Data screening method, network index prediction method, device, equipment and medium

Abstract

The embodiment of the application provides a data screening method, a network index prediction method, a device, equipment and a medium, and belongs to the technical field of artificial intelligence and network communication. The method comprises the following steps: obtaining original prediction data and actual network index data of a network index to be predicted; performing dimension reduction processing on the original prediction data to obtain preliminary prediction data; screening the original prediction data to obtain basic prediction data and to-be-evaluated data; inputting the basic prediction data and the to-be-evaluated data into a preset prediction model to perform network index prediction, and obtaining prediction network index data; obtaining first error data of the to-be-evaluated data according to the prediction network index data and the actual network index data; sorting the to-be-evaluated data according to the first error data to obtain to-be-excluded data; and filtering the to-be-excluded data from the preliminary prediction data to obtain target prediction data. The embodiment of the application can reduce the calculation amount of the prediction model, thereby improving the prediction efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence and the technical field of network communication, and in particular relates to a data screening method, a network index prediction method, a device, equipment and a medium. BACKGROUND

[0002] User experience rate refers to the data rate obtained by a user per unit of time, and is used to represent the transmission rate available to the user in a real network environment. Therefore, the user experience rate can be used as a performance index for network optimization. In related technologies, the user experience rate is predicted by using a machine learning method such as a deep neural network. Specifically, a prediction model is built, and data collected from a real network environment is used as input of the prediction model, and the output of the prediction model is the user experience rate. However, in the above method, there are data in the input data of the model that are actually irrelevant to the user experience rate, thereby increasing the calculation amount of the prediction model. Therefore, how to screen the input data to reduce the calculation amount of the prediction model and improve the prediction efficiency has become a technical problem to be solved. SUMMARY

[0003] The main purpose of the embodiments of the present application is to provide a data screening method, a network index prediction method, a device, equipment and a medium, which aims to reduce the calculation amount of the prediction model and improve the prediction efficiency.

[0004] To achieve the above purpose, a first aspect of the embodiments of the present application provides a data screening method, which comprises:

[0005] obtaining original prediction data of a network index to be predicted and actual network index data;

[0006] performing dimension reduction processing on the original prediction data to obtain preliminary prediction data;

[0007] screening basic prediction data and to-be-evaluated data from the original prediction data;

[0008] inputting the basic prediction data and the to-be-evaluated data into a preset prediction model to perform network index prediction, and obtaining prediction network index data;

[0009] obtaining first error data of the to-be-evaluated data according to the prediction network index data and the actual network index data;

[0010] sorting the to-be-evaluated data according to the first error data to obtain to-be-excluded data;

[0011] filtering the to-be-excluded data from the preliminary prediction data to obtain target prediction data.

[0012] In some embodiments, the dimensionality reduction processing of the original prediction data to obtain preliminary prediction data includes:

[0013] The original prediction data is classified to obtain data groups;

[0014] Perform a loop operation until all the data groups are traversed; wherein, the loop operation includes: filtering out the data groups from the original prediction data to obtain first prediction data; inputting the first prediction data into the prediction model to predict network indicators to obtain first network indicator data;

[0015] The second error data of the data set is obtained based on the first network indicator data and the actual network indicator data;

[0016] The original prediction data is filtered based on the data group corresponding to the second error data to obtain the preliminary prediction data.

[0017] In some embodiments, the step of filtering the original prediction data based on the data group corresponding to the second error data to obtain the preliminary prediction data includes:

[0018] The data group is sorted according to the second error data to obtain sorting position data;

[0019] The data group is added to the preset original filtered dataset according to the sorting position data to obtain the preliminary filtered dataset;

[0020] The preliminary filtered dataset is filtered out from the original prediction data to obtain the second prediction data;

[0021] The second prediction data is input into the prediction model to predict network indicators, thereby obtaining the second network indicator data.

[0022] The target data group is determined from the data group based on the second network indicator data and the actual network indicator data;

[0023] The preliminary prediction data is obtained based on the target data set.

[0024] In some embodiments, obtaining the first error data of the data to be evaluated based on the predicted network metric data and the actual network metric data includes:

[0025] The difference between the predicted network indicator data and the actual network indicator data is calculated to obtain the predicted difference.

[0026] The first error data is calculated by comparing the predicted difference with the actual network indicator data.

[0027] To achieve the above objectives, a second aspect of this application proposes a network metric prediction method, wherein the network metric to be predicted includes user experience rate, the method comprising:

[0028] A preliminary data type for the target prediction data is determined, and the actual prediction data for the user experience rate is obtained based on the preliminary data type; wherein the target prediction data is obtained according to the method described in the first aspect;

[0029] The actual predicted data is input into the prediction model to predict the user experience rate, thereby obtaining the actual rate; wherein the target predicted data is obtained according to the method described in the first aspect.

[0030] In some embodiments, the prediction model includes an intermediate layer, and before determining the preliminary data type of the target prediction data, the method further includes updating the prediction model, specifically including:

[0031] The target prediction data is input into the prediction model to predict network indicators, thereby obtaining the first network indicator data.

[0032] By increasing or decreasing the number of intermediate layers, a preliminary prediction model can be obtained;

[0033] The original prediction data is input into the preliminary prediction model to predict network indicators, thereby obtaining the second network indicator data.

[0034] The target prediction data is input into the preliminary prediction model to predict network indicators, thereby obtaining the third network indicator data.

[0035] A third error data is obtained based on the first network indicator data and the actual network indicator data. A fourth error data is obtained based on the second network indicator data and the actual network indicator data. A fifth error data is obtained based on the third network indicator data and the actual network indicator data. The third error data, the fourth error data, and the fifth error data are compared numerically. If the comparison result indicates that the third error data value is the smallest, the prediction model is updated according to the preliminary prediction model.

[0036] To achieve the above objectives, a third aspect of this application provides a data filtering apparatus, the apparatus comprising:

[0037] The first data acquisition module is used to acquire the original predicted data and actual network indicator data of the network indicators to be predicted.

