CN116709394A

CN116709394A - Method, device, computer equipment and storage medium for determining quality of dummy network element

Info

Publication number: CN116709394A
Application number: CN202310677383.5A
Authority: CN
Inventors: 王宇晖; 陈洋; 王智; 柯姣; 黄子杰; 习钰; 陈雯; 刘晶; 李小玄; 孙成丽
Original assignee: China Telecom Corp Ltd
Current assignee: China Telecom Corp Ltd
Priority date: 2023-06-08
Filing date: 2023-06-08
Publication date: 2023-09-05

Abstract

The application discloses a method, a device, computer equipment and a storage medium for determining the quality of a dummy network element. The method can be applied to the technical fields of wireless communication and terminal application, and specifically comprises the following steps: constructing a target RSRP distribution curve graph of the to-be-tested dummy network element according to the reference signal receiving power RSRP of the to-be-tested dummy network element reported by the user terminal in the test period; inputting the target RSRP distribution curve graph into a dummy network element classification model to obtain a target classification result of the dummy network element to be detected; and determining the quality of the to-be-detected dummy network element according to the target classification result. According to the scheme, the quality of the to-be-detected dummy network element can be accurately determined according to the Reference Signal Received Power (RSRP) of the to-be-detected dummy network element reported by the terminal, so that the dummy network element with poor quality can be easily found, and the network quality of the dummy network element can be ensured.

Description

Method, device, computer equipment and storage medium for determining quality of dummy network element

Technical Field

The application relates to the technical field of wireless communication and terminal application, in particular to a method, a device, computer equipment and a storage medium for determining the quality of a dummy network element.

Background

A room item is an item that performs coverage, optimization, and distribution of wireless signals within a building. Displaying according to the query related public information: in order to improve network service quality, a wireless telecommunication company is generally involved in arranging a series of base stations and antennas in a building or a certain area, and connecting the base stations and antennas by wired or wireless means, so as to realize coverage and optimization of services such as telephone, short message, data transmission and the like in the area. The indoor division project can solve the problem of poor signal coverage in the building, improves the communication quality and speed, improves the stability and wireless coverage of mobile phone signals, and is particularly suitable for indoor environments such as large malls, hotels, hospitals, subways and the like.

However, along with the wide application of large-capacity indoor division projects, the application of mobile dummy network elements is more and more extensive, especially in high-density crowd areas such as residential houses, large commercial buildings, business office buildings, movie theatres and the like, and the network quality of the dummy network elements is difficult to guarantee due to the characteristics of high design complexity, more devices and complex networking modes of the areas.

Disclosure of Invention

Based on this, it is necessary to provide a method, an apparatus, a computer device and a storage medium for determining quality of a dummy network element according to the reference signal received power RSRP of the dummy network element to be measured reported by a terminal.

In a first aspect, the present application provides a method for determining quality of a dummy network element, where the method includes:

constructing a target RSRP distribution curve graph of the to-be-tested dummy network element according to the reference signal receiving power RSRP of the to-be-tested dummy network element reported by the user terminal in the test period;

inputting the target RSRP distribution curve graph into a dummy network element classification model to obtain a target classification result of the dummy network element to be detected;

and determining the quality of the to-be-detected dummy network element according to the target classification result.

In one embodiment, the dummy network element classification model is trained by:

acquiring a sample RSRP distribution curve graph and a sample labeling classification result of the sample RSRP distribution curve graph of each sample dummy network element;

and training the convolutional neural network by adopting each sample RSRP distribution curve graph and sample labeling classification results of each sample RSRP distribution curve graph to obtain a dummy network element classification model.

In one embodiment, obtaining a sample annotation classification result of each sample RSRP distribution graph includes:

clustering the RSRP distribution graphs of each sample to obtain a plurality of distribution graph classes;

and uniformly labeling each distribution curve graph class to obtain a sample labeling classification result of each sample RSRP distribution curve graph.

In one embodiment, clustering the RSRP distribution graphs of each sample to obtain a plurality of distribution graph classes includes:

and clustering the RSRP distribution curves of the samples according to the curve distribution similarity of the RSRP distribution curves of the samples to obtain a plurality of distribution curve graphs.

In one embodiment, the convolutional neural network comprises an input layer, a processing layer, a fully connected layer, a linear weighting layer and a loss function layer;

training the convolutional neural network by adopting each sample RSRP distribution curve graph and sample labeling classification results of each sample RSRP distribution curve graph to obtain a dummy network element classification model, wherein the method comprises the following steps:

inputting the RSRP distribution curve graphs of the samples into a processing layer through an input layer for processing to obtain characteristic representation of the RSRP distribution curve graphs of the samples;

respectively carrying out nonlinear conversion on each characteristic representation through the full connection layer to obtain conversion characteristics of each characteristic representation;

respectively carrying out linear weighting on each conversion characteristic through a linear weighting layer to obtain a sample prediction classification result of each sample RSRP distribution curve graph;

determining training loss according to a sample prediction classification result and a sample labeling classification result of each sample RSRP distribution curve graph through a loss function layer;

Training the convolutional neural network by adopting training loss to obtain a dummy network element classification model.

