Carrier frequency resource allocation method and system, device, medium, and program product

Abstract

A carrier frequency resource allocation method and system, a device, a medium, and a program product. The method comprises: acquiring covered population characteristics of a target base station and a network quality classification result of population covered by the target base station, and determining feature information of the target base station on the basis of the covered population characteristics and the network quality classification result; predicting network utilization rate information of the target base station in the next time period on the basis of the feature information; determining uplink and downlink carrier frequency increments and decrements of the target base station on the basis of network utilization rate information of the target base station in the current time period and the network utilization rate information of the target base station in the next time period; and determining a total carrier frequency increment or decrement of the target base station on the basis of the uplink and downlink carrier frequency increments and decrements and a carrier frequency increment caused by service migration, and performing carrier frequency resource allocation by applying the total carrier frequency increment or decrement. Thus, the accuracy of carrier frequency resource allocation can be effectively improved, and the problems of resource waste or insufficient user assurance are avoided.

Description

Carrier frequency resource allocation method, system, device, medium and program product

[0001] Cross-reference to related applications

[0002] The present disclosure is based on and claims priority from Chinese Patent Application No. 202410433623.1 filed on April 11, 2024, the entire contents of which are incorporated herein by reference. TECHNICAL FIELD

[0003] The present disclosure relates to the technical field of wireless networks, and in particular to a carrier frequency resource allocation method, system, device, medium and program product. BACKGROUND

[0004] With the large-scale construction of 5G networks, the coverage target has been achieved, and the focus of future network planning will gradually shift to the protection of capacity perception needs. In the 5G era, the explosive growth of network traffic and diverse business needs pose higher challenges to carrier frequency resource allocation. Therefore, how to accurately predict network utilization, intelligently calculate carrier frequency resource demand, and realize dynamic and accurate deployment of 5G network carrier frequency resources has become an important issue in current network planning.

[0005] Existing carrier frequency allocation techniques are mostly based on waveform coding or convex optimization function models of communication systems. Although they achieve optimal allocation of resources to some extent, they often aim to maximize the rate of a single user, making it difficult to fully reflect the overall network carrying and operating efficiency. In addition, the application range of theoretical algorithms is limited, and their accuracy and adaptability are limited for complex and variable network environments, limiting their ability to improve the accuracy of carrier frequency resource allocation.

[0006] Intelligent utilization prediction usually converts network utilization prediction into a time series prediction model for modeling and solving. However, traditional time series prediction algorithms such as ARIMA (Autoregressive Integrated Moving Average model) are often based on linear models and are difficult to accurately describe the complexity and nonlinearity of network utilization changes. In particular, the periodicity and trend of individual cell utilization are not significant, making the application effect of traditional algorithms greatly discounted, resulting in low prediction accuracy for long periods and affecting carrier frequency resource allocation.

[0007] Therefore, the network utilization prediction accuracy of the above related technologies is not high, and the carrier frequency resource allocation is not accurate. SUMMARY

[0008] The present disclosure provides a carrier frequency resource allocation method, system, device, medium and program product to solve the problem of low prediction accuracy of network utilization and inaccurate carrier frequency resource allocation in the related art.

[0009] The present disclosure provides a carrier frequency resource allocation method, comprising: obtaining coverage population characteristics of a target base station and network quality division results of the population covered by the target base station, and determining feature information of the target base station based on the coverage population characteristics and the network quality division results; predicting network utilization information of the target base station in the next period based on the feature information of the target base station; determining uplink and downlink carrier frequency increment and decrement of the target base station based on network utilization information of the target base station in the current period and network utilization information of the target base station in the next period; and determining total carrier frequency increment and decrement of the target base station based on the uplink and downlink carrier frequency increment and decrement and carrier frequency increment caused by service migration, and applying the total carrier frequency increment and decrement for carrier frequency resource allocation.

[0010] In some embodiments, the obtaining of the coverage population characteristics of the target base station and the network quality division results of the population covered by the target base station comprises: obtaining the area population and the area size of the area where the target base station is located; determining the coverage population size of the target base station based on the single-station coverage area of the target base station, the area population and the area size, wherein the coverage population characteristics comprise the single-station coverage area and the coverage population size; selecting a plurality of sampling points within the single-station coverage area, and performing network quality division on the coverage population size based on the channel signal strength and the uplink and downlink rate of each sampling point to obtain the network quality division results.

[0011] In some embodiments, the network quality division on the coverage population size based on the channel signal strength and the uplink and downlink rate of each sampling point to obtain the network quality division results comprises: calculating the network quality division threshold of each sampling point based on the network modulation order distribution characteristics of each sampling point in the current period; and performing network quality division on the coverage population size based on the network quality division threshold of each sampling point, and the channel signal strength and the uplink and downlink rate of each sampling point to obtain the network quality division results.

[0012] In some embodiments, the determination of the feature information of the target base station based on the coverage population characteristics and the network quality division results comprises: removing outliers from the network quality division results, and performing data aggregation on the network quality division results after removing outliers, the coverage population characteristics and the network traffic data of the target base station to obtain the feature information of the target base station.

[0013] In some embodiments, the predicting the network utilization information of the target base station in the next time period based on the feature information of the target base station comprises: inputting the feature information into a network utilization prediction model corresponding to the target base station to obtain the network utilization information of the target base station in the next time period output by the network utilization prediction model; and the network utilization prediction model is obtained by training based on sample feature information and sample network utilization information of the target base station and model accuracy evaluation.

[0014] In some embodiments, the objective function of the network utilization prediction model comprises a training loss function based on a square root of a relative error and a regularization loss function, and the training step of the network utilization prediction model comprises: obtaining a training data set and a test data set of the target base station, wherein the training data set comprises sample feature information and sample network utilization information, and the test data set comprises test feature information and test network utilization information; inputting the sample feature information into an initial network utilization prediction model to obtain first predicted network utilization information output by the initial network utilization prediction model; performing parameter iteration on the initial network utilization prediction model based on the first predicted network utilization information and the sample network utilization information to obtain a pre-trained network utilization prediction model; performing model accuracy evaluation on the pre-trained network utilization prediction model by applying the test feature information and the test network utilization information to obtain the network utilization prediction model.

[0015] In some embodiments, the performing model accuracy evaluation on the pre-trained network utilization prediction model by applying the test feature information and the test network utilization information to obtain the network utilization prediction model comprises: inputting the test feature information into the pre-trained network utilization prediction model to obtain second predicted network utilization information output by the pre-trained network utilization prediction model; comparing the second predicted network utilization information with the test network utilization information and determining a model accuracy evaluation result of the pre-trained network utilization prediction model based on a comparison result; in a case where the model accuracy evaluation result is inaccurate, rearranging features of the test feature information and removing interference features to obtain new test feature information; and performing model accuracy evaluation on the pre-trained network utilization prediction model again by applying the new test feature information until the model accuracy evaluation result is accurate, and obtaining the network utilization prediction model.

[0016] In some embodiments, the determining the uplink and downlink carrier frequency increment and decrement amount of the target base station based on the network utilization information of the current time period of the target base station and the network utilization information of the next time period of the target base station comprises: performing carrier frequency increment and decrement necessity evaluation on the area covered by the target base station based on the network utilization information of the current time period of the target base station and the network utilization information of the next time period of the target base station; and determining the uplink and downlink carrier frequency increment and decrement amount of the target base station based on the evaluation result and the average limit capacity correlation model.

[0017] In some embodiments, the performing carrier frequency increment and decrement necessity evaluation on the area covered by the target base station based on the network utilization information of the current time period of the target base station and the network utilization information of the next time period of the target base station comprises: determining uplink and downlink traffic growth coefficients based on the uplink and downlink traffic in the network utilization information of the current time period and the uplink and downlink traffic in the network utilization information of the next time period; determining uplink and downlink utilization growth coefficients based on the uplink and downlink channel utilization in the network utilization information of the current time period and the uplink and downlink channel utilization in the network utilization information of the next time period; determining uplink and downlink carrier frequency resource bearing efficiency growth coefficients based on the uplink and downlink traffic growth coefficients and the uplink and downlink utilization growth coefficients; and performing uplink and downlink carrier frequency increment and decrement necessity evaluation on the area covered by the target base station based on the uplink and downlink traffic growth coefficients, the uplink and downlink utilization growth coefficients, and the uplink and downlink carrier frequency resource bearing efficiency growth coefficients.

