wireless network planning device and a delay evaluation device according to an embodiment of the present application are described in detail.

[0147] Figure 12 This is a structural diagram of the wireless network planning device 22 provided in the embodiment of the present application. Figure 12 As shown, the wireless network planning device 22 includes:

[0148] An acquisition module 221 is configured to acquire multiple groups of indicator values of M wireless network performance indicators of a wireless network; wherein M is a positive integer, and one group of indicator values in the multiple groups of indicator values is a value of each indicator in the M wireless network performance indicators corresponding to multiple moments in a time window;

[0149] a processing module 222 configured to determine, based on the multiple groups of indicator values and the delay status corresponding to each group of indicator values in the multiple groups of indicator values, a correlation between each indicator in the M wireless network performance indicators and the delay status; wherein the delay status corresponding to any one group of indicator values in the multiple groups of indicator values is a state that meets the target delay requirement or a state that does not meet the target delay requirement;

[0150] A determination module 223 is configured to determine N wireless network performance indicators from the M wireless network performance indicators, wherein the correlation of each wireless network performance indicator in the N wireless network performance indicators meets the requirement, N is a positive integer, and N is less than or equal to M;

[0151] The planning module 224 determines a value range of each of the N wireless network performance indicators according to the delay status to plan the wireless network.

[0152] In one example, the processing module is specifically configured to:

[0153] Training a first network model using a plurality of first training sample pairs, wherein each of the plurality of first training sample pairs includes a set of indicator values from the plurality of sets of indicator values and a delay state corresponding to the set of indicator values;

[0154] Determine the Gini impurity between each of the M wireless network performance indicators and the delay state based on the trained first network model; wherein the Gini impurity represents the correlation.

[0155] In one example, the determining module is specifically configured to:

[0156] The N wireless network performance indicators are determined according to the Gini impurity; wherein the Gini impurity between each indicator of the N wireless network performance indicators and the delay state is greater than or equal to a threshold.

[0157] In one example, the planning module is specifically configured to:

[0158] Determine a value range for each indicator in the N wireless network performance indicators based on at least one set of indicator values that meets the standard among the multiple sets of indicator values, wherein the delay state corresponding to the at least one set of indicator values that meets the standard is a state that meets the target delay requirement.

[0159] In one example, the wireless network planning apparatus further includes:

[0160] a delay state determination module, configured to input each group of indicator values in the plurality of groups of indicator values into a delay evaluation model of the wireless network to obtain a delay value corresponding to each group of indicator values;

[0161] The delay state corresponding to each group of indicator values is determined according to whether the delay value corresponding to each group of indicator values meets the target delay requirement. If so, the delay state is a state that does not meet the target delay requirement; if not, the delay state is a state that meets the target delay requirement.

[0162] In another example, the delay evaluation model is a second network model trained by multiple second training sample pairs, wherein one second training sample pair among the multiple second training sample pairs is composed of the indicator value of each indicator of M wireless network performance indicators in one time window among multiple time windows and the delay value corresponding to the one time window.

[0163] In another example, the second network model is an LSTM model.

[0164] In another example, the M wireless network performance indicators include at least one or more of: a coverage status indicator, a retransmission error indicator, a channel scheduling indicator, a cell switching indicator, a distance indicator from a base station, and a congestion index indicator.

[0165] The functions of each functional module in the device can be Figures 1 to 11 Therefore, the specific working process of the device provided in the embodiment of the present application will not be repeated here.

[0166] Figure 13 This is a structural diagram of the delay evaluation device 21 provided in the embodiment of the present application. Figure 13 As shown, the delay evaluation device 21 includes:

[0167] An acquisition module 211 is configured to acquire multiple groups of indicator values of M wireless network performance indicators of the wireless network; wherein M is a positive integer, and one group of indicator values in the multiple groups of indicator values is a value of each indicator in the M wireless network performance indicators corresponding to multiple moments in a time window;

[0168] a model generation module 212 configured to generate a delay evaluation model, wherein the delay evaluation model is a second network model trained by a plurality of second training sample pairs, wherein one of the plurality of second training sample pairs is composed of an indicator value of each of M wireless network performance indicators within one of a plurality of time windows and a delay value corresponding to the time window;

[0169] An input module 213 is configured to input each of the multiple groups of indicator values into a wireless network delay evaluation model to obtain a delay value corresponding to each group of indicator values;

[0170] The output module 214 is configured to output the delay value corresponding to each group of indicator values obtained by the delay evaluation model.

[0171] In one example, the second network model is an LSTM model.

[0172] In an example, the length of the time window ranges from 10 to 50 seconds.

[0173] In one example, the M wireless network performance indicators are determined based on expert experience and data statistical analysis.

