Training method of rate prediction model, rate prediction method and rate prediction system

Abstract

The invention provides a training method of a rate prediction model, a rate prediction method and a rate prediction system, relates to the technical field of communication, and is used for improving the prediction accuracy of a wireless downlink application layer rate. The method comprises the following steps: acquiring original simulation characteristics of a simulation sample; the original simulation features comprise a reference signal to noise ratio, a radio bearer channel code and a downlink application layer rate; determining a context rate priority factor based on the simulation channel sample set of the simulation sample; the context rate priority factor is used for representing the relative priority of the downlink application layer rate of the simulation sample under the similar condition of the signal-to-noise ratio of the reference signal; constructing a rate prediction feature vector of the simulation sample based on the context rate priority factor, the radio bearer channel code and the signal-to-noise ratio of the reference signal; and inputting the rate prediction feature vector of the simulation sample into a random forest regression model for training to obtain a rate prediction model which is used for predicting the rate of a downlink application layer.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a training method for a rate prediction model, a rate prediction method, and a rate prediction system. Background Technology

[0002] In recent years, as wireless communication networks have evolved towards 5G ultra-dense heterogeneous networks and 6G intelligent fully connected networks, network scenarios have exhibited three major characteristics: enhanced user mobility, more complex channel environments, and more dynamic scheduling strategies, leading to severe challenges in predicting wireless downlink rates.

[0003] Existing downlink rate prediction methods typically involve determining the rate by looking up tables or using the MCS-scheduling resource block mapping table in the 3GPP standard, processing it layer by layer across multiple modules such as the physical layer, link layer, and MAC layer, or using commercial simulation tools for chain-like derivation, that is, calculating the theoretical rate step by step based on scheduling resources, CQI mapping, modulation and coding rate, etc.

[0004] However, existing rate prediction methods suffer from drawbacks such as complex formulas, high computational overhead, and difficulty in reflecting application layer rate differences caused by scheduling strategy differences under similar reference signal-to-noise ratio conditions, resulting in inaccurate wireless downlink application layer rate prediction. Summary of the Invention

[0005] This application provides a training method for a rate prediction model, a rate prediction method, and a rate prediction system to improve the accuracy of predicting wireless downlink application layer rates.

[0006] In a first aspect, this application provides a training method for a rate prediction model, comprising: acquiring the original simulation features of simulation samples; wherein, the simulation samples are a structured representation of the original simulation data generated by downlink wireless link simulation operation; the original simulation features include the reference signal-to-noise ratio (SNR), wireless bearer channel coding, and downlink application layer rate; determining a context rate priority factor based on the simulation channel sample set of the simulation samples; the context rate priority factor is used to characterize the relative priority of the downlink application layer rate of the simulation samples under the condition of similar reference signal-to-noise ratio; the simulation channel sample set includes multiple key simulation channel samples that meet the sample selection rules; the sample selection rules are used to characterize the simulation channel samples whose SNR difference with the simulation samples does not exceed a preset SNR difference as key simulation channel samples; constructing a rate prediction feature vector of the simulation samples based on the context rate priority factor, wireless bearer channel coding, and reference signal-to-noise ratio; inputting the rate prediction feature vector of the simulation samples into a random forest regression model for training to obtain a rate prediction model, which is used to predict the downlink application layer rate.

[0007] The technical solution provided in this application obtains the original simulation features of the simulation samples, which can provide data support for subsequent feature construction. Based on the set of simulation channel samples of the simulation samples, the context rate priority factor is determined, which can measure the data transmission advantage of the simulation samples under similar reference signal-to-noise ratio conditions, and provide a basis for constructing a rate prediction feature vector with dynamic channel adaptability. Based on the context rate priority factor, wireless bearer channel coding, and reference signal-to-noise ratio, the rate prediction feature vector of the simulation samples is constructed. This can retain the spatiotemporal correlation between key features while reducing feature redundancy through dimensionality reduction, providing a structured data support basis for the subsequent construction of the rate prediction model. The rate prediction feature vector of the simulation samples is input into a random forest model for training to obtain a rate prediction model for predicting the wireless downlink rate. The accuracy of the wireless downlink rate prediction can be improved based on multi-dimensional feature fusion and dynamic splitting optimization of the regression decision tree.

[0008] In one possible implementation, the context rate priority factor is determined based on the set of simulated channel samples from the simulated samples, including: determining the average simulated context rate based on the simulated downlink rate of each key simulated channel sample in the set of simulated channel samples; and determining the context rate priority factor based on the downlink application layer rate and the average simulated context rate of the simulated samples.

[0009] Another possible implementation, the formula for calculating the context rate priority factor is as follows:

[0010]

[0011] Where CRP is the context rate priority factor; R is the downlink application layer rate of the simulation sample; μ represents the average rate of the context simulation; and ∈ is a constant for avoiding division-by-zero anomalies, with a value of 10. -6 .

[0012] Another possible implementation involves constructing a rate prediction feature vector for the simulation sample based on the context rate priority factor, radio bearer channel coding, and reference signal signal-to-noise ratio. This includes: determining the bearer efficiency of the simulation sample based on the radio bearer channel coding; and arranging the context rate priority factor, radio bearer channel coding, bearer efficiency, and reference signal signal-to-noise ratio according to a preset feature arrangement order to generate the rate prediction feature vector for the simulation sample.

[0013] Another possible implementation involves determining the carrying efficiency of the simulation sample based on wireless bearer channel coding, including: determining the carrying efficiency of the simulation sample based on wireless bearer channel coding and a preset mapping table; the preset mapping table contains multiple mapping relationship entries between wireless bearer channel coding and carrying efficiency.