[0018] The present disclosure further provides a carrier frequency resource allocation system, comprising: a feature determination module configured to acquire coverage population features of a target base station and network quality division results of the population covered by the target base station, and determine feature information of the target base station based on the coverage population features and the network quality division results; a prediction evaluation module configured to predict network utilization information of a next time period of the target base station based on the feature information of the target base station; a carrier frequency allocation module configured to determine uplink and downlink carrier frequency increment and decrement amount of the target base station based on the network utilization information of a current time period of the target base station and the network utilization information of the next time period of the target base station, determine total carrier frequency increment and decrement amount of the target base station based on the uplink and downlink carrier frequency increment and decrement amount and carrier frequency increment caused by service migration, and perform carrier frequency resource allocation based on the total carrier frequency increment and decrement amount.

[0019] The present disclosure further provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the carrier frequency resource allocation method according to any one of the above embodiments when executing the program.

[0020] The disclosure also provides a non-transitory computer-readable storage medium having stored thereon a computer program, which, when executed by a processor, implements the carrier frequency resource allocation method according to any one of the above.

[0021] The disclosure also provides a computer program product comprising a computer program, which, when executed by a processor, implements the carrier frequency resource allocation method according to any one of the above.

[0022] The carrier frequency resource allocation method, system, device, medium and program product provided by the disclosure can determine the feature information of the target base station based on the coverage population characteristics of the target base station and the network quality division result of the population covered by the target base station, and apply the feature information, so as to predict accurate network utilization information to support the allocation of carrier frequency resources, thereby effectively solving the network utilization prediction problem caused by long-period prediction difficulty and insufficient model generalization capability. By determining the uplink and downlink carrier frequency increment of the target base station based on the network utilization information of the target base station in the current period and the network utilization information of the target base station in the next period, and determining the total amount of carrier frequency increment of the target base station based on the uplink and downlink carrier frequency increment and the carrier frequency increment caused by service migration, the total amount of carrier frequency increment can be applied to carrier frequency resource allocation, an intelligent carrier frequency allocation method for service development and bearing efficiency change is realized, the influence of uplink and downlink service change and service migration caused by 4G frequency reduction is quantitatively evaluated, and the accuracy of carrier frequency resource allocation can be effectively improved, and the problems of investment waste or insufficient user protection are avoided. BRIEF DESCRIPTION OF DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the disclosure or the related art, the following will briefly introduce the drawings needed to be used in the embodiments or related art description. Obviously, the drawings in the following description are some embodiments of the disclosure, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0024] Fig. 1 is a flowchart of the carrier frequency resource allocation method provided by the disclosure.

[0025] Fig. 2 is a schematic diagram of single station coverage area calculation provided by the disclosure.

[0026] Fig. 3 is a flowchart of network utilization prediction model training provided by the disclosure.

[0027] Fig. 4 is a flowchart of the carrier frequency resource allocation method provided by the disclosure.

[0028] Fig. 5 is a flowchart of the network quality feature division algorithm based on single station coverage population decomposition provided by the disclosure.

[0029] Fig. 6 is a flowchart of network utilization prediction and accuracy evaluation according to the present disclosure.

[0030] Fig. 7 is a flowchart of intelligent allocation of carrier frequencies based on 5G uplink and downlink service development and 4G reduced frequency service migration according to the present disclosure.

[0031] Fig. 8 is a structural diagram of a carrier frequency resource allocation system according to the present disclosure.

[0032] Fig. 9 is a structural diagram of an electronic device according to the present disclosure. DETAILED DESCRIPTION

[0033] For the purpose, technical solutions and advantages of the present disclosure to be clearer, the technical solutions in the present disclosure will be described clearly and completely below in conjunction with the drawings in the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, rather than all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present disclosure.

[0034] At present, most of the carrier frequency resource allocation technologies are based on the convex optimization function model of waveform coding or communication system, which converts the problem into an optimal resource allocation problem with the maximum user rate as the target and the user transmission power and service quality as the constraint conditions. By establishing a deep learning joint optimal strategy to solve the optimization problem, a subcarrier allocation scheme is obtained. However, the application range of the theoretical algorithm result is limited, and it is difficult to obtain the objective measurement of the overall network carrying and running efficiency with the maximum single user rate as the target, and the ability to improve the accuracy of carrier frequency resource allocation is limited. In addition, the resource allocation theory based on deep learning usually outputs a closed-form solution space, and its strategy targets the user rate, which is difficult to obtain a globally optimal quantitative solution from the perspective of resource allocation, and lacks consideration of the real network service carrying efficiency and macro-level business demand changes.

[0035] The intelligent utilization rate prediction method is to transform the 5G network carrier frequency resource allocation problem into a time series prediction model for modeling and solving. Traditional time series prediction algorithms such as ARIMA are only linear models that consider the periodicity and trend of the sequence. The prediction is performed by first differentiating to obtain a stationary sequence. However, the network utilization rate does not have a significant periodicity. Compared with other time series prediction problems, the time series characteristics of a single cell utilization rate are more complex, and even difficult to describe with traditional time series models that calculate periodicity, trend or linearity. In addition, traditional time series prediction algorithms are usually based on single-step prediction, and the predicted value is added to the historical data as the true value to predict the next time step, which has certain error accumulation and is suitable for short-term time series prediction requirements. However, for long-period prediction, it will lead to low accuracy. At the same time, due to the lack of macro user and network data guidance, it is difficult to generalize the single-cell model to the multi-cell prediction scene, and it is difficult to meet the requirements of accuracy and efficiency for millions of data.

[0036] To this end, the embodiment of the present disclosure provides a carrier frequency resource allocation method, which can support carrier frequency resource allocation and guarantee user perception by proposing a network quality feature division algorithm based on single-station coverage population decomposition, thereby effectively solving the network utilization rate prediction problem caused by long-period prediction difficulty and insufficient model generalization ability. At the same time, by considering the uplink and downlink service changes and the service migration influence caused by 4G frequency reduction, the carrier frequency intelligent allocation for service development and bearing efficiency change is realized, which is scientific, efficient and highly implementable.

[0037] FIG. 1 is a flowchart of the carrier frequency resource allocation method provided by the present disclosure, as shown in FIG. 1, the method comprises the following steps 110-140.

[0038] Step 110, obtain the coverage population characteristics of the target base station and the network quality division result of the population covered by the target base station, and determine the feature information of the target base station based on the coverage population characteristics and the network quality division result.

[0039] It should be noted that the target base station refers to a specific base station that needs to be allocated carrier frequency resources, which can be any base station in a certain administrative region, or multiple base stations in a certain administrative region. By applying the carrier frequency resource allocation method provided by the embodiment of the present disclosure, the carrier frequency adjustment and allocation scheme at the base station level and each administrative division level can be output.

[0040] Specifically, the coverage population characteristics of the target base station refer to the characteristics of the residents or users within the coverage range of the base station, including but not limited to population size, distribution density, etc. The network quality classification result of the population covered by the target base station refers to the evaluation and classification of the network quality experienced by the users within the coverage area of the target base station, in order to understand the differences in network experience of different areas or user groups. The feature information of the target base station is a comprehensive analysis based on the coverage population characteristics and the network quality classification result, which provides a detailed description and attributes of the target base station. These feature information helps to predict network utilization, so as to make more accurate allocation of carrier frequency resources.