[0174] Specifically, the M wireless network performance indicators include at least one or more of a coverage status indicator, a retransmission error indicator, a channel scheduling indicator, a cell switching indicator, a distance indicator from a base station, and a congestion index indicator.

[0175] In one example, wireless network performance indicator marking data corresponding to M wireless network performance indicators in a CHR of a base station network management system of a wireless network is collected; and a value of each indicator in the M wireless network performance indicators is obtained according to the marking data.

[0176] The functions of each functional module in the device can be Figures 1 to 11 Therefore, the specific working process of the device provided in the embodiment of the present application will not be repeated here.

[0177] It should be explained that the wireless network planning method and wireless network planning device, wireless network delay evaluation method and delay evaluation device provided in the embodiments of the present application can be applied to heterogeneous wireless networks or non-heterogeneous wireless networks.

[0178] The present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed in a computer, the computer is caused to execute any one of the above methods.

[0179] The present application also provides a computer program or a computer program product, which includes instructions. When the instructions are executed, the computer is caused to perform any of the above methods.

[0180] Figure 14 This is a schematic diagram of the structure of the computing device provided in this application. The computing device 140 can be the wireless network planning device or wireless network delay evaluation device described above, and the computing device 140 can include a processor 142, a memory 143, and a bus 141. The processor 142 and memory 143 in the device 140 can establish a communication connection via the bus 141.

[0181] The processor 142 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0182] In addition to the data bus, the bus 141 may also include a power bus, a control bus, and a status signal bus. Figure 14 In the text, various buses are marked as buses.

[0183] The memory 143 may include a volatile memory, such as a random-access memory (RAM); the memory 143 may also include a non-volatile memory, such as a read-only memory (ROM), a flash memory, a hard disk drive (HDD), or a solid state drive (SSD); the memory 143 may also include a combination of the above types of memory.

[0184] Processor 142 is used to couple with memory 143, and read and execute instructions in memory 143; when processor 142 is running, it executes the above instructions, so that processor 142 implements the functions of the above wireless network planning device and / or delay evaluation device, and the above wireless network planning method and / or delay evaluation method.

[0185] It is understandable that Figure 14 Only a simplified design of the computing device is shown. In actual applications, the computing device may also include any number of processors, controllers, display screens, memories, etc.

[0186] Those skilled in the art should further appreciate that the modules, units, and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0187] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0188] The specific implementation methods described above further illustrate the purpose, technical solutions and beneficial effects of this application. It should be understood that the above description is only the specific implementation methods of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of this application should be included in the scope of protection of this application.

Claims

1. A wireless network planning method, characterized in that: include: Obtain multiple groups of indicator values of M wireless network performance indicators of the wireless network; wherein M is a positive integer, and one group of indicator values in the multiple groups of indicator values is a value of each indicator in the M wireless network performance indicators corresponding to multiple moments in a time window; Determining, based on the multiple groups of indicator values and the delay states corresponding to each group of indicator values in the multiple groups of indicator values, a correlation between each indicator in the M wireless network performance indicators and the delay states; wherein the delay state corresponding to any one group of indicator values in the multiple groups of indicator values is a state that meets the target delay requirement or a state that does not meet the target delay requirement; Determining N wireless network performance indicators from the M wireless network performance indicators, wherein the correlation of each wireless network performance indicator in the N wireless network performance indicators meets the requirement, and N is a positive integer and is less than or equal to M; A value range of each of the N wireless network performance indicators is determined according to the delay state to plan the wireless network.

2. The method according to claim 1, characterized in that The determining, based on the multiple groups of indicator values and the delay state corresponding to each group of indicator values in the multiple groups of indicator values, a correlation between each indicator in the M wireless network performance indicators and the delay state includes: Training a first network model using a plurality of first training sample pairs, wherein each of the plurality of first training sample pairs includes a set of indicator values from the plurality of sets of indicator values and a delay state corresponding to the set of indicator values; Determine the Gini impurity between each of the M wireless network performance indicators and the delay state based on the trained first network model; wherein the Gini impurity represents the correlation.

3. The method according to claim 2, characterized in that Determining N wireless network performance indicators from the M wireless network performance indicators, wherein the correlation of each wireless network performance indicator in the N wireless network performance indicators meets a requirement, includes: The N wireless network performance indicators are determined according to the Gini impurity; wherein the Gini impurity between each indicator of the N wireless network performance indicators and the delay state is greater than or equal to a threshold.

4. The method according to claim 1, wherein The determining, according to the delay state, a value range of each of the N wireless network performance indicators includes: Determine a value range for each indicator in the N wireless network performance indicators based on at least one set of indicator values that meets the standard among the multiple sets of indicator values, wherein the delay state corresponding to the at least one set of indicator values that meets the standard is a state that meets the target delay requirement.