In one embodiment, the processing layers include at least two convolutional pooling layers connected end to end, each comprising one convolutional layer and one pooling layer.

In one embodiment, determining the quality of the to-be-detected dummy network element according to the target classification result includes:

and determining the quality of the to-be-detected dummy network element according to the target classification result based on the corresponding relation between the classification result and the quality.

In a second aspect, the present application further provides a quality determining device for a dummy network element, where the device includes:

the curve graph construction module is used for constructing a target RSRP distribution curve graph of the to-be-tested dummy network element according to the reference signal receiving power RSRP of the to-be-tested dummy network element reported by the user terminal in the test period;

the classification determining module is used for inputting the target RSRP distribution curve graph into the dummy network element classification model to obtain a target classification result of the dummy network element to be detected;

and the quality determining module is used for determining the quality of the to-be-detected dummy network element according to the target classification result.

In a third aspect, the present application also provides a computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the following steps when executing the computer program:

In a fourth aspect, the present application also provides a computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of:

In a fifth aspect, the application also provides a computer program product comprising a computer program which, when executed by a processor, performs the steps of:

According to the method, the device, the computer equipment and the storage medium for determining the quality of the dummy network element, in the period with the test requirement of the dummy network element, the RSRP distribution curve graph drawn according to the RSRP of the dummy network element to be tested, reported by the user terminal in the period, is input into the dummy network element classification model to obtain the target classification result of the dummy network element to be tested, and then the quality of the dummy network element to be tested can be determined according to the target classification result. The method for determining the quality of the dummy network element can accurately determine the quality of the dummy network element to be detected according to the RSRP of the dummy network element to be detected, which is reported by the terminal, so that the dummy network element with poor quality can be easily found, and the network quality of the dummy network element can be ensured.

Drawings

Fig. 1 is an application environment diagram of a method for determining quality of a dummy network element in one embodiment;

fig. 2 is a flow chart of a method for determining quality of a dummy network element in one embodiment;

fig. 3 is a schematic flow chart of obtaining a dummy network element classification model in one embodiment;

FIG. 4 is a flow chart of sample labeling classification results of obtaining RSRP distribution graphs of samples according to an embodiment;

FIGS. 5 (a) -5 (c) are graphs of sample RSRP profiles for three curve profiles in one embodiment;

FIG. 6 is a schematic diagram of a convolutional neural network in one embodiment;

FIG. 7 is a schematic flow chart of training a convolutional neural network in one embodiment;

fig. 8 is a flow chart of a method for determining quality of a dummy network element in another embodiment;

fig. 9 is a block diagram of a quality determining apparatus of a dummy network element in one embodiment;

fig. 10 is a block diagram of a quality determining apparatus for a dummy network element according to another embodiment;

FIG. 11 is an internal block diagram of a computer device in one embodiment.

Detailed Description

The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.

At present, the prior art for judging the abnormal dummy network element by using MR (Measurement Report ) data in China is more, but the abnormal screening of the mobile network dummy network element is lacking. The existing method for monitoring the mobile network dummy network element is mainly realized through the traditional network drive test and the traditional machine learning method. The traditional drive test method generally uses a special test system to collect network related data such as signal intensity, signal quality and the like of a test area through drive test, and the collected data is manually replayed and analyzed in the later period; in the traditional machine learning method, algorithms such as decision trees and the like are adopted to mark MR data information and construct a classification model, and quality difference cells or road sections are screened out, so that abnormal dummy network elements are screened out.

However, with the wide application of large-capacity indoor division projects, the application of the mobile network element is more and more extensive, especially in high-density crowd areas such as residential houses, large commercial buildings, business offices, movie theaters and the like, due to the characteristics of high design complexity, multiple devices and complex networking modes of the areas, the hidden faults of the indoor division projects are often difficult to find, and the network quality of the mobile network element is also difficult to ensure. Based on the above, the embodiment of the application provides a method for determining the quality of a dummy network element, so as to improve the technical problems.

The method for determining the quality of the dummy network element provided by the embodiment of the application can be applied to an application environment shown in fig. 1. In one embodiment, a computer device is provided, which may be a server, the internal structure of which may be as shown in FIG. 1. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is used for storing the acquired data of the quality determination of the dummy network element. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program when executed by a processor implements a method of determining quality of a dummy network element.

The application discloses a method, a device, computer equipment and a storage medium for determining the quality of a dummy network element. The computer equipment obtains a target classification result of the to-be-tested dummy network element by constructing a target RSRP distribution curve graph of the to-be-tested dummy network element and inputting the target RSRP distribution curve graph into the dummy network element classification model, so that the quality of the to-be-tested dummy network element is determined according to the target classification result.

In one embodiment, fig. 2 is a flow chart of a method for determining quality of a dummy network element according to an embodiment of the present application, and is described by taking the application of the method to the server 102 in fig. 1 as an example, the method includes the following steps:

s201, constructing a target RSRP distribution curve graph of the to-be-tested dummy network element according to reference signal received power RSRP of the to-be-tested dummy network element reported by the user terminal in the test period.