[0038] The dimensionality reduction module is used to perform dimensionality reduction processing on the original prediction data to obtain preliminary prediction data;

[0039] The filtering module is used to filter the original prediction data to obtain basic prediction data and data to be evaluated;

[0040] The first prediction module is used to input the basic prediction data and the data to be evaluated into a preset prediction model to predict network indicators and obtain predicted network indicator data.

[0041] An error data determination module is used to obtain first error data of the data to be evaluated based on the predicted network index data and the actual network index data.

[0042] The sorting module is used to sort the data to be evaluated according to the first error data to obtain the data to be excluded;

[0043] The filtering module is used to filter out the data to be excluded from the preliminary prediction data to obtain the target prediction data.

[0044] To achieve the above objectives, a fourth aspect of this application provides a network metric prediction apparatus, the apparatus comprising:

[0045] The second data acquisition module is used to determine the preliminary data type of the target prediction data and acquire the actual prediction data of the user experience rate based on the preliminary data type; wherein the target prediction data is obtained according to the method described in the first aspect;

[0046] The second prediction module is used to input the actual prediction data into the prediction model to predict the user experience rate and obtain the actual rate; wherein the target prediction data is obtained according to the method described in the first aspect.

[0047] To achieve the above objectives, a fifth aspect of the present application provides an electronic device, the electronic device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method described in the first or second aspect above.

[0048] To achieve the above objectives, a sixth aspect of the present application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in the first or second aspect.

[0049] The data filtering method, network indicator prediction method, apparatus, equipment, and medium proposed in this application first perform dimensionality reduction processing on the original prediction data to obtain preliminary prediction data with a smaller data volume. Based on this, data to be excluded from the preliminary data is then filtered out, resulting in target prediction data with a further reduced data volume. Therefore, when predicting network indicators based on the target prediction data and the prediction model, the amount of data input to the prediction model can be reduced, thereby reducing the computational load of the prediction model, lowering the performance requirements of the computing equipment loading the prediction model, and improving the prediction efficiency of the network indicators. Attached Figure Description

[0050] Figure 1 This is a flowchart of the data filtering method provided in the embodiments of this application;

[0051] Figures 2A-2C This is a schematic diagram of the original prediction data;

[0052] Figure 3A yes Figure 1 The flowchart of step S120 in the middle;

[0053] Figure 3B yes Figure 3A The flowchart of step S320 in the middle;

[0054] Figure 4 yes Figure 3A The flowchart of step S340 in the text;

[0055] Figure 5 This is a schematic diagram of the sorting position data provided in an embodiment of this application;

[0056] Figure 6 This is a flowchart of the first error data calculation method provided in the embodiments of this application;

[0057] Figure 7 This is a flowchart of a network metric prediction method provided in an embodiment of this application;

[0058] Figure 8 This is another flowchart of the network metric prediction method provided in the embodiments of this application.

[0059] Figures 9A-9B This is a schematic diagram of the experimental results provided in the embodiments of this application;

[0060] Figure 10 This is a schematic diagram of the structure of the data filtering device provided in the embodiments of this application;

[0061] Figure 11 This is a schematic diagram of the network index prediction device provided in the embodiments of this application;

[0062] Figure 12This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0064] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, and the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0065] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0066] First, let's analyze some of the terms used in this application:

[0067] Artificial intelligence (AI) is a new branch of computer science that studies, develops, and applies theories, methods, technologies, and systems to simulate, extend, and expand human intelligence. It aims to understand the essence of intelligence and produce intelligent machines that can react in a way similar to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing, and expert systems. AI can simulate the information processes of human consciousness and thought. Furthermore, AI utilizes digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceiving the environment, acquiring knowledge, and using that knowledge to achieve optimal results.

[0068] User experience rate (UXR) refers to the rate at which a user receives data per unit of time, representing the transmission rate a user can obtain in a real network environment. Therefore, UXR can serve as a performance indicator for network optimization. Related technologies use machine learning methods such as deep neural networks to predict UXR. Specifically, a prediction model is built, and data collected from the real network environment is used as input; the output of the prediction model is the UXR. However, in these methods, the input data can easily contain data that is actually irrelevant to UXR, thus increasing the computational load of the prediction model. Therefore, how to filter the input data to reduce the computational load of the prediction model and improve prediction efficiency has become an urgent technical problem to be solved.

[0069] Based on this, embodiments of this application provide a data filtering method, a network index prediction method, an apparatus, a device, and a medium, which aim to reduce the computational load of the prediction model and thus improve prediction efficiency.

[0070] The data filtering method, network indicator prediction method, apparatus, device and medium provided in the embodiments of this application are specifically described through the following embodiments. First, the data filtering method in the embodiments of this application is described.

[0071] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0072] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0073] The data filtering method provided in this application relates to the fields of artificial intelligence technology and network communication technology. The data filtering method provided in this application can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, etc.; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application implementing the data filtering method, but is not limited to the above forms.

[0074] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0075] It should be noted that in all specific embodiments of this application, when processing is required based on user information, user behavior data, or data from the user's terminal, the user's permission or consent will be obtained first. Furthermore, the collection, use, and processing of this data will comply with relevant laws, regulations, and standards of the relevant countries and regions. In addition, when embodiments of this application require access to sensitive personal information of users, separate permission or consent from the user will be obtained through pop-ups or redirects to confirmation pages. Only after obtaining the user's separate permission or consent will the necessary user-related data for the normal operation of the embodiments of this application be obtained.

[0076] Figure 1 This is an optional flowchart of the data filtering method provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S110 to S170.

[0077] Step S110: Obtain the original predicted data and actual network indicator data of the network indicators to be predicted;

[0078] Step S120: Dimensionality reduction is performed on the original prediction data to obtain preliminary prediction data;

[0079] Step S130: Filter the raw prediction data to obtain the basic prediction data and the data to be evaluated;

[0080] Step S140: Input the basic prediction data and the data to be evaluated into the preset prediction model to predict the network indicators and obtain the predicted network indicator data.