[0014] Secondly, this application provides a training system for a rate prediction model, comprising: a feature acquisition module for acquiring the original simulation features of simulation samples; wherein the simulation samples are a structured representation of the original simulation data generated by the downlink wireless link simulation operation; the original simulation features include the reference signal-to-noise ratio (SNR), the wireless bearer channel coding, and the downlink application layer rate; a priority analysis module for determining a context rate priority factor based on a set of simulation channel samples of the simulation samples; the context rate priority factor is used to characterize the relative priority of the downlink application layer rate of the simulation samples under the condition of similar reference signal-to-noise ratio; the set of simulation channel samples includes multiple key simulation channel samples that meet the sample selection rules; the sample selection rules include: the difference between the SNR of the simulation samples and the reference signal-to-noise ratio of the simulation samples does not exceed a preset SNR difference; a feature construction module for constructing a rate prediction feature vector of the simulation samples based on the context rate priority factor, the wireless bearer channel coding, and the reference signal-to-noise ratio; and a model construction module for inputting the rate prediction feature vector of the simulation samples into a random forest regression model for training to obtain a rate prediction model, which is used to predict the downlink application layer rate.

[0015] Thirdly, this application provides a rate prediction method, comprising: obtaining a rate prediction feature vector of an application terminal; the application terminal being a user equipment in a wireless downlink; the rate prediction feature vector of the application terminal; inputting the rate prediction feature vector of the application terminal into a rate prediction model to obtain the corresponding downlink application layer rate; wherein the rate prediction model is trained by the method described in the first aspect above.

[0016] Fourthly, this application provides a rate prediction system, comprising: a data acquisition module for acquiring a rate prediction feature vector of an application terminal; the application terminal is a user equipment in a wireless downlink; the application terminal is a user equipment in a wireless downlink; the rate prediction feature vector of the application terminal; and a rate prediction module for inputting the rate prediction feature vector of the application terminal into a rate prediction model to obtain the corresponding downlink application layer rate; wherein the rate prediction model is trained by the method described in the first aspect above.

[0017] Fifthly, this application provides an electronic device comprising: a processor and a memory; the memory storing processor-executable instructions; when the processor is configured to execute the instructions, causing the electronic device to implement the training method of the rate prediction model of the first aspect and / or the rate prediction method of the third aspect.

[0018] Sixthly, this application provides a computer-readable storage medium comprising: computer software instructions; which, when executed in an electronic device, cause the electronic device to implement the training method of the rate prediction model of the first aspect and / or the rate prediction method of the third aspect.

[0019] In a seventh aspect, this application provides a computer program product comprising a computer program; when the computer program is run in an electronic device, the electronic device implements the training method of the rate prediction model of the first aspect and / or the rate prediction method of the third aspect.

[0020] The beneficial effects of the second to seventh aspects mentioned above are described in the corresponding description of the first aspect and will not be repeated here. Attached Figure Description

[0021] Figure 1 A schematic diagram illustrating the application environment of a rate prediction method provided in this application;

[0022] Figure 2 A flowchart illustrating a training method for a rate prediction model provided in this application;

[0023] Figure 3 A schematic diagram illustrating the process of constructing a rate prediction model provided in this application;

[0024] Figure 4 A schematic diagram of the structure of a training system for a rate prediction model provided in this application;

[0025] Figure 5 A flowchart illustrating a rate prediction method provided in this application;

[0026] Figure 6 A schematic diagram of the structure of a rate prediction system provided in this application;

[0027] Figure 7 This is a schematic diagram of the structure of an electronic device provided in this application. Detailed Implementation

[0028] The following is a detailed description of a call detail record (CDR) data recording method provided in this application, with reference to the accompanying drawings.

[0029] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0030] The terms "first" and "second," etc., used in the specification and drawings of this application are used to distinguish different objects or to distinguish different treatments of the same object, rather than to describe a specific order of objects.

[0031] Furthermore, the terms "comprising" and "having," and any variations thereof, used in the description of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the steps or units listed, but may optionally include other steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus.

[0032] It should be noted that in the embodiments of this application, the words "exemplary" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0033] To facilitate a clear description of the technical solutions of the embodiments of this application, the terms "first" and "second" are used in the embodiments of this application to distinguish the same or similar items with essentially the same function and effect. Those skilled in the art can understand that the terms "first" and "second" are not intended to limit the quantity or execution order.

[0034] In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0035] In recent years, as wireless communication networks have evolved towards 5G ultra-dense heterogeneous networks and 6G intelligent fully connected networks, network scenarios have exhibited three major characteristics: enhanced user mobility, more complex channel environments, and more dynamic scheduling strategies, leading to severe challenges in predicting wireless downlink rates.

[0036] Existing downlink rate prediction methods typically involve determining the rate by looking up tables or using the MCS-scheduling resource block mapping table in the 3GPP standard, processing it layer by layer across multiple modules such as the physical layer, link layer, and MAC layer, or using commercial simulation tools for chain-like derivation, that is, calculating the theoretical rate step by step based on scheduling resources, CQI mapping, modulation and coding rate, etc.

[0037] However, existing rate prediction methods suffer from drawbacks such as complex formulas, high computational overhead, and difficulty in reflecting application layer rate differences caused by scheduling strategy differences under similar reference signal-to-noise ratio conditions, resulting in inaccurate wireless downlink rate prediction.

[0038] Based on this, this application provides a training method for a rate prediction model. By acquiring the original simulation features of the simulation samples, data support can be provided for subsequent feature construction. Based on the set of simulation channel samples of the simulation samples, a context rate priority factor is determined, which can measure the data transmission advantage of the simulation samples under similar reference signal-to-noise ratio conditions, providing a basis for constructing a rate prediction feature vector with dynamic channel adaptability. Based on the context rate priority factor, wireless bearer channel coding, and reference signal-to-noise ratio, a rate prediction feature vector of the simulation samples is constructed. This method can retain the spatiotemporal correlation between key features while reducing feature redundancy through dimensionality reduction, providing a structured data support basis for the subsequent construction of the rate prediction model. The rate prediction feature vector of the simulation samples is input into a random forest model for training to obtain a rate prediction model for predicting wireless downlink rate. The accuracy of wireless downlink rate prediction can be improved based on multi-dimensional feature fusion and dynamic splitting optimization of regression decision trees.