[0041] It can be understood that in order to obtain the coverage population characteristics of the target base station, geographic information system and population census data can be used to determine the geographic area and population distribution within the coverage range of the target base station. Macro population data can also be introduced from the corresponding statistical website and decomposed to a single base station. Macro population data and distribution are used to calculate population density coefficient, so as to derive the population size covered by a single station, in order to obtain the coverage population characteristics of the target base station.

[0042] After obtaining the coverage population characteristics of the target base station, sampling can be performed within the coverage area of the target base station, further decomposing the population size covered by a single station, and evaluating the network quality of each sampling point within the same period according to the reference signal strength of the broadcast channel, the reference signal strength of the service channel, the uplink and downlink rate, etc. According to the evaluation result, the coverage area of the target base station can be divided into different network quality levels, so as to obtain the population size within the coverage area in different network quality level ranges.

[0043] Combined with the coverage population characteristics and the network quality classification result, a feature information library of the target base station can be constructed, which can include population statistical characteristics, network quality indicators, user behavior patterns and other dimensions. These feature information will provide important reference basis for subsequent network utilization prediction and carrier frequency resource allocation.

[0044] Step 120, based on the feature information of the target base station, predicting the network utilization information of the target base station in the next period.

[0045] It should be noted that the network utilization information of the target base station in the next period refers to the prediction or estimation of the network usage of the target base station in the future time period. Here, in the 5G network, network utilization information is crucial for the allocation of carrier frequency resources, because it can help operators predict future traffic demand, so as to allocate resources in advance and ensure the stability and efficiency of the network.

[0046] Specifically, after obtaining the characteristic information of the target base station, various prediction algorithms or models can be used to predict the network utilization information of the target base station in the next period. For example, a suitable prediction model can be selected, the model can be trained using historical data, and the characteristic information of the target base station can be input into the trained model, so as to calculate the network utilization prediction information in the next period.

[0047] In addition, the prediction result of the model can be evaluated for accuracy to optimize the prediction effect. For the base station that fails to pass the evaluation, the features in the base station feature information can be sorted based on feature importance, the interference features can be removed, and the training can be repeated until the accuracy evaluation passes, so as to provide accurate input for subsequent carrier frequency resource allocation.

[0048] Step 130, based on the network utilization information of the target base station in the current period and the network utilization information of the target base station in the next period, determining the uplink and downlink carrier frequency increment and decrement of the target base station.

[0049] Specifically, based on the network utilization information of the target base station in the current period and the predicted network utilization information in the next period, comparison and analysis can be performed to determine the uplink and downlink carrier frequency increment and decrement of the target base station. Here, the uplink and downlink carrier frequency increment and decrement of the target base station refers to the specific value calculated based on the network utilization information for adjusting the number of carrier frequency resources of the target base station, which can include uplink and downlink carrier frequency increment and decrement. The uplink carrier frequency refers to the frequency band used by the user equipment to send data to the base station, and the downlink carrier frequency refers to the frequency band used by the base station to send data to the user equipment. In the embodiments of the present disclosure, the network utilization prediction information in the current period and the network utilization information in the next period are both predicted based on the characteristic information of the target base station. The prediction method of the network utilization prediction information in the current period can refer to the prediction method of the network utilization information in the next period, which will not be described here.

[0050] It can be understood that the network utilization information predicted in the above step 120 can include uplink and downlink traffic channel utilization, downlink control channel utilization, uplink and downlink traffic, and predicted coverage population, etc. According to the network utilization information of the target base station in the current period and the predicted network utilization information in the next period, the traffic, network utilization, and bearing efficiency joint traffic growth coefficient of the uplink and downlink can be calculated. Through the combination of these coefficients, the necessity of uplink and downlink carrier frequency increment and decrement can be evaluated, and combined with the single station average limit bearing capacity mutual correlation model, the uplink and downlink carrier frequency increment and decrement can be calculated.

[0051] Step 140, based on the uplink and downlink carrier frequency increment and decrement and the carrier frequency increment caused by service migration, determining the total amount of carrier frequency increment and decrement of the target base station, and applying the total amount of carrier frequency increment and decrement for carrier frequency resource allocation.

[0052] It should be noted that the carrier frequency increment caused by service migration refers to the fact that, since the 4G network is undergoing a frequency reduction operation (i.e., reducing the carrier frequency resources of the 4G network), the services originally running on the 4G network need to be migrated to the 5G network. In order to meet the demand of these newly added 5G services, the amount of carrier frequency resources needs to be increased in the 5G network. Here, the carrier frequency increment caused by service migration is in terms of each target base station.

[0053] Specifically, after calculating the uplink and downlink carrier frequency increment and decrement of the target base station, the total amount of carrier frequency increment and decrement of the target base station can be calculated by further combining the equivalent carrier folding increment of 4 / 5G, so that the total amount of carrier frequency increment and decrement can be used for carrier frequency resource allocation to output the base station level and each administrative division level carrier frequency adjustment and allocation scheme. Here, the total amount of carrier frequency increment and decrement of the target base station refers to the total amount of carrier frequency resources that the target base station finally needs to adjust after considering the carrier frequency increment and decrement required by the target base station based on network utilization changes and the carrier frequency increment caused by 4G frequency reduction service migration. The total amount of carrier frequency increment and decrement reflects the demand of 5G uplink and downlink service development changes on carrier frequency resources, and also considers the impact of network upgrade and service migration on resource allocation.

[0054] Specifically, the carrier frequency increment caused by 4G frequency reduction service migration can be calculated first, for example, the carrier frequency increment caused by 4G frequency reduction service migration can be calculated in an equivalent folding manner by using the comprehensive data of the carrier frequency scale, carrier bandwidth and wireless utilization rate of each frequency band of 4G, so as to obtain the carrier frequency increment caused by service migration. Subsequently, in order to guarantee the user perception of uplink and downlink and the demand of 4G frequency reduction service migration to 5G, ensure accurate resource allocation and avoid resource waste, the carrier frequency increment can be combined with the uplink and downlink carrier frequency increment and decrement required by the target base station based on network utilization changes, so as to obtain the total amount of carrier frequency increment and decrement of the target base station. Finally, according to the total amount of carrier frequency increment and decrement, the carrier frequency resources of the target base station can be allocated and adjusted specifically, which can include increasing or decreasing the number of carrier frequencies and other operations, so as to ensure that the target base station can meet the current and future network demand, while maintaining the reasonable utilization and efficient operation of network resources.

[0055] The method provided by the embodiments of the present disclosure can determine the feature information of the target base station based on the coverage population characteristics of the target base station and the network quality division result of the population covered by the target base station, and apply the feature information to predict accurate network utilization information to support allocation of carrier frequency resources, thereby effectively solving the network utilization prediction problem caused by long-period prediction difficulty and insufficient model generalization capability. By determining the uplink and downlink carrier frequency increment and decrement of the target base station based on the network utilization information of the target base station in the current period and the network utilization information of the target base station in the next period, and determining the total amount of carrier frequency increment and decrement of the target base station based on the uplink and downlink carrier frequency increment and decrement and the carrier frequency increment caused by service migration, the total amount of carrier frequency increment and decrement can be applied to carrier frequency resource allocation, an intelligent carrier frequency allocation method for service development and bearing efficiency change is realized, the influence of uplink and downlink service change and service migration caused by 4G frequency reduction is quantitatively evaluated, and the accuracy of carrier frequency resource allocation can be effectively improved to avoid investment waste or insufficient user protection.

[0056] Based on the above embodiments, in step 110, the coverage population characteristics of the target base station and the network quality division result of the population covered by the target base station are obtained, including steps 111-113 (not shown in the drawings).

[0057] In step 111, the number of regional population and the area of the region where the target base station is located are obtained.