5. The method according to claim 1, wherein The method further comprises: Inputting each group of indicator values in the multiple groups of indicator values into the delay evaluation model of the wireless network to obtain a delay value corresponding to each group of indicator values; The delay state corresponding to each group of indicator values is determined according to whether the delay value corresponding to each group of indicator values meets the target delay requirement. If so, the delay state is a state that meets the target delay requirement; or if not, the delay state is a state that does not meet the target delay requirement.

6. The method according to claim 5, characterized in that The delay evaluation model is a second network model trained by multiple second training sample pairs, wherein one second training sample pair among the multiple second training sample pairs is composed of an indicator value of each indicator of M wireless network performance indicators in one time window among multiple time windows and a delay value corresponding to the one time window.

7. The method according to claim 6, characterized in that The second network model is an LSTM model.

8. The method according to any one of claims 1 to 7, characterized in that The M wireless network performance indicators include at least one or more of a coverage status indicator, a retransmission error indicator, a channel scheduling indicator, a cell switching indicator, a distance indicator from a base station, and a congestion index indicator.

9. A wireless network planning device, characterized in that: include: an acquisition module, configured to acquire multiple groups of indicator values of M wireless network performance indicators of the wireless network; wherein M is a positive integer, and one group of indicator values in the multiple groups of indicator values is a value of each indicator in the M wireless network performance indicators corresponding to multiple moments in a time window; a processing module, configured to determine, based on the multiple groups of indicator values and the delay status corresponding to each group of indicator values in the multiple groups of indicator values, a correlation between each indicator in the M wireless network performance indicators and the delay status; wherein the delay status corresponding to any one group of indicator values in the multiple groups of indicator values is a state that meets the target delay requirement or a state that does not meet the target delay requirement; a determination module, determining N wireless network performance indicators from the M wireless network performance indicators, wherein the correlation of each wireless network performance indicator in the N wireless network performance indicators meets the requirement, and N is a positive integer and is less than or equal to M; A planning module determines a value range of each of the N wireless network performance indicators according to the delay state to plan the wireless network.

10. The device according to claim 9, characterized in that The processing module is specifically used for: Training a first network model using a plurality of first training sample pairs, wherein each of the plurality of first training sample pairs includes a set of indicator values from the plurality of sets of indicator values and a delay state corresponding to the set of indicator values; Determine the Gini impurity between each of the M wireless network performance indicators and the delay state based on the trained first network model; wherein the Gini impurity represents the correlation.

11. The device according to claim 10, characterized in that The determining module is specifically configured to: The N wireless network performance indicators are determined according to the Gini impurity; wherein the Gini impurity between each indicator of the N wireless network performance indicators and the delay state is greater than or equal to a threshold.

12. The device according to claim 9, characterized in that The planning module is specifically used for: Determine a value range for each indicator in the N wireless network performance indicators based on at least one set of indicator values that meets the standard among the multiple sets of indicator values, wherein the delay state corresponding to the at least one set of indicator values that meets the standard is a state that meets the target delay requirement.

13. The device according to claim 9, characterized in that The device further comprises: a delay state determination module, configured to input each group of indicator values in the plurality of groups of indicator values into a delay evaluation model of the wireless network to obtain a delay value corresponding to each group of indicator values; The delay state corresponding to each group of indicator values is determined according to whether the delay value corresponding to each group of indicator values meets the target delay requirement. If so, the delay state is a state that meets the target delay requirement; if not, the delay state is a state that does not meet the target delay requirement.

14. The device according to claim 13, characterized in that The delay evaluation model is a second network model trained by multiple second training sample pairs, wherein one second training sample pair among the multiple second training sample pairs is composed of an indicator value of each indicator of M wireless network performance indicators in one time window among multiple time windows and a delay value corresponding to the one time window.

15. The device according to claim 14, characterized in that The second network model is an LSTM model.

16. The device according to any one of claims 9 to 15, characterized in that The M wireless network performance indicators include at least one or more of a coverage status indicator, a retransmission error indicator, a channel scheduling indicator, a cell switching indicator, a distance indicator from a base station, and a congestion index indicator.

17. A wireless network planning device, comprising a memory and a processor, characterized in that: The memory stores executable code, and the processor executes the executable code to implement the method according to any one of claims 1 to 8.

18. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed in a computer, the computer is caused to execute the method according to any one of claims 1 to 8.

19. A computer program product, characterized in that The computer program product comprises instructions, and when the instructions are executed, the method according to any one of claims 1 to 8 is implemented.

Images