Alternatively, the test period may be a period in which the quality of the dummy network element to be tested needs to be tested. In one embodiment, the to-be-measured dummy network element may transmit the reference signal outwards, and the user terminal may report RSRP (Reference Signal Received Power ) to the to-be-measured dummy network element after receiving the reference signal. The reference signal received power RSRP is a power average value of the received reference signal reported by the ue.

Specifically, the user terminal reports RSRP of the to-be-tested dummy network element in the test period, counts the reported value and quantity of the RSRP of the to-be-tested dummy network element, takes the RSRP value as a horizontal axis (the value range is 0-145), takes the quantity of the RSRP as a vertical axis, and utilizes the python drawing function to construct a target RSRP distribution curve graph of the to-be-tested dummy network element.

S202, inputting the target RSRP distribution curve graph into a dummy network element classification model to obtain a target classification result of the dummy network element to be detected.

Alternatively, the dummy network element classification model may be a trained convolutional neural network model capable of classifying the target RSRP distribution graph.

Specifically, the target RSRP distribution curve graph is input into a dummy network element classification model, and the target classification result of the dummy network element to be detected is determined by utilizing the classification function of the dummy network element classification model.

And S203, determining the quality of the to-be-detected dummy network element according to the target classification result.

Optionally, a table of correspondence between the classification result and the quality is preset, and after the target classification result is obtained, the quality of the dummy network element corresponding to the target classification result can be determined according to the table. Specifically, based on the corresponding relation between the classification result and the quality, the quality of the to-be-detected dummy network element is determined according to the target classification result. For example, in the table of the correspondence between the classification result and the quality, the quality corresponding to the classification result 1 is excellent, the quality corresponding to the classification result 2 is good, and the quality corresponding to the classification result 3 is poor. When the obtained target classification result corresponding to the target RSRP distribution curve graph is the classification result 2, the quality of the to-be-detected dummy network element can be determined to be good.

According to the method for determining the quality of the dummy network element, in the period with the test requirement of the dummy network element, the RSRP distribution curve graph drawn according to the reference signal received power RSRP of the dummy network element to be tested, reported by the user terminal in the period, is input into the dummy network element classification model to obtain the target classification result of the dummy network element to be tested, and then the quality of the dummy network element to be tested can be determined according to the target classification result. The method for determining the quality of the dummy network element can accurately determine the quality of the dummy network element to be detected according to the Reference Signal Received Power (RSRP) of the dummy network element to be detected, which is reported by the terminal, so that the dummy network element with poor quality can be easily found, and the network quality of the dummy network element can be ensured.

In order to explain the procedure of obtaining the dummy network element classification model in the above embodiment. A flow diagram of the training process of the dummy network element classification model in one embodiment is shown in fig. 3. Optionally, as shown in fig. 3, the following implementation process is included:

s301, a sample RSRP distribution curve graph and a sample label classification result of the sample RSRP distribution curve graph of each sample dummy network element are obtained.

Optionally, the sample RSRP distribution graph of each sample dummy network element may be constructed according to the reference signal received power RSRP of each sample dummy network element reported by the user terminal; the sample labeling and classifying result can be a result obtained by performing clustering operation and labeling operation on the sample RSRP distribution curve graph, and specifically can be a sample label of the sample RSRP distribution curve graph.

Specifically, a part of RSRP distribution curves are selected from the constructed RSRP distribution curves to serve as sample RSRP distribution curves, and the sample RSRP distribution curves are subjected to clustering marking to obtain sample marking classification results of the sample RSRP distribution curves.

S302, training a convolutional neural network by adopting each sample RSRP distribution curve graph and sample labeling classification results of each sample RSRP distribution curve graph to obtain a dummy network element classification model.

Specifically, inputting each sample RSRP distribution curve graph into a convolutional neural network for training to obtain a sample prediction classification result of each sample RSRP distribution curve graph, and training loss according to the sample prediction classification result of each sample RSRP distribution curve graph and a sample labeling classification result of each sample RSRP distribution curve graph; when the training loss reaches a preset loss threshold value, the obtained trained convolutional neural network is the dummy network element classification model.

It can be understood that in this embodiment, a sample RSRP distribution curve graph of a sample dummy network element and a sample prediction classification result of the sample RSRP distribution curve graph are introduced, and a dummy network element classification model can be obtained by training a convolutional neural network, so as to lay a foundation for finally determining the quality of the dummy network element.

On the basis of the embodiment, the step of obtaining the sample labeling classification result of each sample RSRP distribution curve graph is decomposed and refined. Optionally, as shown in fig. 4, the following implementation process is included:

s401, clustering the RSRP distribution graphs of the samples to obtain a plurality of distribution graph classes.

Alternatively, the clustering operation may be an operation of classifying the RSRP distribution graphs of the samples, and each classification of the RSRP distribution graphs of the samples obtained is regarded as a plurality of distribution graph classes.