[0081] Step S150: Obtain the first error data of the data to be evaluated based on the predicted network indicator data and the actual network indicator data;

[0082] Step S160: Sort the data to be evaluated according to the first error data to obtain the data to be excluded;

[0083] Step S170: Filter out the data to be excluded from the preliminary prediction data to obtain the target prediction data.

[0084] It should be noted that, in the embodiments of this application, the network metric to be predicted refers to the network metric that needs to be predicted, such as user experience rate, latency, regional transmission capacity, peak data rate, and other metrics related to network communication. This application does not specifically limit the specific network metric to be predicted. However, for ease of explanation, this application uses user experience rate as an example.

[0085] In step S110 of some embodiments, actual network indicator data refers to the real data of the network indicator to be predicted under actual conditions, while original prediction data refers to data related to the network indicator to be predicted. When the network indicator to be predicted is the user experience rate, the actual network indicator data is the real value corresponding to the user experience rate obtained in the actual network environment, and the original prediction data refers to data related to the user experience rate. It is understood that "related" in the embodiments of this application refers to the following situation: when data A can be used to predict data B, it indicates that data A is related to data B. Figures 2A-2C As shown, the data related to user experience rate (i.e., the raw prediction data) includes 74 dimensions. This 74-dimensional data can be obtained from the user's terminal or the base station, and this 74-dimensional data is only exemplary; that is, the data can be added or deleted according to the actual situation.

[0086] In step S120 of some embodiments, dimensionality reduction refers to the operation of converting high-dimensional data into low-dimensional data. The preliminary prediction data is obtained after dimensionality reduction of the original prediction data; therefore, the amount of preliminary prediction data is less than the amount of original prediction data. Dimensionality reduction can be performed using any of the following methods: empirical methods, measurement algorithms, statistical analysis methods (including variance filtering, chi-square filtering, etc.), machine learning algorithms, etc. The specific methods of dimensionality reduction in the embodiments of this application will be described in detail below.

[0087] In step S130 of some embodiments, the basic prediction data refers to the data in the original prediction data that has a weak correlation with the network metric to be predicted, and the data to be evaluated refers to all data in the original prediction data other than the basic prediction data. For example, if the basic prediction data is 28-dimensional data, then the data to be evaluated is 46-dimensional data. It is understood that the method for determining the basic prediction data can be any of the historical experience method, machine learning algorithm, etc., and this application embodiment does not specifically limit this.

[0088] In steps S140 and S150 of some embodiments, the prediction model is a pre-built model with the ability to predict the network metric to be predicted. This application embodiment does not specifically limit the model structure of the prediction model. Basic prediction data and data to be evaluated are used as input data to the prediction model, enabling the prediction model to predict the predicted value of the network metric to be predicted (i.e., predicted network metric data, such as the predicted user experience rate) based on the basic prediction data and the data to be evaluated. Since the basic prediction data is data with a weak correlation to the network metric to be predicted, the obtained predicted network metric data is actually greatly affected by the input data to be evaluated. Therefore, in step S150, a first error data representing the prediction error can be obtained based on the predicted network metric data and the actual network metric data. The specific calculation method for the first error data will be described below. The operations described in steps S140 and S150 are performed on each piece of data to be evaluated to determine the first error data for each piece of data to be evaluated.

[0089] In steps S160 and S170 of some embodiments, the data to be evaluated is sorted according to the first error data, for example, by numerical value, either from largest to smallest or smallest to largest. Taking sorting from largest to smallest as an example, if the first error data is small, it indicates that after inputting the corresponding data to be evaluated and the weakly correlated basic prediction data into the prediction model, the prediction error of the prediction model decreases, meaning the corresponding data to be evaluated has a strong correlation with the network index to be predicted. Conversely, if the first error data is large, it indicates that after inputting the corresponding data to be evaluated and the weakly correlated basic prediction data into the prediction model, the prediction error of the prediction model increases, meaning the corresponding data to be evaluated has a weak correlation with the network index to be predicted. Therefore, the data to be evaluated that is ranked higher is considered data to be excluded. Since the preliminary prediction data obtained in step S120 is data obtained through preliminary screening, there may still be data with weak correlation in the preliminary prediction data. Therefore, in step S170, the data to be excluded in the preliminary prediction data is filtered out, and the remaining data (i.e., the target prediction data) are all data with strong correlation to the network indicators to be predicted. It can be seen that the target prediction data is data obtained from the original prediction data after two screenings, and the amount of data is reduced to a certain extent compared with the original prediction data.

[0090] Steps S110 to S170, as illustrated in the embodiments of this application, involve first performing dimensionality reduction on the original prediction data to obtain preliminary prediction data with a smaller data volume. Based on this, data to be excluded from the preliminary data is then filtered out, resulting in target prediction data with a further reduced data volume. Therefore, when predicting the network indicators to be predicted based on the target prediction data and the prediction model, the amount of data input to the prediction model can be reduced, thereby reducing the computational load of the prediction model, lowering the performance requirements of the computing equipment loading the prediction model, and improving the prediction efficiency of the network indicators to be predicted.

[0091] The dimensionality reduction method will be explained in detail below. (Refer to...) Figure 3A and Figure 3B In some embodiments, step S120 includes, but is not limited to, steps S310 to S340.

[0092] Step S310: Classify the original prediction data to obtain data groups;

[0093] Step S320: Perform a loop operation until all data groups have been traversed;

[0094] Step S330: Obtain the second error data of the data set based on the first network index data and the actual network index data;

[0095] Step S340: Filter the original prediction data according to the data group corresponding to the second error data to obtain preliminary prediction data.

[0096] The loop operation in step S320 includes, but is not limited to, steps S321 to S322.

[0097] Step S321: Filter out data groups from the original prediction data to obtain the first prediction data;

[0098] Step S322: Input the first prediction data into the prediction model to predict the network index and obtain the first network index data.