[0058] Specifically, the number of regional population of the region where the target base station is located refers to the total number of population living or moving in the administrative region where the target base station is located. This data can be obtained through population census, official data of the statistical bureau or related population research, etc. The area of the region where the target base station is located refers to the size of the area of the administrative region where the target base station is located. For example, the population size data of each province can be introduced from the statistical bureau website. Taking A province as an example, the land area is 124,000 square kilometers, and the population size is 41,540,000. The detailed information is shown in Table 1 as follows:

[0059] Table 1 Population size table

[0060] The population size of the province and city can be divided into different administrative regions according to rural areas, county towns, towns and cities. It can be independently divided according to geographical range, or it can be divided according to the official standard of city planning, which is not limited in the embodiments of the present disclosure. Since the population distribution is uneven between regions and the station density is uneven, a population density coefficient d can be introduced to represent different regions:

[0061] Q(x)=De -δr

[0062] Wherein, where p(x) represents the population distribution probability in different regions x, Q(x) represents the population distribution model in different regions, r is the Euclidean distance between the region center and the city center, δ is the attenuation coefficient, and D is a constant term representing the benchmark value or initial value of the population information entropy. The administrative village and the natural village are high-density non-continuous population distribution, and are far away from the city center. Therefore, d 农村 <d 乡镇县城 <d 城市 .

[0063] In step 112, the coverage population size of the target base station is determined based on the single-station coverage area of the target base station, the population number of the region, and the area of the region. The coverage population feature includes the single-station coverage area and the coverage population size.

[0064] Specifically, the single-station coverage area of the target base station refers to the geographical area covered by the target base station. The single-station coverage area is represented by BS_area, which can be estimated by the Thiessen polygon algorithm according to the station size in the region and the station site structure relationship.

[0065] FIG. 2 is a schematic diagram of single-station coverage area calculation provided by the present disclosure. As shown in FIG. 2, taking a single target base station as an example, the connecting lines of N neighboring stations around the base station are traversed, and N median lines are sequentially drawn as the N-polygon coverage boundary of the station. The number of station sites N around the neighboring stations determines the number of coverage boundaries. The connecting lines of the N-polygon coverage boundaries form a closed polygon to calculate the area of the coverage surface, thereby obtaining the single-station coverage area of the target base station.

[0066] Subsequently, according to the single-station coverage area of the target base station, the population number of the region, and the area of the region, the coverage population size of the target base station can be determined. Here, the coverage population size of the target base station can be the average coverage population number of the target base station derived. Specifically, it can be calculated by the following formula:

[0067] wherein P_cover represents the coverage population size of the target base station, S_area represents the area of the region, P_area represents the population number of the region, and BS_area represents the Thiessen polygon estimated area. The area of the region S_area, the population number of the region P_area, and the single-station coverage area BS_area are macroscopically known data.

[0068] Taking the three typical base stations shown in Table 2 as an example, the base station 484585 is located in a rural area, corresponding to a non-continuous network hotspot coverage, the single-station coverage area is large (1.21 square kilometers) but the population density is low (36 people), and the population density coefficient is 0.69 calculated according to the above formula. The average coverage population number of the three base stations can be calculated in the same way The results are as follows: The results are consistent with the expectations.

[0069] Table 2 Population characteristics covered by base stations

[0070] Step 113, selecting a plurality of sampling points within the single-station coverage area of the target base station, dividing the population size covered by the network quality based on the channel signal strength and uplink and downlink rates of each sampling point, to obtain the network quality division result.

[0071] Specifically, after calculating the population size covered by the target base station, it can be further decomposed, that is, different sampling points are selected within the single-station coverage area of the target base station, and the network quality of each sampling point in the same period is evaluated according to the channel signal strength and uplink and downlink rates of each sampling point. Here, a plurality of sampling points are selected within the single-station coverage area for network quality evaluation and data collection. The selection of sampling points should be representative and able to reflect the network conditions in the coverage area. The channel signal strength of the sampling point refers to the strength of the signal received at the sampling point, which can include the reference signal strength of the broadcast channel and the reference signal strength of the service channel. The uplink and downlink rates of the sampling point refer to the rates of uploading and downloading data between the user equipment and the base station at the sampling point, wherein the uplink rate refers to the data rate from the user equipment to the base station, and the downlink rate refers to the data rate from the base station to the user equipment.

[0072] For example, the network quality level can be divided into good, medium and poor. For the population size covered by a sampling point in the target base station , the coverage position of the sampling point can be evaluated according to the reference signal strength P SS-RSRP of the broadcast channel at the sampling point CSI-RSRP , the reference signal strength P UP_RATE of the service channel at the sampling point DOWN_RATE , the uplink rate V good and the downlink rate V gereral whether the coverage position is in the good point set, the medium point set or the poor point set. The specific decision method is as follows:

[0073] First, determine whether the coverage position of the sampling point is in the good point set A good , and the decision condition is:

[0074] If the condition is not met, continue to determine whether the coverage position of the sampling point is in the medium point set A gereral , and the decision condition is:

[0075] If the above two conditions are not met, the coverage position of the sampling point is automatically attributed to the poor point set A bad , and the explicit decision condition is:

[0076] In the above formula, s i ,s j are the upper and lower limits of the decision threshold, i.e., the upper and lower limits of the network quality division threshold, which is a first-order threshold function and can be calculated according to the network modulation order distribution characteristics. Here, the network modulation order distribution characteristics can be the MCS (Modulation and Coding Scheme) order distribution characteristics, which refers to the use or distribution of different modulation and coding scheme orders in network communication.

[0077] Based on the above embodiment, step 113 specifically includes: calculating the network quality division threshold of each sampling point based on the network modulation order distribution characteristics of each sampling point in the current period; and dividing the coverage population size based on the network quality division threshold of each sampling point, and the channel signal strength and uplink and downlink rate of each sampling point, to obtain the network quality division result.

[0078] Specifically, according to the MCS order distribution characteristics, the proportion of QPSK, 16QAM, 64QAM and 256QAM modulation modes in the same period is counted, and s i ,s j can be calculated, where s i is the upper limit and s j is the lower limit. Here, QPSK (Quadrature Phase Shift Keying) is a digital modulation method that can carry 2 bits of information per symbol; 16QAM (16-Quadrature Amplitude Modulation) can carry 4 bits of information per symbol; 64QAM can carry 6 bits of information per symbol; and 256QAM can carry 8 bits of information per symbol. These modulation methods are used in wireless communication to increase data transmission rate, and as the modulation order increases (from QPSK to 256QAM), the amount of information that can be carried per symbol also increases. It should be understood that the higher the MCS order, the higher the optimization quality and bearing efficiency of the network RB (Resource Block) resource, and the higher the adaptive threshold s i for dividing good and poor points, and the lower the s j .

[0079] The upper limit s i can be determined by calculating the maximum value of the residence time of the highest order and the user ratio:

[0080] In the above formula, k refers to the number of users in the sampling point coverage area, and tm is the time length of user m opening the highest order N max , T is the total time length of user staying, R m is the proportion of users opening the highest order, and R is the total number of users staying. The maximum value of the proportion calculated respectively is multiplied by N max , N max Generally, it can be set to 8. Here, the total time length of user staying refers to the total time of all users staying in the network or on a specific service.

[0081] The lower limit s j can be determined by calculating the minimum value of the staying time length of the lowest order and the proportion of users:

[0082] t n is the time length of user n opening the lowest order N min , T is the total time length of user staying, R n is the proportion of users opening the lowest order, and R is the total number of users staying. The minimum value of the proportion calculated respectively is multiplied by N min , N min Generally, it can be set to 2.

[0083] For a sampling point, after the upper limit s i and the lower limit s j of the network quality division threshold corresponding to the sampling point are calculated, whether the sampling point is a good point, a medium point or a poor point can be determined according to the determination condition in the above embodiment. The detailed determination is shown in Table 3 as follows:

[0084] Table 3 Network quality division result information table

[0085] It should be noted that the reference signal strength of the broadcast channel, the reference signal strength of the service channel, the uplink rate and the downlink rate in Table 3 above are all values after normalization processing after obtaining the original data. By comparing these values with the upper limit and the lower limit, the network quality division result can be obtained.