In the embodiment of the present application, clustering of the RSRP distribution graphs of each sample is performed according to the curve distribution of the RSRP distribution graph of each sample, and the RSRP distribution graph of the sample with similar curve distribution is used as a distribution graph class. Specifically, according to the curve distribution similarity of the RSRP distribution curves of each sample, clustering the RSRP distribution curves of each sample to obtain a plurality of distribution curve graphs. Furthermore, the clustering process in the embodiment of the application uses a k-means clustering algorithm, that is, a set of RSRP distribution curve graphs of one sample and a plurality of preset clustering centers are given, then the distance between the distribution average value of each RSRP distribution curve graph and each clustering center is calculated, and each RSRP distribution curve graph is distributed to the closest clustering center. The cluster centers and the RSRP profiles assigned to them represent a profile class. The clustering calculation process is shown in formula (1), wherein p is a characteristic point of each sample RSRP distribution curve graph in space, ci is a clustering center, and k value in the embodiment of the application is preferably 3, namely, the k value is divided into three distribution curve graph types.

For example, the embodiment of the present application describes a clustering operation by taking the final classification into three distribution graph classes as an example. Fig. 5 (a) -5 (c) show sample RSRP profiles for three curve profiles in one embodiment. As shown in fig. 5 (a), the corresponding RSRP curve distribution is more right, and the data is concentrated between-85-65; as shown in FIG. 5 (b), the corresponding RSRP curve distribution is biased towards normal distribution, and the data set is between-105-85; as shown in FIG. 5 (c), the corresponding RSRP curve distribution trend is more left, and the data is concentrated between-120-105.

And S402, uniformly labeling each distribution curve graph class to obtain a sample labeling classification result of each sample RSRP distribution curve graph.

Specifically, after obtaining a plurality of distribution curve graph classes, uniformly marking the same mark for each distribution curve graph class, and obtaining a sample marking and classifying result of each sample RSRP distribution curve graph. For example, as shown in fig. 5 (a), the data is concentrated between-85 and-65, the quality of the corresponding dummy network element is excellent, and the rank 0 sample labels are uniformly labeled, so as to obtain the sample labeling classification result of each sample RSRP distribution curve chart in the distribution curve chart class; as shown in fig. 5 (b), the quality of the corresponding dummy network element is good between-105 and-85 in the data set, and the rank 1 sample labels are uniformly marked so as to obtain sample marking classification results of each sample RSRP distribution curve chart in the distribution curve chart class; as shown in fig. 5 (c), the data are concentrated between-120 and-105, the quality of the corresponding dummy network element is poor, and the rank 2 sample labels are uniformly labeled, so as to obtain the sample labeling classification result of each sample RSRP distribution curve chart in the distribution curve chart class. And finally obtaining sample labeling classification results of the RSRP distribution graphs of all the samples.

It can be understood that in this embodiment, after the RSRP distribution graphs of each sample are clustered into multiple distribution graph classes, each distribution graph class is marked uniformly, so that repeated marking of a large amount of data can be avoided, time and labor are saved, and a foundation is laid for determining the quality of the final GBA dummy network element.

On the basis of the embodiment, the step of training the convolutional neural network is decomposed and refined. The convolutional neural network in the embodiment of the application comprises an input layer, a processing layer, a full connection layer, a linear weighting layer and a loss function layer, wherein each processing layer comprises at least two end-to-end convolutional pooling layers, and each convolutional pooling layer comprises one convolutional layer and one pooling layer. The convolution neural network is a lightweight convolution neural network obtained by structurally changing the general convolution neural network, and the calculation amount is reduced by changing the network structure through the lightweight convolution neural network.

A schematic structural diagram of a convolutional neural network in an embodiment of the present application is shown in fig. 6. As shown in fig. 6, the convolutional neural network includes an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, a third pooling layer, a full-connection layer, a linear weighting layer, and a loss function layer. The input and output of the input layer are all pixel matrixes of 32 x 3 converted from a sample RSRP distribution curve graph, the input of each subsequent layer is the output of the previous layer, and the convolution layer is used for carrying out convolution operation by using a preset convolution check input pixel matrix; the pooling layer is used for pooling the input pixel matrix with a preset pooling window; the full connection layer is used for carrying out nonlinear conversion on each characteristic representation which is output and converted by the third pooling layer to obtain conversion characteristics of each characteristic representation; the full connection layer is used for respectively carrying out nonlinear conversion on each characteristic representation to obtain conversion characteristics of each characteristic representation; the loss function layer is used for determining training loss according to the sample prediction classification result and the sample labeling classification result of each sample RSRP distribution curve graph. After the training loss is determined, the convolutional neural network can be trained according to the training loss, so that a dummy network element classification model is obtained.

On the basis of the convolutional neural network structure, a flow chart of a training process of the convolutional neural network in the embodiment of the application is shown in fig. 7. Training the convolutional neural network, as described in claim 7, comprises the following implementation process:

s701, inputting the RSRP distribution curve graph of each sample into a processing layer through an input layer for processing, and obtaining the characteristic representation of the RSRP distribution curve graph of each sample.

Optionally, the input layer is used for transmitting data, and the input and output of the input layer can be a pixel matrix of 32×32×3 converted from the sample RSRP distribution graph.