[0039] The embodiments provided in this application will now be described in detail with reference to the accompanying drawings.

[0040] The rate prediction method provided in this application can be applied to, for example... Figure 1 The application environment shown. For example... Figure 1 As shown, the application environment includes a rate prediction device 101 and a terminal 102. The rate prediction device 101 includes a rate prediction system 600 and a rate prediction model training system 400. The rate prediction system 600 and the rate prediction model training system 400 are interconnected with the terminal 102.

[0041] In some embodiments, the rate prediction device 101 may be a server cluster consisting of multiple servers, a single server, a computer, or a processor or processing chip in a server or computer, etc. This application does not limit the specific device form of the rate prediction device 101. Figure 1 The rate prediction device 101 is used as an example of a single server.

[0042] In some embodiments, terminal 102 may be a device capable of generating wireless network simulation data and acquiring real-time data. Examples include base stations, mobile terminal devices (such as laptops and mobile phones), and edge computing nodes. This application does not limit the specific device form of terminal 102. Figure 1 The example shown is based on terminal 102 as a base station.

[0043] In some embodiments, terminal 102 is used to generate a wireless network simulation dataset and acquire real-time communication data generated by various application devices (e.g., mobile phones, smartphones, IoT devices (e.g., smart meters, environmental sensors), vehicle terminals, etc.) communicating in the wireless downlink.

[0044] Terminal 102 is used to transmit wireless network simulation datasets to the training system 400 of the rate prediction model, and to transmit real-time communication data (such as real-time channel state information (CSI), reference signal-to-noise ratio (RS-SINR), etc.) to the rate prediction system 600.

[0045] The rate prediction model training system 400 constructs features and trains the model based on the wireless network simulation data transmitted by the terminal 102, thereby building the rate prediction model. The rate prediction model training system 400 returns the model construction results to the terminal 102.

[0046] The rate prediction system 600 predicts the downlink application layer rate based on the real-time communication data transmitted by the terminal 102 and the rate prediction model transmitted by the rate prediction model training system 400. The rate prediction system 600 returns the prediction result of the downlink application layer rate to the terminal 102.

[0047] It should be noted that the system architecture described in the embodiments of this application is for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and does not constitute a limitation on the technical solutions provided in the embodiments of this application. As those skilled in the art will know, with the evolution of system architecture, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0048] See Figure 2 This application provides a method for training a rate prediction model, comprising:

[0049] S201. Obtain the original simulation features of the simulation samples; wherein, the simulation samples are the structured representation of the original simulation data generated by the downlink wireless link simulation operation; the original simulation features include the reference signal-to-noise ratio, wireless bearer channel coding, and downlink application layer rate.

[0050] The original simulation data consists of unstructured data in binary log or text format output by simulation tools (such as NS-3) when a wireless communication system (such as 4G LTE, 5G NR, etc.) is running in a downlink wireless link simulation scenario.

[0051] The simulation samples are simulation data records extracted from the wireless network simulation dataset output from the downlink wireless link simulation scenario of a wireless communication system. These simulation data records are structured tabular data obtained by parsing the raw simulation data generated during the downlink wireless link simulation. The simulation data records can be stored in CSV or Excel format.

[0052] The wireless network simulation dataset consists of multiple simulation samples organized in a time series. Each simulation sample corresponds to a complete downlink wireless link state snapshot at a point in time, including but not limited to original simulation feature fields such as reference signal-to-noise ratio, wireless bearer channel coding, and downlink application layer rate.

[0053] The wireless network simulation dataset consists of discrete time series data.

[0054] Simulation data records include, but are not limited to, timestamps, modulation schemes, number of MIMO layers, reference signal-to-noise ratio, radio bearer channel coding, coding efficiency, and downlink application layer rate.

[0055] The original simulation features are those extracted from simulation samples using simulation data processing software (such as MATLAB, NS-3) or scripts (such as Python's pandas library). These original simulation features include, but are not limited to, the reference signal-to-noise ratio, radio bearer channel coding, and downlink application layer rate.

[0056] The reference signal-to-noise ratio (SNR) is the SNR of the receiver in the downlink during a simulation scenario, used to reflect signal quality.

[0057] Wireless bearer channel coding refers to the channel coding method (such as Turbo code, LDPC code, Polar code, etc.) used in the downlink in the simulation scenario. It is used for error correction and to improve transmission reliability, reflecting the allocation of system resources.

[0058] The downlink application layer rate is the actual application layer data transmission rate received in the wireless downlink in the simulation scenario.

[0059] For example, the application layer data transmission rate can be the effective data transmission rate that can actually be achieved at the application layer when transmitting data from the base station to the user equipment in a cellular network.

[0060] S202. Based on the simulated channel sample set, determine the context rate priority factor; the context rate priority factor is used to characterize the relative priority of the downlink application layer rate of the simulated sample under the condition of similar signal-to-noise ratio of the reference signal.

[0061] The simulated channel sample set includes multiple key simulated channel samples that meet the sample selection rules. The sample selection rules include: the difference between the signal-to-noise ratio (SNR) of the simulated sample and the reference signal does not exceed the preset SNR difference.

[0062] The reference signal-to-noise ratio difference is the absolute value of the difference between the reference signal-to-noise ratio of the simulated sample and the reference signal-to-noise ratio of the simulated channel sample.