[0099] In step S310 of some embodiments, the original prediction data is classified to obtain multiple data groups. The data classification criteria can include a preset quantity, acquisition source, and data type. The preset quantity refers to evenly distributing the original prediction data so that each data group includes the same preset quantity of original prediction data. The acquisition source refers to grouping original prediction data from the same source into a single data group. The data type refers to grouping original prediction data of the same data type into a single data group. Since the correlation between original prediction data of the same data type and the network indicator to be predicted is relatively small, this embodiment of the application classifies data based on data type, thereby enabling the simultaneous determination of the correlation between multiple original prediction data of the same data type and the network indicator to be predicted.

[0100] Specifically, methods for determining data types can include symbolic methods and clustering methods. Symbolic methods refer to pre-assigning labels (including characters, colors, tags, etc.) to the original predicted data, since the types of original predicted data are known beforehand, while the numerical values ​​corresponding to the original predicted data are unknown. Different labels are assigned to different types of original predicted data, and the same labels are assigned to original predicted data of the same type, allowing data classification based on the labels. Clustering methods refer to classifying the original predicted data using clustering algorithms. Taking characters as an example of symbols, refer to... Figures 2A-2C A different identifier character is assigned to each piece of raw prediction data. For example, the identifier character for the raw prediction data corresponding to the average downlink rank in dimension 38 is AveDLRank, and the identifier character for the raw prediction data corresponding to the average modulation and coding scheme in dimension 55 is MIMOAveMCS. Raw prediction data of the same data type have the same character prefix, while raw prediction data of different data types have different character prefixes. For example... Figures 2A-2CThe 74-dimensional original prediction data shown can be divided into fifteen categories. Category 1 includes dimensions 1-8; Category 2 includes dimensions 9-16; Category 3 includes dimensions 17-24; Category 4 includes dimensions 25-27; Category 5 includes dimensions 28-37; Category 6 includes dimensions 38-40; Category 7 includes dimensions 41-45; Category 8 includes dimensions 46-52; Category 9 includes dimensions 53-59; Category 10 includes dimensions 60-62; Category 11 includes dimensions 63-65; Category 12 includes dimensions 66-68; Category 13 includes dimensions 69-71; Category 14 includes dimensions 72-74; and Category 15 includes dimensions 9-24. In other words, the original prediction data can be divided into fifteen groups using the above method.

[0101] In steps S321 to S322 of some embodiments, the first predicted data refers to the data remaining after filtering out any group of data from the original predicted data. The first predicted data is used as input data to the prediction model, enabling the prediction model to predict network metrics based on the first predicted data, thus obtaining the first network metric predicted data. This operation is performed on each group of data until all groups of data have been traversed; that is, the above operation needs to be repeated fifteen times.

[0102] In step S330 of some embodiments, second error data for the corresponding data set is calculated based on the first network indicator data and the actual network indicator data. The second error data represents the error in the prediction model's prediction of network indicators based on the first prediction data. The specific calculation method for the second error data can refer to the calculation method for the first error data described below. For example, still using... Figures 2A-2C For example, the second error data corresponding to each data group is shown in Table 1 below.

[0103]

[0104]

[0105] Table 1

[0106] In Table 1, "Deleted Feature Class" refers to the data group corresponding to the deleted feature class, and "Remaining Data Dimensions" refers to the amount of data in the first prediction data. For example, deleting the first class means deleting the data group consisting of dimensions 1-8 from the 74-dimensional original prediction data, leaving a total of 66 dimensions in the first prediction data. Inputting this 66-dimensional data into the prediction model for network indicator prediction, the second error data calculated based on the predicted first network indicator data and the actual network indicator data is 35.79%. It can be understood that a deleted feature class of "0" means no deletion, and the 74-dimensional original prediction data is input into the prediction model for network indicator prediction. The error data calculated based on the prediction model's output data and the actual network indicator data is 34.98%. Other data in Table 1 are obtained through the same operation. This 34.98% error data can be used as reference error data. The role of the reference error data is to constrain the data filtering operation, that is, to ensure that data filtering is performed when the error data changes little, meaning that the data filtering operation should not affect the accuracy of the prediction.

[0107] In step S340 of some embodiments, after deleting each data group from the original prediction data, the larger the value of the second error data obtained from the corresponding first prediction data, the stronger the correlation between the deleted data group and the network index to be predicted; conversely, the smaller the value of the second error data obtained from the corresponding first prediction data, the weaker the correlation between the deleted data group and the network index to be predicted. Therefore, the data groups with strong correlation to the network index to be predicted can be determined based on the second error data, thereby filtering preliminary prediction data from the original prediction data based on the corresponding data groups. The specific filtering method for the preliminary prediction data will be described in detail below.

[0108] The advantage of steps S310 to S340 is that it enables preliminary screening of the original prediction data, and the screening is performed in the form of data groups, which can improve the efficiency and accuracy of preliminary screening to a certain extent.

[0109] The following section elaborates on the methods used to select preliminary forecast data. (Refer to...) Figure 4 In some embodiments, step S340 includes, but is not limited to, steps S410 to S460.

[0110] Step S410: Sort the data group according to the second error data to obtain the sorting position data;

[0111] Step S420: Add data groups to the preset original filtered dataset according to the sorting position data to obtain the preliminary filtered dataset;

[0112] Step S430: Filter out the preliminary filtered dataset from the original prediction data to obtain the second prediction data;

[0113] Step S440: Input the second prediction data into the prediction model to predict the network index and obtain the second network index data;

[0114] Step S450: Determine the target data group from the data group based on the second network indicator data and the actual network indicator data;

[0115] Step S460: Obtain preliminary prediction data based on the target data set.

[0116] In step S410 of some embodiments, the data groups are sorted from largest to smallest or smallest to largest based on the second error data to obtain sorting position data. Sorting position data refers to the position of a data group within all sorted data groups, for example, referring to... Figure 5 The sorting position of data group A is 1, the sorting position of data group B is 2, and so on. Since there are a total of 15 data groups, the maximum value of the sorting position is 15.