Optionally, the input of the first convolution layer of the first processing layer is a 32 x 3 matrix of pixels output from the input layer, wherein the convolution kernel size of the first convolution layer is 3 x 32, and the first convolution layer is filled with all 0 s, the step length is 1, and the output of the first convolution layer is a pixel matrix of 32 x 32.

The input of the first pooling layer of the first processing layer is a 32 x 32 matrix of pixels of the output of the first convolution layer. The size of the pooling window of the first pooling layer is 2×2, the step sizes of the length and the width are both 2, and the output of the first pooling layer is a pixel matrix of 16×16×32.

The input to the second convolution layer of the second processing layer is a 16 x 32 matrix of pixels from the output of the first pooling layer. The convolution kernel size of the second convolution layer is 3×3×64, the second convolution layer is filled with all 0 s, the step size is 1, and the output of the second convolution layer is a pixel matrix of 16×16×64.

The input of the second pooling layer of the second processing layer is a matrix of 16 x 64 pixels of the output of the second convolution layer. The size of the pooling window of the second pooling layer is 2×2, the step sizes of the length and the width are both 2, and the output of the second pooling layer is 8×8×64 pixel matrix.

The input of the third convolution layer of the third processing layer is an 8 x 64 matrix of pixels from the output of the second pooling layer. The convolution kernel size of the third convolution layer is 3×3×128, the third convolution layer is filled with all 0 s, the step size is 1, and the output of the third convolution layer is a pixel matrix of 8×8×128.

The input of the third pooling layer of the third processing layer is an 8 x 128 matrix of pixels output by the third convolution layer. The size of the pooling window of the third pooling layer is 2×2, the step sizes of the length and the width are both 2, and the output of the third pooling layer is a pixel matrix of 4×4×128, namely, the characteristic representation of the RSRP distribution curve graph of the sample.

S702, respectively performing nonlinear conversion on each feature representation through the full connection layer to obtain conversion features of each feature representation.

Optionally, the input of the full connection layer is a 4×4×128 pixel matrix output by the third pooling layer, and the number of output nodes is 512, so as to perform nonlinear conversion on the pixel matrix output by the third pooling layer to output conversion features represented by each feature.

S703, respectively carrying out linear weighting on each conversion characteristic through a linear weighting layer to obtain a sample prediction classification result of each sample RSRP distribution curve graph.

Optionally, the input of the linear weighting layer is a conversion feature represented by each feature output by the full-connection layer, and the number of output nodes is 3, so as to linearly weight each conversion feature of the full-connection layer, and output a sample prediction classification result of each sample RSRP distribution curve graph.

S704, determining training loss according to a sample prediction classification result and a sample labeling classification result of each sample RSRP distribution curve graph through a loss function layer.

Specifically, after the sample prediction classification result of each sample RSRP distribution curve graph is obtained, the sample labeling classification result of each sample RSRP distribution curve graph is combined, and the difference value between the sample prediction classification result and the sample labeling classification result of each sample RSRP distribution curve graph is calculated through a loss function in a loss function layer to be used as training loss.

Further, the loss function used by the loss function layer of the convolutional neural network is a cross entropy loss function, and a calculation formula of the cross entropy loss function is shown as the following formula (2):

wherein n is the number of sample RSRP distribution graphs; p (y) _i |x _i ) Graph x representing sample RSRP distribution _i Classification to y via convolutional neural network _i Probability of (2); y is _i And (5) representing a sample labeling classification result corresponding to the sample RSRP distribution curve graph.

And S705, training the convolutional neural network by adopting training loss to obtain a dummy network element classification model.

Optionally, when the training loss is greater than a preset loss threshold, updating and optimizing the weight parameter and the bias parameter of each layer in the convolutional neural network by adopting a back propagation algorithm to obtain an updated convolutional neural network, and continuing to train the updated convolutional neural network until the training loss is less than or equal to the preset loss threshold, and stopping training, so as to obtain the dummy network element classification model.

Furthermore, in order to speed up the decrease of the loss value to shorten the training time, the back propagation algorithm in the embodiment of the application adopts Adam-Optimizer as an Optimizer. And the initial value of the weight parameter may use a random value of the truncated normal distribution output with a standard deviation of 0.1, and the initial value of the bias parameter may be set to 0.

It should be noted that, in the convolutional neural network training process, in order to introduce nonlinearity into the convolutional neural network to strengthen learning ability of the convolutional neural network, a linear rectification function (Linear rectification function, reLU) is introduced as a neuron activation function, and the ReLU is expressed as the following formula (3):

Where x is a sample RSRP distribution graph.

It can be understood that in this embodiment, a lightweight convolutional neural network is introduced, and each sample RSRP distribution graph is input into the lightweight convolutional neural network, and the convolutional neural network is trained according to the training loss, so as to finally obtain a dummy network element classification model, thereby laying a foundation for final GBA dummy network element quality determination.