[0063] The preset signal-to-noise ratio difference is a critical value set in advance based on actual scenario requirements (such as scheduling granularity and channel change rate) and is used to screen key simulation channel samples.

[0064] For example, the preset signal-to-noise ratio difference can be 0.25dB.

[0065] The key simulation channel sample is a simulation channel sample in which the absolute value of the difference between the reference signal signal-to-noise ratio and the reference signal signal-to-noise ratio of the simulation sample does not exceed the preset signal-to-noise ratio difference.

[0066] The simulated channel sample is a record of simulated channel parameters extracted from the wireless network simulation dataset under the downlink wireless link simulation scenario of a wireless communication system. This simulated channel parameter record includes, but is not limited to, timestamps, reference signal-to-noise ratio, received signal strength indication, signal-to-interference-plus-noise ratio, modulation and coding parameters, transmission rate and performance indicators (such as downlink application layer rate and physical layer rate), and simulation scenario identifiers.

[0067] For example, the formula for the set of simulated channel samples for the i-th simulated sample is as follows:

[0068] N δ (x i )={j:|x j -x i |≤δ} Formula (1)

[0069] Where, N δ (x i Let x represent the set of simulated channel samples for the i-th simulated sample; i Let x represent the signal-to-noise ratio of the reference signal for the i-th simulated sample. i ∈R; δ represents the preset signal-to-noise ratio difference; x j Let S be the reference signal-to-noise ratio of the j-th key simulated channel sample.

[0070] In some embodiments, determining a context rate priority factor based on a set of simulated channel samples includes: determining the average context simulation rate based on the simulated downlink rate of each key simulated channel sample in the set of simulated channel samples; and determining the context rate priority factor based on the downlink application layer rate and the average context simulation rate of the simulated samples.

[0071] The simulated downlink rate is the downlink application layer rate included in the key simulated channel samples. The simulated downlink rate is used to reflect the actual transmission performance under different simulated channel conditions.

[0072] For example, the formula for calculating the context simulation mean rate of the i-th simulation sample is as follows:

[0073]

[0074] Where, μ i V is the context simulation mean rate for the i-th simulation sample; j Let be the simulated downlink rate of the j-th key simulated channel sample.

[0075] In some embodiments, the context rate priority factor is calculated as follows:

[0076]

[0077] Where CRP is the context rate priority factor; V is the downlink application layer rate of the simulation sample; μ represents the average rate of the context simulation; and ∈ is a constant for avoiding division-by-zero anomalies, with a value of 10. -6 .

[0078] The context rate priority factor characterizes the relative priority of the downlink rate of the simulated sample among samples with similar channel quality. A larger value of the context rate priority factor indicates that the downlink application layer rate of the simulated sample is better scheduled among samples with similar channel quality, and is more likely to be selected first during scheduling, thereby improving the scheduling effect.

[0079] Therefore, based on the downlink rate of the channel sample with a similar signal-to-noise ratio to the reference signal of the simulation sample, a context rate priority factor is constructed. This factor can measure the data transmission advantage of the simulation sample under the condition of similar signal-to-noise ratio of the reference signal, and provide a valid basis for the subsequent feature vector construction.

[0080] S203. Based on the context rate priority factor, wireless bearer channel coding, and reference signal signal-to-noise ratio, a rate prediction feature vector for the simulation sample is constructed.

[0081] In some embodiments, a rate prediction feature vector for a simulation sample is constructed based on a context rate priority factor, radio bearer channel coding, and a reference signal signal-to-noise ratio (SNR). This includes: determining the bearer efficiency of the simulation sample based on the radio bearer channel coding; and arranging the context rate priority factor, radio bearer channel coding, bearer efficiency, and reference signal SNR according to a preset feature arrangement order to generate a rate prediction feature vector for the simulation sample.

[0082] Among these, the reference signal-to-noise ratio (SNR) is strongly correlated with the downlink application layer rate. The radio bearer channel coding is strongly correlated with bearer efficiency.

[0083] The context rate priority factor has low correlation with wireless bearer channel coding, bearer efficiency, and reference signal-to-noise ratio. The context rate priority factor is used to improve model explanatory power and nonlinear fitting capability.

[0084] For example, the preset feature arrangement order is a pre-defined rule for sorting context rate priority factor, radio bearer channel coding, bearer efficiency, and reference signal-to-noise ratio. The preset feature arrangement order can be: [reference signal-to-noise ratio, context rate priority factor, radio bearer channel coding, bearer efficiency].

[0085] For example, the rate prediction feature vector X of the i-th simulated sample i It can be:

[0086] X i =[x i CRP i Bearer Index, η b ] Formula (4)

[0087] Where, x i CRP is the signal-to-noise ratio of the reference signal for the i-th simulated sample. i η is the context rate priority factor for the i-th simulation sample; BearerIndex is the radio bearer channel coding for the i-th simulation sample; b Let be the load-bearing efficiency of the i-th simulation sample.

[0088] Therefore, by arranging the context rate priority factor, wireless bearer channel coding, bearer efficiency, and reference signal signal-to-noise ratio based on a preset feature arrangement order, an enhanced feature vector can be generated. While preserving the spatiotemporal correlation between key features, feature redundancy is reduced through dimensionality reduction, providing structured data support for the subsequent construction of rate prediction models.

[0089] S204. Input the rate prediction feature vector of the simulation sample into the random forest regression model for training to obtain the rate prediction model. The rate prediction model is used to predict the downlink application layer rate.

[0090] Among them, the Random Forest Regression Model is an ensemble learning model built on the Bagging ensemble framework. The Random Forest Regression Model reduces variance by training multiple independent regression decision trees in parallel and combining the prediction results of all trees (such as by taking the mean), thereby effectively fitting the complex relationships between nonlinear features. It has the advantages of strong robustness, good noise resistance, and low overfitting.