[0086] In the embodiment of the present disclosure, through the good, medium and poor point feature division of the population scale, firstly, the population density of different areas can be finely distinguished. In areas with high density, the network load is high, and the capacity potential is large, which is a necessary input feature of the subsequent network utilization prediction model. Secondly, the population of different base stations at different positions can be finely divided in terms of network quality, which can further depict the differences in network utilization caused by different spectrum efficiencies of good points, medium points and poor points. The closer the sampling point of the population covered by a single station is to the network device, the better P CSI-RSRP / P SS-RSRP and V UP_RATE / VDOWN_RATE The higher, the more sampling points exceeding s i The lower, the fewer sampling points below s j The higher, the higher the spectrum efficiency, and the lower the network utilization under the condition of equal traffic. Therefore, it is also a necessary input feature of the network utilization prediction model.

[0087] Based on any of the above embodiments, in step 110, determining the feature information of the target base station based on the coverage population characteristics and the network quality division result includes: removing outliers from the network quality division result, and data grouping the network quality division result after removing outliers, the coverage population characteristics, and the network traffic data of the target base station to obtain the feature information of the target base station.

[0088] Specifically, for each target base station, after obtaining the network quality division result of the population it covers, the result can be subjected to outlier removal. For example, the base station ID repeated values can be removed, and the single-station coverage population sampling point number repeated values can be removed. If the reference signal strength P SS-RSRP , the reference signal strength P CSI-RSRP , the uplink rate V UP_RATE , the downlink rate V DOWN_RATE of a certain sampling point are all 0 or null values or \t, they are removed, and if one of them is not 0, the entire row is retained.

[0089] After the outlier removal process, the network quality division result, the coverage population characteristics, and the network traffic data of the target base station can be subjected to data grouping, thereby forming a feature table of the target base station, i.e., obtaining the feature information of the target base station. As shown in Table 4 below, the formed feature information can include the following 20 typical features:

[0090] Table 4 Feature Information

[0091] The features with serial numbers 1-9 are the coverage population features and network quality classification results obtained by the above embodiments, and specifically include base station name, base station ID, area, coverage population size, single station coverage area, population density coefficient, population sampling size in good point, medium point and poor point. The features with serial numbers 10-20 are derived from the introduced cell engineering parameters (referring to cell engineering parameters, which are a set of parameters used to describe the characteristics of a wireless communication cell), service development, 4G MR (Measurement Report), 5G MR, market user and other network data. These network data mainly include: daily average 5G traffic, daily average effective RRC (Radio Resource Control) connection average number, 5G single station average uplink and downlink utilization rate, ARPU (Average Revenue Per User) greater than 120 yuan user number, 5G terminal number, 5G SA (Standalone) call sheet user traffic in the 4G network, SS-RSRP (Synchronization Signal Reference Signal Received Power) ≤-110 dBm sampling point number > 5% sector number, uplink and downlink average rate, average RSRP and other system data.

[0092] It can be understood that, considering that different databases or data storage areas store data in different rules or ways, when data is grouped, data stored in a certain data storage area can be taken as central data, and data from other sources can be integrated and grouped according to the rules or ways of the data storage area. For example, O-domain data can be taken as central data for grouping to form a base station feature table, that is, ecgi+base station ID of O-domain data is taken as a joint primary key for splicing, and other data is grouped to O-domain data, and M-domain is grouped to O-domain after deduplication with base station ID as the primary key. Here, ecgi refers to the abbreviation of E-UTRAN Cell Global Identifier, which is an identifier for uniquely identifying an LTE cell. In addition, after data grouping is completed, the proportion of splicing failed cells can be counted, and if the proportion is >20%, the accuracy of the result analysis will be affected, and the data needs to be supplemented and improved before being calculated again.

[0093] After the feature information of the target base station is obtained through the above steps, it can be stored in a data unit for subsequent prediction of network utilization information and training and testing of a network utilization prediction model.

[0094] Based on any of the above embodiments, step 120 specifically comprises: inputting the feature information into the network utilization prediction model corresponding to the target base station to obtain the network utilization information of the target base station in the next time period output by the network utilization prediction model; and the network utilization prediction model is obtained by training based on sample feature information and sample network utilization information of the target base station and after model accuracy evaluation.

[0095] It should be noted that the network utilization prediction model is a tool based on an AI (Artificial Intelligence) algorithm and machine learning technology, which is used to predict the network utilization information of the target base station in a future time period. This model learns and identifies the relationship between network utilization and various influencing factors by using historical data (i.e., sample feature information and sample network utilization information), and predicts future network utilization based on these relationships.

[0096] Specifically, before predicting the network utilization information of the target base station in the next time period, the network utilization prediction model corresponding to the target base station can be trained in advance, and the training can be performed by the following steps: first, a large amount of historical data of the target base station is collected, which includes feature information of the base station and corresponding network utilization information; then, the initial model is trained using these data, and the prediction result is optimized by continuously adjusting the model parameters. During the training process, some evaluation indicators can be used to evaluate the accuracy of the model, such as mean square error, accuracy, etc. These evaluation indicators help to understand the performance of the model in predicting network utilization, and the model is optimized accordingly.

[0097] When the model training is completed and passes the accuracy evaluation, it can be used to predict the network utilization information of the target base station in the future time period. It should be understood that the network utilization prediction model is not fixed, and as the network environment and business requirements change, the model also needs to be updated and retrained regularly to maintain its prediction accuracy and effectiveness.

[0098] Based on any of the above embodiments, FIG. 3 is a flowchart of training of the network utilization prediction model provided by the present disclosure, as shown in FIG. 3, the objective function of the network utilization prediction model includes a training loss function based on a relative error square root and a regularization loss function, and the training steps of the network utilization prediction model include steps 310-340.

[0099] Step 310, obtaining a training data set and a test data set of the target base station, the training data set including sample feature information and sample network utilization information, and the test data set including test feature information and test network utilization information.

[0100] Step 320, input the sample feature information into the initial network utilization prediction model to obtain first predicted network utilization information output by the initial network utilization prediction model.

[0101] Step 330, based on the first predicted network utilization information and the sample network utilization information, parameter iteration is performed on the initial network utilization prediction model to obtain a pre-trained network utilization prediction model.

[0102] Step 340, the test feature information and the test network utilization information are applied to perform model accuracy evaluation on the pre-trained network utilization prediction model to obtain the network utilization prediction model.

[0103] It should be noted that when training a machine learning model, a function or a set of parameters is usually tried to be found to make the performance of the model on the given data optimal, and this performance is usually quantified by an objective function. The objective function defines the optimization goal of the model. For the network utilization prediction model, the objective function can include a training loss function and a regularization loss function. Among them, the training loss function is used to measure the degree of inconsistency between the predicted value of the model and the true value. By minimizing the training loss function, a set of model parameters can be found to make the performance of the model on the training data optimal. The regularization loss function is used to prevent model overfitting. By combining the training loss function and the regularization loss function, the objective function of the network utilization prediction model can be formed. By minimizing the objective function, a model with good performance on training data and strong generalization ability can be found.

[0104] The objective function is composed of two parts: the training loss function and the regularization loss function:

[0105] obj(θ)=L(θ)+Ω(θ)

[0106] Where L(θ) is the training loss function and Ω(θ) is the regularization loss function. The training loss function based on the square root of relative error is designed as follows:

[0107] Considering that the predicted target network utilization is a percentage data between 0% and 100%, it is not suitable for the absolute error method with high starting point and high magnitude for traffic prediction and user number prediction. Therefore, the square root of relative error method can be used to measure the training loss between the predicted value y i and the true value , which is standardized to 1% calculation, avoiding the problem of positive numbers offsetting negative numbers. At the same time, the square root design can avoid the accuracy decline caused by the extreme loss caused by the appearance of the maximum predicted network utilization (>95%).

[0108] The regularization loss function Ω(θ) is used to penalize large feature weights, implicitly reduce the number of free parameters during model training, and control overfitting. The regularization loss function is represented as follows:

[0109] Ω(θ) = γ∑|ω|

[0110] where γ is a hyperparameter used to control the size of the regularization term, and ω represents the weight parameters of the model.