[0117] In step S420 of some embodiments, the original filtered dataset is a pre-set dataset, which is either an empty dataset or a dataset containing specific data in its initial state. Data groups are added sequentially to the original filtered dataset according to their sorting positions to obtain multiple preliminary filtered datasets with different amounts of data. The order in which the data groups are added is determined based on the correlation between the data groups and the metrics of the network to be predicted. Specifically, data groups with weak correlation are added first, followed by data groups with strong correlation, to avoid bias in the judgment of data groups with weak correlation when adding data groups with strong correlation first. Therefore, when sorted from largest to smallest, the data group at the top of the sorting is added to the original filtered dataset first, i.e., the data group with the smallest sorting position is added first. When sorted from smallest to largest, the data group at the bottom of the sorting is added to the original filtered dataset first, i.e., the data group with the largest sorting position is added first. The specific operation of adding data sequentially is as follows: In the first addition operation, the data group determined according to the above method (such as data group A) is added to the original filtered dataset in its initial state to obtain a preliminary filtered dataset. In the second addition operation, another data group (such as data group B) is determined from the remaining data groups using the method described above, and this data group (data group B) is added to the preliminary filtered dataset obtained in the first addition operation to update the preliminary filtered dataset. This process is repeated until, after a certain addition operation, the updated preliminary filtered dataset contains fifteen data groups. Therefore, the sequential addition operation continuously adds new data groups to update the number of data groups contained in the preliminary filtered dataset, ultimately resulting in multiple preliminary filtered datasets with different data sizes.

[0118] In steps S430 to S440 of some embodiments, a preliminary filtered dataset is removed from the original prediction data, and the remaining data is used as the second prediction data. The second prediction data is used as input data for the prediction model, enabling the model to predict the network metric to be predicted based on the second prediction data, thus obtaining the second network metric data. Error data is calculated based on the second network metric data and the actual network metric data; the calculation method for this error data can refer to the calculation method for the first error data. The above operation is performed on each preliminary filtered dataset to obtain multiple error data sets. For example, the error data shown in Table 2 below are obtained.

[0119]

[0120]

[0121] Table 2

[0122] Since the fifteenth category is a special category composed of the second and third categories, it differs from the method described above for determining the initial filtered dataset. First, the fifteenth category is filtered out from the original prediction data, and an error value of 34.79% is calculated based on the remaining data and the prediction model. In other words, the data group corresponding to the fifteenth category is used as the specific data included in the original filtered dataset. Based on this, the data group corresponding to the fourteenth category is added, and the original filtered dataset is updated according to the data group corresponding to the fourteenth category, resulting in the initial filtered dataset (including the data groups corresponding to the fifteenth and fourteenth categories). At this point, an error value of 34.89% is calculated. Then, based on the above operation, the data group corresponding to the twelfth category is added, and the initial filtered dataset (including the data groups corresponding to the fifteenth, fourteenth, and twelfth categories) is updated according to the data group corresponding to the twelfth category, resulting in an error value of 34.71%, and so on.

[0123] In step S450 of some embodiments, the larger the error value calculated from the second network indicator data obtained based on an updated preliminary filtering dataset and the actual network indicator data, the stronger the correlation between the data group that triggered the update of the preliminary filtering dataset and the network indicator to be predicted, and the data group is taken as the target data group. It is understood that the term "triggered" described in the embodiments of this application refers to the following situation: when data group H is added to the preliminary filtering dataset obtained in the previous operation, the preliminary filtering dataset is updated based on data group H; at this time, it is said that data group H can trigger the update of the preliminary filtering dataset.

[0124] It is understood that the number of target data groups can be adaptively set according to the actual situation, and this application example does not make a specific limitation. However, for ease of explanation, since the error values ​​corresponding to adding the seventh, fifth, first, and sixth categories are the largest in Table 2, this application embodiment uses the following four data groups as target data groups: the data group corresponding to the seventh category, the data group corresponding to the fifth category, the data group corresponding to the first category, and the data group corresponding to the sixth category.

[0125] In step S460 of some embodiments, the determined multiple target data groups are combined to obtain preliminary prediction data. For example, the four target data groups are combined to obtain 28-dimensional preliminary prediction data.

[0126] The advantage of steps S410 to S460 is that by adding data groups sequentially by sorting position data to update the method of initially filtering the dataset, the correlation between each data group and the data to be predicted can be accurately determined, and the influence of data groups with strong correlation on data groups with weak correlation can be avoided, thereby improving the accuracy of determining the target data group.

[0127] The steps S130 to S170 will be explained in detail below with reference to examples from the above embodiments.

[0128] Reference Figures 2A-2C The 28 original prediction data dimensions (1-27 and 60) are used as the basic prediction data, while the remaining original prediction data are used as the data to be evaluated. Understandably, due to the special nature of wave velocity, dimensions 1-8 are usually used in combination. In one operation, the basic prediction data and one data to be evaluated are used together as input data to the prediction model, resulting in a first error data. The above operation is performed on all the data to be evaluated, resulting in the first error data shown in Table 3 below.

[0129]

[0130]

[0131] Table 3

[0132] In Table 3, adding "0" to the data to be evaluated indicates that only the basic prediction data is used as input data for the prediction model. Table 3 shows how the first error data changes after adding data to be evaluated, and thus the strength of the correlation between each data point to be evaluated and the network indicator to be predicted. For example, if the first error data is greater than 62.99% after adding a certain data point to be evaluated, it indicates a weak correlation between that data point and the network indicator; if the first error data is less than 62.99%, it indicates a strong correlation.

[0133] Among them, reference Figure 6 In some embodiments, the method for calculating the first error data includes steps S610 to S620.

[0134] Step S610: Calculate the difference between the predicted network indicator data and the actual network indicator data to obtain the predicted difference.

[0135] Step S620: Calculate the ratio between the predicted difference and the actual network index data to obtain the first error data.