[0091] The rate prediction model is a strong predictor built through multiple rounds of iterative training based on a random forest regression model and rate prediction feature vectors. The rate prediction model based on the random forest regression model can accurately capture the nonlinear relationship of the rate prediction feature vectors, achieving high-precision, low-latency prediction of downlink application layer rates in complex wireless communication environments.

[0092] For example, suppose the prediction sample of the rate prediction model is a feature vector X. i If the prediction function is the ensemble average of multiple regression decision trees, then the output of the rate prediction model, i.e., the feature vector X, is... i wireless downlink rate The calculation formula is as follows:

[0093]

[0094] Where B is the number of regression decision trees; T b This is the output of the b-th regression decision tree. Each regression decision tree can be constructed by bootstrapping from the training dataset to create a sample subset, and the training space can be recursively partitioned on the feature subset to form an ensemble of weak learners.

[0095] Therefore, by training a random forest regression model based on the rate prediction feature vector, a rate prediction model can be constructed, which can provide an intelligent prediction engine for rate prediction. This rate prediction model can accurately capture the dynamic impact of complex interactions between features on the downlink rate, and achieve accurate prediction of the downlink application layer rate.

[0096] Therefore, by obtaining the original simulation features of the simulation samples, data support can be provided for subsequent feature construction. Based on the set of simulation channel samples, the context rate priority factor can be determined to measure the data transmission advantage of the simulation samples under similar reference signal-to-noise ratio conditions, providing a basis for constructing a rate prediction feature vector with dynamic channel adaptability. Based on the context rate priority factor, wireless bearer channel coding, and reference signal-to-noise ratio, the rate prediction feature vector of the simulation samples can be constructed. This can preserve the spatiotemporal correlation between key features while reducing feature redundancy through dimensionality reduction, providing a structured data support basis for the subsequent construction of the rate prediction model. The rate prediction feature vector of the simulation samples can be input into a random forest model for training to obtain a rate prediction model for predicting the wireless downlink rate. The accuracy of the wireless downlink rate prediction can be improved based on multi-dimensional feature fusion and dynamic splitting optimization of the regression decision tree.

[0097] In one embodiment, determining the carrying efficiency of a simulation sample based on wireless bearer channel coding includes: determining the carrying efficiency of the simulation sample based on wireless bearer channel coding and a preset mapping table; the preset mapping table contains multiple entries mapping relationships between wireless bearer channel coding and carrying efficiency.

[0098] The preset mapping table can be a table pre-established based on factors such as different bearer service types, Quality of Service (QoS) requirements, and channel conditions, combined with the 3GPP standard framework. The preset mapping table contains multiple entries mapping the relationship between radio bearer channel coding and bearer efficiency.

[0099] Bearer efficiency is used to characterize the physical layer modulation capability and resource utilization efficiency. Bearer efficiency is closely related to the physical modulation scheme (such as QPSK, 16QAM, 64QAM, 256QAM) and can be used to reflect the ability of a wireless communication system to select an efficient modulation scheme under channel quality conditions.

[0100] For example, the formula for obtaining the load capacity is as follows:

[0101] η b =Efficiency(b), b∈Bearer Index formula (6)

[0102] Where, η b For carrying efficiency; b∈N + It is a set of coding schemes for wireless bearer channels.

[0103] Therefore, based on the wireless bearer channel coding mapping bearer efficiency, we can introduce the ability to model the impact of bearer characteristics on rate, providing a valid basis for subsequent model construction.

[0104] In one embodiment, test simulation data is input into the rate prediction model to obtain test simulation prediction results; based on the test simulation prediction results and set performance prediction indicators, the prediction performance evaluation results of the rate prediction model are determined.

[0105] Therefore, by conducting performance tests and evaluations on the rate prediction model, the model's prediction accuracy and generalization ability for wireless downlink rates can be quantified, thereby providing a data-driven decision-making basis for subsequent model iteration and optimization.

[0106] The test simulation data is simulation data obtained in advance in the simulation scenario of the downlink wireless link of the wireless communication system to verify the prediction performance of the rate prediction model.

[0107] The test simulation prediction result is the predicted wireless downlink rate output by inputting the test simulation data into the rate prediction model.

[0108] For example, performance prediction metrics are set to quantify the predictive performance of the rate prediction model. These performance prediction metrics can be the prediction error RMSE, the coefficient of determination R, etc. 2 .

[0109] The formula for calculating the prediction error RMSE is expressed as follows:

[0110]

[0111] Coefficient of determination R 2 The calculation formula is expressed as:

[0112]

[0113] Among them, c a Let a represent the actual wireless downlink rate corresponding to the a-th test simulation data, where a = 1, 2, 3, ..., n, and n represents the total number of test simulation data. This represents the predicted wireless downlink rate corresponding to the a-th test simulation data output by the rate prediction model. It represents the average of the corresponding real wireless downlink rates for all test simulation data.

[0114] Predictive performance evaluation results are based on set performance prediction metrics (such as RMSE, R...). 2 The quantitative evaluation value of the rate prediction model is used to determine whether the rate prediction model meets the prediction performance requirements.

[0115] In some embodiments, see Figure 3 The process for constructing a rate prediction model can be as follows:

[0116] S301. Simulation data import and data processing.

[0117] The simulation data refers to wireless network simulation data in CSV or Excel format output by the simulation project.

[0118] The simulation project can be a modeling and simulation project for intelligent performance of wireless networks for optimizing the Quality of Service (QoS) and User Experience (QoE) of cellular systems. It aims to generate multi-scenario wireless network datasets through high-fidelity simulation to support intelligent optimization of downlink application layer rate prediction.

[0119] Data processing involves extracting fields such as the reference signal-to-noise ratio, radio bearer channel coding, coding efficiency, and actual downlink application layer rate from the wireless network simulation dataset.