On the basis of the embodiment, the sample RSRP distribution graph is divided into a training set and a testing set to train the convolutional neural network, wherein the step of training the convolutional neural network is as follows: sample RSRP profile was plotted at 7:3, dividing the ratio into a training set and a testing set; inputting randomly selected m (with values of 64 to 128) sample RSRP distribution graphs in all sample RSRP distribution graphs in a training set into a convolutional neural network by utilizing the convolutional neural network to train to obtain a trained convolutional neural network; and testing the trained convolutional neural network by using a test set to obtain a test result, comparing the test result with a real result to determine an error, and updating and optimizing the weight parameter and the bias parameter of each layer in the convolutional neural network by using a back propagation algorithm when the error does not meet the requirement to obtain an updated convolutional neural network, and repeating the steps until the error meets the requirement, thereby obtaining the dummy network element classification model.

It can be understood that in this embodiment, the sample RSRP distribution graph is divided into a training set and a testing set, and the training set and the testing set are used to train the convolutional neural network, so as to finally obtain the dummy network element classification model, thereby laying a foundation for final GBA dummy network element quality determination.

In addition, in one embodiment, the present application further provides an optional example of a method for determining quality of a dummy network element, and fig. 8 is a schematic flow chart of a method for determining quality of a dummy network element in another embodiment, and in combination with fig. 8, the method specifically includes the following implementation procedures:

s801, a sample RSRP distribution curve graph of each sample dummy network element is obtained.

S802, clustering the RSRP distribution graphs of the samples according to the curve distribution similarity of the RSRP distribution graphs of the samples to obtain a plurality of distribution graph types.

And S803, uniformly labeling each distribution curve graph class to obtain a sample labeling classification result of each sample RSRP distribution curve graph.

S804, inputting the RSRP distribution curve graph of each sample into a processing layer through an input layer for processing, and obtaining the characteristic representation of the RSRP distribution curve graph of each sample.

And S805, respectively performing nonlinear conversion on each feature representation through the full connection layer to obtain conversion features of each feature representation.

And S806, respectively carrying out linear weighting on each conversion characteristic through a linear weighting layer to obtain a sample prediction classification result of each sample RSRP distribution curve graph.

S807, determining training loss according to the sample prediction classification result and the sample labeling classification result of each sample RSRP distribution curve graph through the loss function layer.

S808, training the convolutional neural network by adopting training loss to obtain a dummy network element classification model.

S809, constructing a target RSRP distribution curve graph of the to-be-tested dummy network element according to the reference signal received power RSRP of the to-be-tested dummy network element reported by the user terminal in the test period.

S810, inputting the target RSRP distribution curve graph into a dummy network element classification model to obtain a target classification result of the dummy network element to be detected.

S811, determining the quality of the to-be-detected dummy network element according to the target classification result based on the corresponding relation between the classification result and the quality.

According to the scheme, the quality of the to-be-detected dummy network element can be accurately determined according to the Reference Signal Received Power (RSRP) of the to-be-detected dummy network element reported by the terminal, so that the dummy network element with poor quality can be easily found, and the network quality of the dummy network element can be ensured.

The specific process of S801 to S811 may refer to the description of the foregoing method embodiment, and its implementation principle and technical effect are similar, and will not be described herein.

It should be understood that, although the steps in the flowcharts related to the embodiments described above are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in the flowcharts described in the above embodiments may include a plurality of steps or a plurality of stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of the steps or stages is not necessarily performed sequentially, but may be performed alternately or alternately with at least some of the other steps or stages.

Based on the same inventive concept, the embodiment of the application also provides a device for determining the quality of the dummy network element, which is used for realizing the method for determining the quality of the dummy network element. The implementation scheme of the solution to the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitation in the embodiments of the device for determining quality of one or more dummy network elements provided below may refer to the limitation of the method for determining quality of a dummy network element hereinabove, and will not be repeated herein.

In one of the embodiments, a block diagram of the quality determining device of the dummy network element in one embodiment is shown by fig. 9. As shown in fig. 9, there is provided a quality determining apparatus 9 for a dummy network element, the apparatus comprising: a graph construction module 90, an input classification module 91, and a determination module 92, wherein:

the graph construction module 90 is configured to construct a target RSRP distribution graph of the to-be-tested dummy network element according to the reference signal received power RSRP of the to-be-tested dummy network element reported by the user terminal in the test period;

an input classification determining module 91, configured to input the target RSRP distribution graph into a dummy network element classification model, to obtain a target classification result of the to-be-tested dummy network element;

and the quality determining module 92 is configured to determine the quality of the to-be-detected dummy network element according to the target classification result.

According to the device for determining the quality of the dummy network element, in the period with the test requirement of the dummy network element, the RSRP distribution curve graph drawn according to the reference signal received power RSRP of the dummy network element to be tested, reported by the user terminal in the period, is input into the dummy network element classification model to obtain the target classification result of the dummy network element to be tested, and then the quality of the dummy network element to be tested can be determined according to the target classification result. The method for determining the quality of the dummy network element can accurately determine the quality of the dummy network element to be detected according to the Reference Signal Received Power (RSRP) of the dummy network element to be detected, which is reported by the terminal, so that the dummy network element with poor quality can be easily found, and the network quality of the dummy network element can be ensured.