[0136] As shown in Table 3, the first error data essentially refers to the absolute percentage value between the predicted network indicator data and the actual network indicator data. Therefore, in step S610, the difference between the predicted network indicator data and the actual network indicator data is first calculated, and the absolute value of the difference is used as the prediction difference. In step S620, the ratio of the prediction difference to the actual network indicator data is calculated to obtain the first error data.

[0137] After obtaining the first error value for each piece of data to be evaluated using the above method, the data to be evaluated can be sorted according to the first error data, and preliminary exclusion data can be determined based on the sorting results. It is understood that preliminary exclusion data refers to data among multiple pieces of data to be evaluated that have a weak correlation with the prediction network index. Subsequently, the preliminary exclusion data is combined to obtain combined data. The combination method can be arbitrary, based on experience, based on data type, based on relevance, etc., and this embodiment of the application does not specifically limit this. Multiple combined data are obtained, and the combined data in the preliminary prediction data is filtered out. The remaining data is used as input data for the prediction model to obtain prediction data, and then error data. Finally, the combined data corresponding to the error data with the smallest difference from the reference error data is taken as data to be excluded. After filtering out this data from the preliminary prediction data, the amount of data (i.e., data dimension) of the remaining data is also relatively small. The advantage of this is that it can further reduce the amount of data based on the preliminary prediction data while ensuring that the prediction accuracy remains basically unchanged. Table 4 below illustrates the filtering of combined data in the preliminary prediction data.

[0138]

[0139]

[0140] Table 4

[0141] Based on the error data in Table 4, it can be determined that the combined data with a data dimension of 23 and an error rate of 34.92% in Table 4 is data to be excluded. This data to be excluded is filtered out from the 28-dimensional preliminary prediction data obtained according to steps S410 to S460, resulting in the 23-dimensional target prediction data shown in Table 5 below:

[0142] TOP 0-7 Beam RSRP AveDL Rank, AveDL M CS, AveDL UE PRB Used MIMO AveDL Sch Rank Rb MIMO AveDL Sch Rb MIMO AveM CS MIMO AveDL Mu Sch Rank Rb MIMO AveDL Mu Sch Rb MIMO AveMu M CS MIMO AveBeam Pair UE Measlnfo.AccumMeaslnfes.DLPrbUsedNum UE Measlnfo.AccumMeaslnfes.DlIniTranNackNumCode0 UE Measlnfo.AccumMeaslnfes.DlIniTranTotalNumCode0 UE Measlnfo.AccumMeaslnfes.DlRetranNumCode0 UE Measlnfo.AccumMeaslnfes.DlMCSCode0

[0143] Table 5

[0144] Reference Figure 7 This application also provides a network metric prediction method, wherein the network metric to be predicted is the user experience rate. The network metric prediction method includes, but is not limited to, steps S710 to S720.

[0145] Step 710: Determine the preliminary data type of the target prediction data, and obtain the actual prediction data of the user experience rate based on the preliminary data type;

[0146] Step 720: Input the actual prediction data into the prediction model to predict the user experience rate and obtain the actual rate.

[0147] In step S710 of some embodiments, a preliminary data type of the target prediction data determined according to the above embodiments is obtained. The preliminary data type is used to represent the data type of each target prediction data. In practical applications, the corresponding data in the real network is obtained according to this data type to obtain the actual prediction data. Taking Table 5 as an example, 23-dimensional actual prediction data will be obtained.

[0148] In step S710 of some embodiments, the actual predicted data is used as input data for the prediction model obtained according to the above embodiments to predict the user experience rate and obtain the actual rate.

[0149] The advantage of steps S710 to S720 is that it can reduce the amount of actual prediction data obtained, thereby improving the prediction efficiency of user experience rate.

[0150] In some embodiments, before performing step S710, that is, before making actual predictions, the prediction model, target prediction data, and original prediction data can be experimentally compared to determine the feasibility of the data filtering method described in the above embodiments, thereby improving the accuracy of predictions in actual predictions.

[0151] The experimental comparison methods are explained in detail below. (Refer to...) Figure 8 Prior to step S710, the network metric prediction method also includes, but is not limited to, steps S810 to S850.

[0152] Step S810: Input the target prediction data into the prediction model to predict the network index and obtain the first network index data;

[0153] Step S820: Increase or decrease the number of intermediate layers to obtain a preliminary prediction model;

[0154] Step S830: Input the original prediction data into the preliminary prediction model to predict network indicators and obtain the second network indicator data;

[0155] Step S840: Input the target prediction data into the preliminary prediction model to predict network indicators and obtain the third network indicator data;

[0156] Step S850: Obtain the third error data based on the first network indicator data and the actual network indicator data; obtain the fourth error data based on the second network indicator data and the actual network indicator data; obtain the fifth error data based on the third network indicator data and the actual network indicator data; compare the values ​​of the third error data, the fourth error data, and the fifth error data; if the comparison result indicates that the value of the third error data is the smallest, update the prediction model based on the preliminary prediction model.

[0157] In some embodiments, the predictive model structure includes an input layer, intermediate layers, and an output layer, wherein the intermediate layers include hidden layers. Hidden layers are used to abstract the features of the input data to another dimensional space to represent more abstract features of the input data, thereby enabling better linear partitioning. When there are multiple hidden layers, it represents a multi-level abstraction of the features of the input data. Therefore, the number of hidden layers is related to the output of the predictive model. Thus, in this embodiment, based on the original predictive model, the number of intermediate layers and the dimensionality of the input data are changed to determine the impact of different numbers of layers and different dimensions of input data on the prediction results.

[0158] Specifically, in step S810 of some embodiments, the target prediction data is input into the prediction model to predict the network index, and the first network index data is obtained. The first network index data is the prediction result obtained based on the number of intermediate layers of the original prediction model and the filtered target prediction data.

[0159] In step S820 of some embodiments, the number of intermediate layers in the original prediction model is increased or decreased. For example, if the prediction model has 6 layers in step S810, this step adds 1 intermediate layer, resulting in a preliminary prediction model with 7 intermediate layers. It is understood that the number of intermediate layers added each time can be the same or different. For example, this step adds 1 layer, and subsequently, 2, 3, etc., layers can be added. The operation of decreasing the number of intermediate layers is similar to the operation of increasing the number of intermediate layers described above, and will not be repeated here.