[0111] Specifically, after the objective function is constructed, a suitable initial network utilization prediction model can be selected and trained using a pre-collected training dataset. Here, the initial network utilization prediction model refers to the model structure and initial parameters set before training begins, which can be based on a certain machine learning algorithm or deep learning architecture. For example, the initial network utilization prediction model can be a multilayer perceptron (MLP), a convolutional neural network (CNN), or a recurrent neural network (RNN), etc., which is not specifically limited by the embodiments of the present disclosure. The training dataset refers to the dataset used to train the machine learning model, which can include sample feature information (i.e., input features) and sample network utilization information (i.e., target output or label). The training dataset is used to adjust the parameters of the model so that the model can learn the mapping relationship from the input features to the output labels.

[0112] During model training, the difference between the network utilization information predicted by the model and the sample network utilization information (i.e., training loss) can be calculated, and the gradient can be calculated in combination with the regularization loss, so that the parameters of the model can be updated using an optimization algorithm (such as gradient descent) to reduce the target loss. Through multiple training iterations, until the performance of the model on the training dataset reaches a pre-set standard or the number of iterations reaches an upper limit, a pre-trained network utilization prediction model can be obtained.

[0113] Subsequently, the test feature information in the test dataset can be used as output to let the pre-trained network utilization prediction model predict the network utilization of the target base station, and the predicted network utilization information can be compared with the test network utilization information (i.e., the true label) in the test dataset to calculate the accuracy of the model on the test dataset, so that the network utilization prediction model after model accuracy evaluation can be obtained for predicting network utilization information in actual scenarios.

[0114] The method provided by the embodiments of the present disclosure fully utilizes the relative error square root loss function of the percentage data in the intelligent algorithm to fit the complex and nonlinear network utilization prediction model, constructs the network quality features based on single station coverage population decomposition, proposes an accuracy evaluation method for long-period network utilization prediction, and significantly improves the adaptability and generalization ability of the business scenario.

[0115] Based on any of the above embodiments, step 340 specifically comprises: inputting the test feature information into the pre-trained network utilization prediction model to obtain second predicted network utilization information output by the pre-trained network utilization prediction model; comparing the second predicted network utilization information with the test network utilization information, and determining a model accuracy evaluation result of the pre-trained network utilization prediction model based on the comparison result; in the case that the model accuracy evaluation result is inaccurate, rearranging features of the test feature information and removing interference features to obtain new test feature information; and applying the new test feature information to re-evaluate the model accuracy of the pre-trained network utilization prediction model until the model accuracy evaluation result is accurate, and obtaining the network utilization prediction model.

[0116] Specifically, for each target base station, the monthly snapshot data of the last N months can be used for training, and the prediction effect of the model can be tested on the monthly snapshot data of the N+1 to N+6 months. Here, the monthly snapshot data refers to detailed data records about base station feature information and network utilization information collected at a fixed time point in each month. For the same base station ID, the relative error of the six-month predicted values of N+1, N+2, N+3, N+4, N+5, N+6 and the true values is calculated respectively, and the scale and list of the base stations with a relative error less than 20% are counted and labeled as “accurate”. The proportion of the number of “accurate” labels to the total number of labels is counted station by station and month by month, which is the accurate label proportion M in the following. From the N+1 month to the N+6 month, if the accurate label proportion M continuously satisfies the following requirements, it is considered that the base station network utilization prediction is accurate and effective, and can be applied in the system. Otherwise, it is considered that the model is not accurate enough, and the importance feature sorting of the true value needs to be continued to remove the interference items.

[0117] M≥100-bm, m∈[1, 6]

[0118] In the above formula, m represents the month, and b represents that the longer the prediction time period is, the lower the accuracy standard is set.

[0119] As shown in Table 5, the accurate label proportions M of the base station C-HRHH for the last six months all satisfy the above requirements, so the long-period prediction evaluation result of this station is accurate. The accurate label proportions of the base station B-ZRHH for the last six months do not meet the requirements, so the prediction evaluation result is inaccurate.

[0120] Table 5 Model Accuracy Evaluation Information Table

[0121] For the base station whose prediction evaluation result is inaccurate, the feature permutation importance evaluation is performed station by station, the interference term is removed to enhance the model performance, and the model is retrained according to the new month data and new feature combination. Randomly reorder (shuffle) the features in Table 4, calculate the degree of decrease of the relative prediction error, and quantify the contribution size of the feature to the model. The input of the algorithm is the trained model M and the training set D, and the performance score of the model M on the data set D is s, then for each feature j in the data set D, the importance score i j :

[0122] For each iteration k in K times of repeated experiments, randomly rearrange the features j to construct a contaminated data set Data set The feature column is unchanged, and the upper and lower order in the column is shuffled. The operation is repeated K times, and only one feature is shuffled each time for comparison. The performance score s of the model M on the data set is calculated k,j . The importance score i j of the feature j is larger, the contribution of the feature to the utilization feature is larger, and the feature is more important.

[0123] Taking the evaluation result of base station B-ZRHH as an example, the importance ranking t of each feature is obtained by calculating the importance score of each feature, which provides a basis for the construction of the average limit capacity mutual correlation model in the subsequent embodiments. The detailed information is shown in Table 6:

[0124] Table 6 Feature importance ranking information table

[0125] According to the feature importance reordering, the features ranked in the last 10% or the importance score i j <10 can be screened, and the model weight value is adjusted downward or set to 0, and then the model accuracy is evaluated again. If the evaluation result is accurate, it proves that the new model after updating the feature has substantial improvement and can be applied and deployed. If the evaluation conclusion is still inaccurate, it is necessary to further expand the interference term range to the features ranked in the last 20% or the importance score i j <20, and the evaluation effect is iteratively tested.

[0126] After the network utilization prediction model is evaluated, it can be applied to the actual scene to predict the network utilization prediction information of the target base station in the next period. The model result output dataset includes site-level {predicted uplink traffic channel utilization prb_up_pre, predicted downlink traffic channel utilization prb_dl_pre, predicted downlink control channel utilization cce_pre, predicted uplink traffic D_up_pre, predicted downlink traffic D_down_pre, and predicted coverage population U_pre}, that is, the network utilization prediction information, which can be stored as the input of the subsequent intelligent allocation algorithm of the carrier frequency.

[0127] Based on any of the above embodiments, step 130 specifically includes:

[0128] Step 131, based on the network utilization information of the target base station in the current period and the network utilization information of the target base station in the next period, evaluating the necessity of increasing or decreasing the carrier frequency of the area covered by the target base station; and

[0129] Step 132, based on the evaluation result and the average limit carrying capacity mutual correlation model, determining the uplink and downlink carrier frequency increase and decrease amount of the target base station.

[0130] Specifically, the constructed feature information of the target base station is input into the network utilization prediction model, and the predicted uplink and downlink traffic channel utilization, the predicted downlink control channel utilization, the predicted uplink and downlink traffic, and the predicted coverage population can be obtained. According to the uplink and downlink traffic channel utilization, the downlink control channel utilization, the uplink and downlink traffic of the target base station in the current period, and the predicted results output by the model, the traffic, network utilization, and carrying efficiency joint business growth coefficient of the uplink and downlink can be calculated respectively. Through the combination of these coefficients, the necessity of increasing or decreasing the carrier frequency of the area covered by the target base station can be evaluated to guide the accurate allocation of resources. At the same time, the average limit carrying capacity mutual correlation model can be constructed, and the model is constructed for each base station. Applying the model can guide the calculation of the carrier frequency increase and decrease amount, so as to obtain the uplink and downlink carrier frequency increase and decrease amount of the target base station.