[0120] Data processing also includes analyzing the scatter distribution of the reference signal-to-noise ratio and the actual downlink application layer rate at the receiver in the downlink.

[0121] S302, Feature Enhancement Design.

[0122] Among them, the feature enhancement design can be: after determining the context rate priority factor and the bearer efficiency, the rate prediction feature vector is constructed by combining the reference signal-to-noise ratio of the receiver in the downlink and the radio bearer channel coding.

[0123] S303, Model Building and Training.

[0124] The model construction and training can be performed as follows: a random forest regression model is trained using rate prediction feature vectors to construct a rate prediction model. The purpose of constructing the rate prediction model is to achieve high-precision downlink application layer rate prediction.

[0125] S304, Model Evaluation and Visualization.

[0126] Among them, model evaluation can be carried out by using two indicators, prediction error and coefficient of determination, to evaluate the prediction performance of the rate prediction model.

[0127] Model visualization can be used to construct a prediction comparison chart to visualize how well the predicted wireless downlink rate values ​​output by the model fit the actual wireless downlink rate values.

[0128] For example, based on the actual wireless downlink rate and the predicted wireless downlink rate corresponding to the test simulation data, a prediction comparison chart is constructed with the actual wireless downlink rate as the horizontal axis and the predicted wireless downlink rate as the vertical axis.

[0129] Based on the prediction comparison chart, the degree of fit between the predicted wireless downlink rate value and the actual wireless downlink rate value can be determined. When the degree of fit between the predicted and actual wireless downlink rate values ​​is high, the rate prediction feature vector generated based on the context rate priority factor, wireless bearer channel coding, bearer efficiency, and reference signal-to-noise ratio is determined. The constructed rate prediction model exhibits excellent prediction performance and is suitable for different channel quality levels and scheduling strategies.

[0130] Therefore, by establishing a prediction comparison chart based on the predicted and actual values ​​of the wireless downlink rate, the prediction performance of the rate prediction model can be intuitively displayed, thereby providing a data-driven decision-making basis for subsequent model iteration and optimization.

[0131] As can be seen, the above mainly describes the solutions provided by the embodiments of this application from a methodological perspective. To achieve the above functions, the embodiments of this application provide corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the modules and algorithm steps of the various examples described in the embodiments disclosed herein, the embodiments of this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed by hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention.

[0132] This application embodiment can divide the training system of the rate prediction model into functional modules according to the above method example. For example, each function can be divided into a separate functional module, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware or as a software functional module. Optionally, the module division in this application embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0133] See Figure 4 This application provides a training system 400 for a rate prediction model, comprising:

[0134] The feature acquisition module 401 is used to acquire the original simulation features of the simulation samples; wherein, the simulation samples are a structured representation of the original simulation data generated by the downlink wireless link simulation operation; the original simulation features include the reference signal-to-noise ratio, wireless bearer channel coding, and downlink application layer rate.

[0135] The priority analysis module 402 is used to determine the context rate priority factor based on the set of simulated channel samples. The context rate priority factor is used to characterize the relative priority of the downlink application layer rate of the simulated sample under the condition of similar reference signal-to-noise ratio. The set of simulated channel samples includes multiple key simulated channel samples that meet the sample selection rules. The sample selection rules include: the difference between the simulated sample and the reference signal-to-noise ratio does not exceed the preset signal-to-noise ratio difference.

[0136] The feature construction module 403 is used to construct the rate prediction feature vector of the simulation sample based on the context rate priority factor, radio bearer channel coding and reference signal signal-to-noise ratio.

[0137] The model building module 404 is used to input the rate prediction feature vector of the simulation sample into the random forest regression model for training to obtain the rate prediction model, which is used to predict the downlink application layer rate.

[0138] In some embodiments, the priority analysis module 402 is specifically used to determine the context simulation average rate based on the simulated downlink rate of each key simulated channel sample in the simulated channel sample set; and to determine the context rate priority factor based on the downlink application layer rate and the context simulation average rate of the simulated samples.

[0139] In some embodiments, the feature construction module 403 is specifically used to determine the carrying efficiency of the simulation sample based on the wireless bearer channel coding; and to arrange the context rate priority factor, wireless bearer channel coding, carrying efficiency and reference signal signal-to-noise ratio based on a preset feature arrangement order to generate the rate prediction feature vector of the simulation sample.

[0140] In some embodiments, the feature construction module 403 is further configured to determine the carrying efficiency of the simulation sample based on the wireless bearer channel coding and a preset mapping table; the preset mapping table contains multiple entries for the mapping relationship between wireless bearer channel coding and carrying efficiency.

[0141] See Figure 5 A rate prediction method provided in this application includes:

[0142] S501. Obtain the rate prediction feature vector of the application terminal in the wireless downlink; the application terminal is the user equipment in the wireless downlink; the rate prediction feature vector of the application terminal includes the context rate priority factor, wireless bearer channel coding, bearer efficiency, and reference signal-to-noise ratio.

[0143] The application terminal is a terminal device that receives downlink signals and processes data. Application terminals include, but are not limited to, mobile phones, smartphones, IoT devices (such as smart meters and environmental sensors), and vehicle terminals.

[0144] For example, obtaining the rate prediction feature vector of the application terminal includes: determining a context rate priority factor based on the channel sample set of the application terminal; the context rate priority factor is used to characterize the relative priority of the downlink application layer rate of the application terminal under the condition of similar reference signal-to-noise ratio; the channel sample set includes multiple key channel samples that meet the sample selection rules; the sample selection rules include: the difference between the reference signal-to-noise ratio and the application terminal does not exceed a preset signal-to-noise ratio difference.