[0160] In steps S830 to S840 of some embodiments, the original prediction data and the target prediction data are respectively input into the preliminary prediction model to predict network indicators, so as to determine the difference in prediction results obtained by the same prediction model based on input data of different dimensions.

[0161] In step S850 of some embodiments, the error between the first network indicator data and the actual network indicator data is calculated to obtain third error data; the error between the second network indicator data and the actual network indicator data is calculated to obtain fourth error data; and the error between the third network indicator data and the actual network indicator data is calculated to obtain fifth error data. The reference error data, the third error data, the fourth error data, and the fifth error data are numerically compared, and the prediction model is updated based on the number of intermediate layers corresponding to the error data with the smallest value. For example, as shown in Table 6 below.

[0162]

[0163] Table 6

[0164] In Table 6, the original prediction model has 6 intermediate layers. When the original prediction data is used as input, the error rate is 34.98% (reference error rate); when the target prediction data is used as input, the error rate is 34.92%. Increasing the number of intermediate layers from 6 to 7, the error rate is 35.09% when the original prediction data is used as input, and 35.03% when the target prediction data is used as input. Decreasing the number of intermediate layers from 6 to 5, the error rate is 34.57% when the original prediction data is used as input, and 34.94% when the target prediction data is used as input. Comparing these six error rates, the results show that 6 intermediate layers and a 23-dimensional input data result in the lowest error rate. Therefore, in this embodiment, the number of layers in the prediction model is not updated. The advantage of doing this is that it allows us to determine whether the target forecast data can replace the original forecast data.

[0165] Assuming the intermediate layers are 7 and the input data has 23 dimensions, the corresponding error data is minimized. Therefore, the intermediate layers of the prediction model are updated to 7. Then, in steps S710 to S720, the prediction model with 7 intermediate layers is used for practical application.

[0166] Taking experimental data with 6 intermediate layers as an example, assuming that ReLU is used as the activation function in each intermediate layer, and Dropout = 0.2 is selected to prevent overfitting, the number of iterations for the prediction model is set to 25, and the learning rate is set to 10.-4 When training the prediction model, the training and test sets are divided in an 8:2 ratio. (Refer to...) Figure 9A This diagram illustrates the error data when the prediction model is trained 25 times using 74-dimensional data (i.e., the original prediction data) as input and based on the prediction results and actual network metrics. (Refer to...) Figure 9B This diagram illustrates the error data when the prediction model is trained 25 times using 23-dimensional data (i.e., target prediction data) as input and based on the prediction results and actual network metrics. Figure 9A and Figure 9B Comparative analysis revealed that 23-dimensional data does not negatively impact the prediction results; therefore, 23-dimensional data can be used instead of 74-dimensional data to predict user experience rate. It is understood that the error data in Tables 1 to 4 and Table 6 above are all obtained from the converged prediction model. The convergence conditions include reaching the preset number of iterations (25 times) as described above, and the loss function decreasing by no more than the preset loss function threshold.

[0167] Please see Figure 10 This application also provides a data filtering device that can implement the above-described data filtering method. The device includes:

[0168] The first data acquisition module 1010 is used to acquire the original prediction data and actual network indicator data of the network indicator to be predicted.

[0169] Dimensionality reduction module 1020 is used to reduce the dimensionality of the original prediction data to obtain preliminary prediction data;

[0170] The filtering module 1030 is used to filter the raw prediction data to obtain the basic prediction data and the data to be evaluated.

[0171] The first prediction module 1040 is used to input the basic prediction data and the data to be evaluated into the preset prediction model to predict the network indicators and obtain the predicted network indicator data.

[0172] Error data determination module 1050 is used to obtain the first error data of the data to be evaluated based on the predicted network index data and the actual network index data.

[0173] The sorting module 1060 is used to sort the data to be evaluated based on the first error data to obtain the data to be excluded;

[0174] The filtering module 1070 is used to filter out the data to be excluded from the preliminary prediction data to obtain the target prediction data.

[0175] The specific implementation of this data filtering device is basically the same as the specific implementation of the data filtering method described above, and will not be repeated here.

[0176] Please see Figure 11 This application also provides a network metric prediction apparatus that can implement the above-described network metric prediction method. The apparatus includes:

[0177] The second data acquisition module 1110 is used to determine the preliminary data type of the target prediction data and acquire the actual prediction data of the user experience rate based on the preliminary data type.

[0178] The second prediction module 1120 is used to input the actual prediction data into the prediction model to predict the user experience rate and obtain the actual rate.

[0179] The specific implementation of this network indicator prediction device is basically the same as the specific implementation of the network indicator prediction method described above, and will not be repeated here.

[0180] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned data filtering method or network indicator prediction method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0181] Please see Figure 12 , Figure 12 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes:

[0182] The processor 1210 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application.

[0183] The memory 1220 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1220 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 1220 and is called and executed by the processor 1210 to execute the data filtering method or network indicator prediction method of the embodiments of this application.

[0184] The input / output interface 1230 is used to implement information input and output.

[0185] The communication interface 1240 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0186] Bus 1250 transmits information between various components of the device (e.g., processor 1210, memory 1220, input / output interface 1230, and communication interface 1240);

[0187] The processor 1210, memory 1220, input / output interface 1230 and communication interface 1240 are connected to each other within the device via bus 1250.

[0188] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described data filtering method or network indicator prediction method.