In one of the embodiments, a block diagram of the quality determining device of the dummy network element in another embodiment is shown by fig. 10. As shown in fig. 10, the above-mentioned quality determining apparatus 9 for a dummy network element further includes:

the result obtaining module 93 is configured to obtain a sample RSRP distribution curve graph of each sample dummy network element and a sample label classification result of the sample RSRP distribution curve graph;

the network training module 94 is configured to train the convolutional neural network to obtain a dummy network element classification model by using each sample RSRP distribution graph and the sample labeling classification result of each sample RSRP distribution graph.

In one embodiment, the result acquisition module 93 is configured to: acquiring a sample RSRP distribution curve graph of each sample dummy network element; clustering the RSRP distribution graphs of each sample to obtain a plurality of distribution graph classes; and uniformly labeling each distribution curve graph class to obtain a sample labeling classification result of each sample RSRP distribution curve graph.

In one embodiment, the result acquisition module 93 is further configured to: and clustering the RSRP distribution curves of the samples according to the curve distribution similarity of the RSRP distribution curves of the samples to obtain a plurality of distribution curve graphs.

The network training module 94 is configured to: inputting the RSRP distribution curve graphs of the samples into a processing layer through an input layer for processing to obtain characteristic representation of the RSRP distribution curve graphs of the samples; respectively carrying out nonlinear conversion on each characteristic representation through the full connection layer to obtain conversion characteristics of each characteristic representation; respectively carrying out linear weighting on each conversion characteristic through a linear weighting layer to obtain a sample prediction classification result of each sample RSRP distribution curve graph; determining training loss according to a sample prediction classification result and a sample labeling classification result of each sample RSRP distribution curve graph through a loss function layer; training the convolutional neural network by adopting training loss to obtain a dummy network element classification model.

In one embodiment, the quality determination module 92 of FIG. 9 above is specifically configured to: and determining the quality of the to-be-detected dummy network element according to the target classification result based on the corresponding relation between the classification result and the quality.

The modules in the above-mentioned dumb network element quality determining device may be implemented in whole or in part by software, hardware, and a combination thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.

In one embodiment, a computer device is provided, which may be a terminal, and the internal structure of which may be as shown in fig. 11. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is used for storing the dummy network element quality determination data. The network interface of the computer device is used for communicating with an external user side through a network connection. The computer program when executed by a processor implements a method of determining quality of a dummy network element.

It will be appreciated by those skilled in the art that the structure shown in FIG. 11 is merely a block diagram of some of the structures associated with the present inventive arrangements and is not limiting of the computer device to which the present inventive arrangements may be applied, and in particular, a computer device may include more or fewer components than shown, or may combine some of the components, or have a different arrangement of components.

In one embodiment, a computer device is provided, comprising a memory and a processor, the memory storing a computer program, the processor implementing the following steps when executing the computer program:

In one embodiment, when the processor executes logic of a training process of the dummy network element classification model in the computer program, the following steps are specifically implemented:

acquiring a sample RSRP distribution curve graph and a sample labeling classification result of the sample RSRP distribution curve graph of each sample dummy network element; and training the convolutional neural network by adopting each sample RSRP distribution curve graph and sample labeling classification results of each sample RSRP distribution curve graph to obtain a dummy network element classification model.

In one embodiment, when the processor executes logic for obtaining a sample labeling classification result of each sample RSRP distribution graph in the computer program, the following steps are specifically implemented:

Clustering the RSRP distribution graphs of each sample to obtain a plurality of distribution graph classes; and uniformly labeling each distribution curve graph class to obtain a sample labeling classification result of each sample RSRP distribution curve graph.

In one embodiment, when the processor performs clustering on RSRP distribution graphs of each sample in the computer program to obtain logic of a plurality of distribution graph types, the following is specifically implemented:

the processor executes a computer program, adopts each sample RSRP distribution curve graph and sample label classification results of each sample RSRP distribution curve graph, trains the convolutional neural network, and realizes the following specific contents when logic of a dummy network element classification model is obtained:

inputting the RSRP distribution curve graphs of the samples into a processing layer through an input layer for processing to obtain characteristic representation of the RSRP distribution curve graphs of the samples; respectively carrying out nonlinear conversion on each characteristic representation through the full connection layer to obtain conversion characteristics of each characteristic representation; respectively carrying out linear weighting on each conversion characteristic through a linear weighting layer to obtain a sample prediction classification result of each sample RSRP distribution curve graph; determining training loss according to a sample prediction classification result and a sample labeling classification result of each sample RSRP distribution curve graph through a loss function layer; training the convolutional neural network by adopting training loss to obtain a dummy network element classification model.

In one embodiment, the following steps are embodied when the processor executes logic in a computer program:

the processing layer comprises at least two end-to-end convolution pooling layers, each convolution pooling layer comprising a convolution layer and a pooling layer.