[0145] Among them, the key channel samples are those whose absolute value of the difference between the reference signal-to-noise ratio (SNR) and the application terminal's reference signal-to-noise ratio (SNR) does not exceed a preset SNR difference. These channel samples are channel parameter records obtained from real downlink wireless link communication scenarios. These channel parameter records include, but are not limited to, timestamps, reference signal-to-noise ratio (SNR), received signal strength indication, signal-to-interference-plus-noise ratio (SINR), modulation and coding parameters, transmission rate and performance indicators, and scenario identification information.

[0146] For example, determining the context rate priority factor based on the channel sample set of the application terminal includes: determining the average context rate based on the downlink rate of each key channel sample in the channel sample set; and determining the context rate priority factor based on the downlink application layer rate and the average context rate of the application terminal.

[0147] The downlink rate refers to the downlink application layer rate contained within the channel sample. This downlink rate is used to reflect the actual transmission performance under different channel conditions.

[0148] The context-averaged rate is the rate obtained by averaging the downlink rates of all acquired key channel samples.

[0149] For example, suppose the downlink application layer rate of the application terminal at the current moment is v. x The context mean rate is μ x

[0150] Then, the current application terminal's context rate priority factor (CRP) x The calculation formula is as follows:

[0151]

[0152] Wherein, ∈ is a constant for avoiding division by zero anomalies, with a value of 10⁻⁶.

[0153] For example, obtaining the rate prediction feature vector of the application terminal further includes: determining the bearer efficiency of the application terminal based on the radio bearer channel coding; and arranging the context rate priority factor, radio bearer channel coding, bearer efficiency, and reference signal-to-noise ratio of the application terminal based on a preset feature arrangement order to generate the rate prediction feature vector of the application terminal.

[0154] For example, determining the bearer efficiency of an application terminal based on the wireless bearer channel coding includes: determining the bearer efficiency of the application terminal based on the wireless bearer channel coding of the application terminal and a preset mapping table; the preset mapping table contains multiple mapping relationship entries between wireless bearer channel coding and bearer efficiency.

[0155] S502. Input the rate prediction feature vector of the application terminal into the rate prediction model to obtain the corresponding downlink application layer rate; wherein, the rate prediction model is obtained through the training method of the above-mentioned rate prediction model.

[0156] Therefore, by constructing the rate prediction feature vector of the application device, the spatiotemporal correlation between key features can be preserved while the feature redundancy is reduced through dimensionality reduction, providing a basis for subsequent rate prediction. By inputting the rate prediction feature vector of the application device into the rate prediction model, the downlink application layer rate at future moments can be accurately predicted based on the dynamic splitting optimization of multi-dimensional feature fusion and regression decision tree, thereby providing a dynamic decision basis for subsequent wireless resource scheduling, link adaptation, etc.

[0157] As can be seen, the above mainly describes the solutions provided by the embodiments of this application from a methodological perspective. To achieve the above functions, the embodiments of this application provide corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the modules and algorithm steps of the various examples described in the embodiments disclosed herein, the embodiments of this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed by hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention.

[0158] This application embodiment can divide the rate prediction system into functional modules according to the above method example. For example, each function can be divided into its own functional module, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware or as a software functional module. Optionally, the module division in this application embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0159] See Figure 6 A rate prediction system 600 provided in this application embodiment includes:

[0160] The data acquisition module 601 is used to acquire the rate prediction feature vector of the application terminal in the wireless downlink; the application terminal is the user equipment in the wireless downlink; the rate prediction feature vector of the application terminal includes the context rate priority factor, the wireless bearer channel coding, the bearer efficiency, and the reference signal-to-noise ratio.

[0161] The rate prediction module 602 is used to input the rate prediction feature vector of the application terminal into the rate prediction model to obtain the corresponding downlink application layer rate; wherein, the rate prediction model is obtained through the training method of the above-mentioned rate prediction model.

[0162] In the case of implementing the functions of the integrated modules described above in hardware, this embodiment of the invention provides a possible structural schematic diagram of the electronic device involved in the above embodiments. For example... Figure 7 As shown, the electronic device 700 includes: a processor 702, a communication interface 703, and a bus 704. Optionally, the electronic device 700 may also include a memory 701.

[0163] Processor 702 may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 702 may be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 702 may also be a combination that implements computing functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0164] The communication interface 703 is used to connect to other devices via a communication network. This communication network can be Ethernet, wireless access network, wireless local area network (WLAN), etc.

[0165] The memory 701 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), disk storage medium or other magnetic storage device, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.

[0166] As one possible implementation, the memory 701 can exist independently of the processor 702. The memory 701 can be connected to the processor 702 via a bus 704 and is used to store instructions or program code. When the processor 702 calls and executes the instructions or program code stored in the memory 701, it can implement the training method and / or rate prediction method of the rate prediction model provided in this embodiment of the invention.

[0167] In another possible implementation, the memory 701 can also be integrated with the processor 702.

[0168] The 704 bus can be an extended industry standard architecture (EISA) bus, etc. The 704 bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 7 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0169] Through the above description of the implementation methods, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the service calling device can be divided into different functional modules to complete all or part of the functions described above.

[0170] This application also provides a computer-readable storage medium. All or part of the processes in the above method embodiments can be executed by computer instructions instructing related hardware. The program can be stored in the aforementioned computer-readable storage medium, and when executed, it can include the processes of the above method embodiments. The computer-readable storage medium can be any of the foregoing embodiments or memory. The aforementioned computer-readable storage medium can also be an external storage device of the aforementioned service invocation device, such as a pluggable hard drive, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the aforementioned service invocation device. Further, the aforementioned computer-readable storage medium can include both internal storage units of the aforementioned service invocation device and external storage devices. The aforementioned computer-readable storage medium is used to store the aforementioned computer program and other programs and data required by the aforementioned service invocation device. The aforementioned computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0171] This application also provides a computer program product comprising a computer program that, when run on a computer, causes the computer to execute the training method and / or rate prediction method of any of the rate prediction models provided in the above embodiments.