[0189] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0190] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0191] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0192] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0193] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0194] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0195] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0196] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0197] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0198] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0199] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0200] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A data filtering method, characterized in that, The method includes: Obtain the original predicted data and actual network indicator data of the network indicator to be predicted; where the original predicted data refers to the data related to the network indicator to be predicted. The original prediction data is subjected to dimensionality reduction processing to obtain preliminary prediction data; Basic prediction data and data to be evaluated are obtained by filtering from the original prediction data; wherein, the basic prediction data refers to the data in the original prediction data that has a weak correlation with the network index to be predicted, and the data to be evaluated refers to all data in the original prediction data other than the basic prediction data. The basic prediction data and the data to be evaluated are input into a preset prediction model to predict network indicators, thereby obtaining predicted network indicator data. The first error data of the data to be evaluated is obtained based on the predicted network index data and the actual network index data; The data to be evaluated is sorted according to the first error data, and preliminary exclusion data is obtained according to the sorting results. The preliminary exclusion data is combined to obtain combined data, and the data to be excluded is obtained according to the combined data. The preliminary exclusion data refers to data that is weakly related to the prediction network. The data to be excluded is filtered out from the preliminary prediction data to obtain the target prediction data; The step of performing dimensionality reduction on the original prediction data to obtain preliminary prediction data includes: The original prediction data is classified to obtain data groups; Perform a loop operation until all the data groups are traversed; wherein, the loop operation includes: filtering out the data groups from the original prediction data to obtain first prediction data; inputting the first prediction data into the prediction model to predict network indicators to obtain first network indicator data; The second error data of the data set is obtained based on the first network indicator data and the actual network indicator data; The original prediction data is filtered based on the data group corresponding to the second error data to obtain the preliminary prediction data.

2. The method according to claim 1, characterized in that, The step of filtering the original prediction data according to the data group corresponding to the second error data to obtain the preliminary prediction data includes: The data group is sorted according to the second error data to obtain sorting position data; The data group is added to the preset original filtered dataset according to the sorting position data to obtain the preliminary filtered dataset; The preliminary filtered dataset is filtered out from the original prediction data to obtain the second prediction data; The second prediction data is input into the prediction model to predict network indicators, thereby obtaining the second network indicator data. The target data group is determined from the data group based on the second network indicator data and the actual network indicator data; The preliminary prediction data is obtained based on the target data set.

3. The method according to any one of claims 1 to 2, characterized in that, The first error data for obtaining the data to be evaluated based on the predicted network metric data and the actual network metric data includes: The difference between the predicted network indicator data and the actual network indicator data is calculated to obtain the predicted difference. The first error data is calculated by comparing the predicted difference with the actual network indicator data.

4. A method for predicting network indicators, characterized in that, The network metrics to be predicted include user experience rate, and the method includes: A preliminary data type for the target prediction data is determined, and the actual prediction data for the user experience rate is obtained based on the preliminary data type; wherein the target prediction data is obtained by the method according to any one of claims 1 to 3; The actual prediction data is input into the prediction model to predict the user experience rate, thereby obtaining the actual rate; wherein the target prediction data is obtained by the method according to any one of claims 1 to 3.

5. The method according to claim 4, characterized in that, The prediction model includes an intermediate layer. Before determining the initial data type of the target prediction data, the method further includes updating the prediction model, specifically including: The target prediction data is input into the prediction model to predict network indicators, thereby obtaining the first network indicator data. By increasing or decreasing the number of intermediate layers, a preliminary prediction model can be obtained; The original prediction data is input into the preliminary prediction model to predict network indicators, thereby obtaining the second network indicator data. The target prediction data is input into the preliminary prediction model to predict network indicators, thereby obtaining the third network indicator data. A third error data is obtained based on the first network indicator data and the actual network indicator data. A fourth error data is obtained based on the second network indicator data and the actual network indicator data. A fifth error data is obtained based on the third network indicator data and the actual network indicator data. The third error data, the fourth error data, and the fifth error data are compared numerically. If the comparison result indicates that the third error data value is the smallest, the prediction model is updated according to the preliminary prediction model.

6. A data filtering device, characterized in that, The device includes: The first data acquisition module is used to acquire the original prediction data and actual network indicator data of the network indicator to be predicted; wherein, the original prediction data refers to the data related to the network indicator to be predicted. The dimensionality reduction module is used to perform dimensionality reduction processing on the original prediction data to obtain preliminary prediction data; The filtering module is used to filter the original prediction data to obtain basic prediction data and data to be evaluated; wherein, the basic prediction data refers to the data in the original prediction data that has a weak correlation with the network index to be predicted, and the data to be evaluated refers to all data in the original prediction data other than the basic prediction data. The first prediction module is used to input the basic prediction data and the data to be evaluated into a preset prediction model to predict network indicators and obtain predicted network indicator data. An error data determination module is used to obtain first error data of the data to be evaluated based on the predicted network index data and the actual network index data. The sorting module is used to sort the data to be evaluated according to the first error data, obtain preliminary excluded data according to the sorting result, combine the preliminary excluded data to obtain combined data, and obtain the data to be excluded according to the combined data; wherein, the preliminary excluded data refers to data that is weakly related to the prediction network; A filtering module is used to filter out the data to be excluded from the preliminary prediction data to obtain the target prediction data; The step of performing dimensionality reduction on the original prediction data to obtain preliminary prediction data includes: The original prediction data is classified to obtain data groups; Perform a loop operation until all the data groups are traversed; wherein, the loop operation includes: filtering out the data groups from the original prediction data to obtain first prediction data; inputting the first prediction data into the prediction model to predict network indicators to obtain first network indicator data; The second error data of the data set is obtained based on the first network indicator data and the actual network indicator data; The original prediction data is filtered based on the data group corresponding to the second error data to obtain the preliminary prediction data.

7. A network indicator prediction device, characterized in that, The device includes: The second data acquisition module is used to determine the preliminary data type of the target prediction data and acquire the actual prediction data of the user experience rate based on the preliminary data type; wherein, the target prediction data is obtained by the method according to any one of claims 1 to 3; The second prediction module is used to input the actual prediction data into the prediction model to predict the user experience rate and obtain the actual rate; wherein the target prediction data is obtained by the method according to any one of claims 1 to 3.

8. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method of any one of claims 1 to 3, or to implement the method of claim 4 or 5.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 3, or the method of claim 4 or 5.

Images