[0131] Based on any of the above embodiments, step 131 specifically includes: determining an uplink and downlink traffic growth coefficient based on the uplink and downlink traffic in the network utilization information of the current period and the uplink and downlink traffic in the network utilization information of the next period; determining an uplink and downlink utilization growth coefficient based on the uplink and downlink channel utilization in the network utilization information of the current period and the uplink and downlink channel utilization in the network utilization information of the next period; determining an uplink and downlink carrier frequency resource carrying efficiency growth coefficient based on the uplink and downlink traffic growth coefficient and the uplink and downlink utilization growth coefficient; and performing uplink and downlink carrier frequency increase and decrease necessity evaluation on the area covered by the target base station based on the uplink and downlink traffic growth coefficient, the uplink and downlink utilization growth coefficient, and the uplink and downlink carrier frequency resource carrying efficiency growth coefficient.

[0132] It should be noted that the service growth coefficient in the above embodiments includes an uplink service growth coefficient and a downlink service growth coefficient, and similarly, the carrier frequency increase and decrease necessity evaluation includes uplink carrier frequency increase and decrease necessity evaluation and downlink carrier frequency increase and decrease necessity evaluation. The final determined uplink and downlink carrier frequency increase and decrease amount of the target base station includes uplink carrier frequency increase amount, uplink carrier frequency decrease amount, downlink carrier frequency increase amount, and downlink carrier frequency decrease amount. The carrier frequency intelligent allocation method based on 5G uplink and downlink service development and 4G frequency reduction service migration provided by the embodiments of the present disclosure specifically includes the following steps:

[0133] ① Calculate the downlink service growth coefficient

[0134] Site-level downlink traffic growth coefficient:

[0135] Where D_down_pre represents the predicted downlink traffic, and D_down represents the downlink traffic in the current period.

[0136] Site-level downlink utilization growth coefficient:

[0137] b = max (prb_dl_pre, cce_pre) - max (prb_dl, cce)

[0138] Where prb_dl_pre represents the predicted downlink service channel utilization, cce_pre represents the predicted downlink control channel utilization, prb_dl represents the downlink service channel utilization in the current period, and cce represents the downlink control channel utilization in the current period. The site resource carrying efficiency growth coefficient c represents the growth of the network downlink frequency resource carrying efficiency. The greater c is, the more carrier frequency configuration is required:

[0139] ② Downlink carrier frequency increase and decrease necessity evaluation

[0140] The necessity of increasing or decreasing the carrier frequency is evaluated by a combination of coefficients, and the method and conclusion are shown in Table 7 to guide the accurate allocation of resources.

[0141] Table 7 Evaluation table of carrier frequency increase / decrease necessity

[0142] ③ Calculation of downlink carrier frequency increase / decrease amount

[0143] Calculation of downlink carrier frequency increase amount: if the conditions a≥1 and b≥0 are met, the carrier frequency resource increase demand d1 follows the principle of maximum guarantee, and is calculated as follows:

[0144] Calculation of downlink carrier frequency decrease amount: if the conditions 0

[0145] ④ Construction of average limit carrying capacity mutual correlation model

[0146] In the above formula, the average limit carrying capacity coefficient term is introduced, where carrier_a corresponds to the single carrier flow limit, carrier_b corresponds to the single carrier utilization limit, carrier_c corresponds to the single carrier carrying efficiency limit (which is usually a fixed value), and U_pre is the predicted coverage population number by multiplexing the network utilization prediction model in the above embodiment. Through massive data mining, a mutual correlation model is constructed:

[0147] carrier_a=xln(carrier_b)+yln(U_pre)+z

[0148] In the mutual correlation model, the coefficient z is a constant.

[0149] ⑤ Model adjustment coefficient scoring

[0150] In the mutual correlation model, the coefficient x is related to 10 indicators of business development and market value. These 10 indicators can be scored first and then normalized to the interval [1, 100] by weighting, so as to obtain the value of the coefficient x. Specifically, the index item can be matched with the feature name in Table 6, the q item matched with the feature name in Table 6 is consistent, the q index is classified as the first batch of weighted model, and the feature importance sorting result t is reused as input to construct a linear equation. The larger t is, the higher w1 is, and the modeling form is as follows:

[0151] w1=kt,k>0

[0152] In the above formula, k is a constant coefficient. The indicators beyond the feature range of Table 6 are taken as the second batch of indicators, and the number of indicators is 10-q. Each indicator calculates the weight by equal division method:

[0153] The detailed scoring rules and weighting models are shown in Table 8.

[0154] Table 8 Coefficient x evaluation rules

[0155] In the cross-correlation model, the coefficient y is related to seven indicators of network quality and bearing efficiency. The seven indicators can be scored and weighted according to the rules shown in Table 9, normalized to the interval [1, 100], to obtain the value of the coefficient y. Specifically, the index item can be matched with the feature name in Table 6, and the r items matched with the feature name in Table 6 are classified as the first batch of weighting models, and the feature importance ranking result t is reused as input to construct a linear equation. The larger t is, the higher w1 is, and the modeling form is as follows:

[0156] w1 = kt, k > 0

[0157] In the above formula, k is a constant term coefficient. The indicators beyond the feature range in Table 6 are taken as the second batch of indicators, and the number of indicators is 7-r. Each indicator calculates the weight by equal division method:

[0158] The detailed scoring rules and weighting models are shown in Table 9.

[0159] Table 9 Coefficient y evaluation rules

[0160] (6) Calculate the uplink traffic growth coefficient of the station

[0161] The uplink traffic growth coefficient of the station is:

[0162] Where D_up_pre represents the predicted uplink traffic, and D_up represents the uplink traffic in the current period.

[0163] The uplink utilization growth coefficient of the station is slightly different from the downlink:

[0164] b' = prb_up_pre - prb_up

[0165] Where prb_up_pre represents the predicted uplink traffic channel utilization rate, and prb_up represents the uplink traffic channel utilization rate in the current period. The station resource bearing efficiency growth coefficient c' represents the network uplink frequency resource bearing efficiency:

[0166] (7) Evaluation of the necessity of uplink carrier frequency increase or decrease

[0167] The necessity of uplink carrier frequency increase or decrease is evaluated according to Table 7.

[0168] ⑧Calculate the uplink carrier frequency increment

[0169] Calculate the uplink carrier frequency increment size: if the condition a≥1, b≥0 is met, the carrier frequency resource increment demand d3 is calculated as follows:

[0170] Calculate the downlink carrier frequency decrement size: if the condition 0

[0171] The average limit carrying capacity correlation model is constructed and the coefficient scoring is consistent with the above table 8 and table 9 method.

[0172] ⑨Calculate the uplink carrier frequency increment

[0173] The 4G frequency band carrier frequency size, carrier bandwidth and wireless utilization rate are used to calculate the 4G frequency reduction demand increment d5 for 5G expansion in an equivalent conversion manner:

[0174] Wherein, num_c_4G is the 4G frequency reduction carrier size, bandwith_c_4G is the available carrier bandwidth of 4G frequency band, prb_c_4G is the 4G carrier wireless utilization rate, bandwith_c_5G is the available carrier bandwidth of 5G, max_prb_c_5G is the theoretical maximum busy time single carrier space utilization rate of 5G.

[0175] ⑩Calculate the total size of carrier frequency increment or decrement

[0176] In order to guarantee the user perception of uplink and downlink and the demand of 4G frequency reduction to 5G service migration, ensure accurate resource allocation and avoid resource waste, the total demand of carrier frequency increment is considered comprehensively:

[0177] d + =max(d1,d3)+d5

[0178] The total demand of carrier frequency reduction is:

[0179] d - =min(d2,d4)+d5

[0180] After calculating the total amount of carrier frequency increment and decrement, the carrier frequency adjustment and allocation scheme can be output. Since the uplink and downlink service distribution proportion of the business is balanced, the uplink and downlink are increased and decreased at the same time, and the increment and decrement amount conclusion is based on the classification of a, b coefficient, which is independent of each other and does not affect each other. According to the calculation results of the above steps, the carrier frequency allocation scheme of base station level and each administrative division can be obtained.