[0172] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A training method for a rate prediction model, characterized in that, The method comprises: obtaining original simulation features of simulation samples; wherein the simulation samples are structured representations of original simulation data generated by downlink wireless link simulation running; the original simulation features comprise reference signal signal-to-noise ratio, wireless bearer channel coding and downlink application layer rate; determining a context rate priority factor based on a simulation channel sample set of the simulation samples; the context rate priority factor is used to represent the relative priority of the downlink application layer rate of the simulation samples under the condition of similar reference signal signal-to-noise ratio; the simulation channel sample set comprises a plurality of key simulation channel samples satisfying a sample screening rule; the sample screening rule comprises that the difference of the reference signal signal-to-noise ratio of the simulation samples does not exceed a preset signal-to-noise ratio difference value; constructing a rate prediction feature vector of the simulation samples based on the context rate priority factor, the wireless bearer channel coding and the reference signal signal-to-noise ratio; inputting the rate prediction feature vector of the simulation samples into a random forest regression model for training to obtain a rate prediction model, wherein the rate prediction model is used to predict the downlink application layer rate. 2.The method of claim 1, wherein, The method of determining the context rate priority factor based on the simulation channel sample set of the simulation samples comprises: determining a context simulation mean rate based on the simulation downlink rate of each key simulation channel sample in the simulation channel sample set; determining the context rate priority factor based on the downlink application layer rate of the simulation samples and the context simulation mean rate. 3.The method of Claim 2, wherein, The calculation formula of the context rate priority factor is as follows: wherein CRP is the context rate priority factor; V is the downlink application layer rate of the simulation samples; μ represents the context simulation mean rate; ∈ is a constant for avoiding division by zero exception, and the value is 10-6. 4.The method of claim 1, wherein, The method of constructing the rate prediction feature vector of the simulation samples based on the context rate priority factor, the wireless bearer channel coding and the reference signal signal-to-noise ratio comprises: determining the bearer efficiency of the simulation samples based on the wireless bearer channel coding; arranging the context rate priority factor, the wireless bearer channel coding, the bearer efficiency and the reference signal signal-to-noise ratio based on a preset feature arrangement order to generate the rate prediction feature vector of the simulation samples. 5.The method of Claim 4, wherein, The method of determining the bearer efficiency of the simulation samples based on the wireless bearer channel coding comprises: determining the bearer efficiency of the simulation samples based on the wireless bearer channel coding and a preset mapping table; the preset mapping table comprises a plurality of mapping relationship entries of the wireless bearer channel coding and the bearer efficiency. 6.A system for training a rate prediction model, characterized in that, The method comprises: a feature acquisition module configured to obtain original simulation features of simulation samples; wherein the simulation samples are structured representations of original simulation data generated by downlink wireless link simulation running; the original simulation features comprise reference signal signal-to-noise ratio, wireless bearer channel coding and downlink application layer rate; The priority analysis module is configured to determine a context rate priority factor based on a set of simulation channel samples of the simulation sample, wherein the context rate priority factor is used to represent a relative priority of a downlink application layer rate of the simulation sample under a similar condition of a reference signal signal-to-noise ratio, and the set of simulation channel samples comprises a plurality of key simulation channel samples satisfying a sample screening rule, and the sample screening rule comprises that a difference of the reference signal signal-to-noise ratio of the simulation sample does not exceed a preset signal-to-noise ratio difference; The feature construction module is configured to construct a rate prediction feature vector of the simulation sample based on the context rate priority factor, the wireless bearer channel coding, and the reference signal signal-to-noise ratio; The model construction module is configured to input the rate prediction feature vector of the simulation sample into a random forest regression model for training to obtain a rate prediction model, and the rate prediction model is used to predict a downlink application layer rate.

7. A rate prediction method characterized by, The method comprises the following steps: The application terminal is a user equipment in a wireless downlink. The rate prediction feature vector of the application terminal comprises a context rate priority factor, a wireless bearer channel coding, a bearer efficiency, and a reference signal signal-to-noise ratio. The rate prediction model is obtained by inputting the rate prediction feature vector of the application terminal into the rate prediction model, and the rate prediction model is obtained by training the rate prediction model according to the training method of the rate prediction model in any one of claims 1 to 5.

8. A rate prediction system characterized by, The method comprises the following steps: The data acquisition module is configured to acquire a rate prediction feature vector of an application terminal, and the application terminal is a user equipment in a wireless downlink. The rate prediction feature vector of the application terminal comprises a context rate priority factor, a wireless bearer channel coding, a bearer efficiency, and a reference signal signal-to-noise ratio. The rate prediction module is configured to input the rate prediction feature vector of the application terminal into a rate prediction model to obtain a corresponding downlink application layer rate, and the rate prediction model is obtained by training the rate prediction model according to the training method of the rate prediction model in any one of claims 1 to 5.

9. An electronic device, comprising: The computer device comprises a processor and a memory, and the processor is coupled to the memory; the memory is used to store computer instructions, and the computer instructions are loaded and executed by the processor to enable the computer device to implement the training method of the rate prediction model in any one of claims 1 to 5 and / or the rate prediction method in claim 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium comprises computer execution instructions, and when the computer execution instructions run on a computer, the computer is enabled to execute the training method of the rate prediction model in any one of claims 1 to 5 and / or the rate prediction method in claim 7.

11. A computer program product, characterised in that, The computer program product comprises a computer program, and when the computer program runs on an electronic device, the electronic device is enabled to execute the training method of the rate prediction model in any one of claims 1 to 5 and / or the rate prediction method in claim 7.

Images