In one embodiment, when the processor executes logic in the computer program for determining the quality of the to-be-detected dummy network element according to the target classification result, the following steps are specifically implemented:

The principles and specific processes of implementing the foregoing embodiments of the foregoing computer device may be referred to the description in the foregoing embodiments of the method for determining quality of a dummy network element, which is not described herein.

In one embodiment, a computer readable storage medium is provided having a computer program stored thereon, which when executed by a processor, performs the steps of:

In one embodiment, the logic of the training process of the dummy network element classification model in the computer program, when executed by the processor, specifically implements the following steps:

In one embodiment, the logic for obtaining the sample labeling classification result of each sample RSRP distribution graph in the computer program is executed by the processor, and further specifically implements the following steps:

In one embodiment, the logic for clustering RSRP distribution graphs of each sample in the computer program to obtain a plurality of distribution graph classes is executed by the processor, and specifically implements the following:

the computer program adopts each sample RSRP distribution curve graph and sample label classification result of each sample RSRP distribution curve graph to train the convolutional neural network, and when logic of obtaining a dummy network element classification model is executed by a processor, the following is specifically realized:

In one embodiment, logic in a computer program, when executed by a processor, performs the steps of:

In one embodiment, when the logic for determining the quality of the to-be-detected dummy network element in the computer program is executed by the processor according to the target classification result, the following steps are specifically implemented:

The principles and specific procedures of implementing the foregoing embodiments of the foregoing computer readable storage medium may be referred to the descriptions in the foregoing embodiments of the method for determining quality of a dummy network element, which are not described herein.

The principles and specific procedures of implementing the foregoing embodiments of the present invention in the foregoing embodiments of the target detection method may be referred to in the foregoing embodiments of the present invention, and are not described herein in detail.

In one embodiment, a computer program product is provided comprising a computer program which, when executed by a processor, performs the steps of:

The principles and specific procedures of implementing the foregoing embodiments of the present application in the foregoing embodiments of the target detection method may be referred to in the foregoing embodiments of the present application, and are not described herein in detail.

It should be noted that, the data (including but not limited to the data in the quality determining process of the dummy network element) related to the present application are all data fully authorized by each party, and the collection, use and processing of the related data need to comply with the related laws and regulations and standards of the related country and region.

Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high density embedded nonvolatile Memory, resistive random access Memory (ReRAM), magnetic random access Memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric Memory (Ferroelectric Random Access Memory, FRAM), phase change Memory (Phase Change Memory, PCM), graphene Memory, and the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), and the like. The databases referred to in the embodiments provided herein may include at least one of a relational database and a non-relational database. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processor referred to in the embodiments provided in the present application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, or the like, but is not limited thereto.

The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.

The foregoing examples illustrate only a few embodiments of the application and are described in detail herein without thereby limiting the scope of the application. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of the application should be assessed as that of the appended claims.

Claims

1. A method for determining quality of a dummy network element, the method comprising:

constructing a target RSRP distribution curve graph of the to-be-tested dummy network element according to reference signal receiving power RSRP of the to-be-tested dummy network element reported by a user terminal in a test period;

And determining the quality of the dummy network element to be detected according to the target classification result.

2. The method according to claim 1, wherein the dummy network element classification model is trained by:

acquiring a sample RSRP distribution curve graph of each sample dummy network element and a sample labeling classification result of the sample RSRP distribution curve graph;

3. The method of claim 2, wherein obtaining a sample annotation classification result for each sample RSRP distribution graph comprises:

4. A method according to claim 3, wherein clustering the RSRP distribution profiles of each sample to obtain a plurality of distribution profile classes comprises:

5. The method of claim 2, wherein the convolutional neural network comprises an input layer, a processing layer, a full connection layer, a linear weighting layer, and a loss function layer;

the training of the convolutional neural network to obtain a dummy network element classification model by adopting each sample RSRP distribution curve graph and sample labeling classification results of each sample RSRP distribution curve graph comprises the following steps:

inputting the RSRP distribution curve graphs of the samples into the processing layer through the input layer for processing to obtain characteristic representations of the RSRP distribution curve graphs of the samples;

respectively carrying out linear weighting on each conversion characteristic through the linear weighting layer to obtain a sample prediction classification result of each sample RSRP distribution curve graph;

determining training loss according to a sample prediction classification result and a sample labeling classification result of each sample RSRP distribution curve graph through the loss function layer;

and training the convolutional neural network by adopting the training loss to obtain a dummy network element classification model.

6. The method of claim 5, wherein the processing layers comprise at least two end-to-end convolutional pooling layers, each comprising one convolutional layer and one pooling layer.

7. The method according to claim 1, wherein the determining the quality of the dummy network element to be tested according to the target classification result comprises:

8. A dummy network element quality determining apparatus, the apparatus comprising:

the curve graph construction module is used for constructing a target RSRP distribution curve graph of the to-be-tested dummy network element according to the reference signal received power RSRP of the to-be-tested dummy network element reported by the user terminal in the test period;

the classification determining module is used for inputting the target RSRP distribution curve graph into a dummy network element classification model to obtain a target classification result of the dummy network element to be detected;

9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 7 when the computer program is executed.

10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 7.

11. A computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 7.