[0181] Exemplarily, the base station level carrier frequency allocation scheme obtained by the above steps can be shown in table 10 as follows:

[0182] Table 10 Base station level carrier frequency allocation output scheme information table

[0183] The carrier frequency increase and decrease scale requirements of the base station are summarized, and the carrier frequency adjustment and allocation scheme of each administrative division level is obtained, and the detailed information is shown in Table 11 as follows:

[0184] Table 11 Carrier frequency allocation output scheme information table of each administrative division level

[0185] In the embodiments of the present disclosure, by starting from the overall judgment of the real network service bearing efficiency and the macroscopic business demand change, the influence of the 5G uplink and downlink business change and the 4G frequency reduction on the business migration is quantitatively evaluated, the scoring rules of key parameters are refined, the accuracy of the precise allocation of carrier frequency resources is effectively improved, and the problems of investment waste or insufficient user protection are avoided.

[0186] Based on any of the above embodiments, FIG. 4 is a flowchart of a carrier frequency resource allocation method provided by the present disclosure, as shown in FIG. 4, which can be applied to a carrier frequency intelligent allocation system to output the carrier frequency adjustment and allocation scheme of the base station level and each administrative division level. The system mainly consists of three modules: one is a network quality feature division algorithm module based on single station coverage population decomposition, the second is a network utilization rate prediction and accuracy evaluation module based on AI algorithm, and the third is a carrier frequency intelligent allocation module based on 5G uplink and downlink business development and 4G frequency reduction business migration. The three modules are connected with each other and logically connected in series. The output of the previous module is the input of the next module. The method specifically includes:

[0187] First, the macro population data is introduced and decomposed to each base station to construct the single station coverage population characteristics, and a multi-dimensional network quality good point, middle point and poor point division method and an adaptive threshold decision method based on network modulation order are proposed to form a feature table to input the network utilization rate prediction model. Then, a relative error square root loss function for network utilization rate percentage data is designed, and a continuous monthly network utilization rate prediction accuracy evaluation method is proposed. For the sites that do not pass the evaluation, interference features are screened out based on feature importance sorting, and the model is trained in a loop until the accuracy evaluation passes, providing business input for the next carrier frequency allocation module. In the carrier frequency allocation module, a traffic, network utilization rate and bearing efficiency combined business increase and decrease coefficient judgment method is proposed, the carrier frequency increase and decrease demand is calculated through a single station average limit bearing capacity mutual correlation model, and the incremental carrier frequency scale is calculated combined with the 4G / 5G equivalent carrier conversion, and the total amount of carrier frequency increase and decrease is calculated to form the carrier frequency adjustment and allocation scheme of the base station level and each administrative division level. The specific implementation steps of the three modules will be introduced respectively as follows:

[0188] FIG. 5 is a flowchart of a network quality feature division algorithm based on single-station coverage population decomposition provided by the present disclosure. As shown in FIG. 5, to accurately quantify the network quality of single-station population density and coverage population sampling points, the present disclosure provides a network quality feature division method based on single-station coverage population decomposition. First, the population size data of each province can be introduced from the statistical bureau website, and the population density coefficient can be calculated according to the population distribution model. At the same time, the single-station coverage area in the region can be calculated by the Thiessen polygon algorithm, so as to further deduce the average number of single-station coverage populations. Subsequently, the network quality good points, medium points and poor points can be divided based on multi-dimensional joint and adaptive threshold. That is, the sampling points are selected in the coverage area of the target base station, and the upper and lower limits of the judgment threshold of each sampling point are calculated according to the network modulation order distribution characteristics. Whether the sampling point is a good point, a medium point or a poor point is judged according to the reference signal strength of the broadcast channel, the reference signal strength of the service channel, and the uplink and downlink rates. Finally, the network quality division result can be processed to remove outliers, and the processed network quality division result, the coverage population characteristics of the base station, and the network traffic data are grouped to form a base station feature table.

[0189] FIG. 6 is a flowchart of network utilization rate prediction and accuracy evaluation provided by the present disclosure. As shown in FIG. 6, the training loss function of the root mean square of relative error and the regularization loss function are constructed, and the target function of model training is obtained by combining the two functions. At the same time, the present disclosure provides a continuous monthly network utilization rate prediction accuracy evaluation method, and proposes an interference feature screening method based on feature importance ranking for sites with low accuracy. Specifically, the model can be trained using N consecutive monthly snapshot data, and the prediction effect of the model can be tested on the data of the N+1 to N+6 months. When the prediction accuracy of the model is accurate, the model can be deployed in the system and applied. When the prediction accuracy of the model is inaccurate, the feature importance of each feature in the base station feature table needs to be evaluated, and the interference items need to be removed to enhance the performance of the model. Subsequently, the model can be retrained according to the new month data and the new feature combination. If the evaluation result is accurate, it proves that the new model after feature update has substantial improvement, and can be applied and deployed. If the evaluation result is still inaccurate, the range of interference items needs to be further expanded, and the evaluation effect needs to be iteratively tested.

[0190] It can be understood that when the above model is used to predict network utilization information, the results obtained by the model output include {predicted uplink service channel utilization rate, predicted downlink service channel utilization rate, predicted downlink control channel utilization rate, predicted uplink traffic, predicted downlink traffic, and predicted coverage population number} at the site level. These information can be stored as input of the next step of the intelligent allocation algorithm of carrier frequency.

[0191] FIG. 7 is a flowchart of a method for intelligent allocation of carrier frequencies based on 5G uplink and downlink service development and 4G frequency reduction service migration according to the present disclosure. As shown in FIG. 7, the intelligent carrier frequency allocation module first calculates the traffic, network utilization, and service growth coefficient of the combined bearing efficiency of the uplink and downlink, and evaluates the necessity of increasing or decreasing the carrier frequency through these coefficients. At the same time, a single-station average limit bearing capacity mutual correlation model and a model adjustment coefficient index and detailed scoring method can be constructed to guide the calculation of the uplink and downlink carrier frequency increase and decrease. Subsequently, a 4G / 5G equivalent carrier conversion carrier frequency allocation overall calculation method can be further constructed, the calculated uplink and downlink carrier frequency increase and decrease are combined with the carrier frequency increase caused by the migration of services to 5G due to 4G frequency reduction, and the total size of the carrier frequency increase and decrease is obtained, so that the carrier frequency allocation scheme at the base station level and at each administrative division level (provincial, city, and regional) can be output.

[0192] The network quality feature division method based on single-station coverage population decomposition provided by the embodiments of the present disclosure can periodically and automatically update and supplement the feature content, ensure sufficient algorithm training data sources, continuously guarantee the implementation effect of the method, reduce the cost of manual intervention, and improve the efficiency of automatic data flow. In addition, by considering the influence of 5G uplink and downlink service development and 4G frequency reduction on service migration, the precision of carrier frequency resource allocation can be effectively improved, which is scientific, efficient, and highly implementable.

[0193] Based on any of the above embodiments, FIG. 8 is a structural diagram of a carrier frequency resource allocation system according to the present disclosure. As shown in FIG. 8, the system includes:

[0194] A feature determination module 810 is configured to obtain coverage population features of a target base station and network quality division results of the coverage population of the target base station, and determine feature information of the target base station based on the coverage population features and the network quality division results.

[0195] A prediction and evaluation module 820 is configured to predict network utilization information of a next time period of the target base station based on the feature information of the target base station.

[0196] A carrier frequency allocation module 830 is configured to determine uplink and downlink carrier frequency increase and decrease amounts of the target base station based on the network utilization information of the current time period of the target base station and the network utilization information of the next time period of the target base station, determine a total carrier frequency increase and decrease amount of the target base station based on the uplink and downlink carrier frequency increase and decrease amounts and a carrier frequency increase amount caused by service migration, and perform carrier frequency resource allocation by applying the total carrier frequency increase and decrease amount.

[0197] The system provided by the embodiments of the present disclosure can determine the characteristic information of the target base station based on the coverage population characteristics of the target base station and the network quality division result of the population covered by the target base station, and can predict accurate network utilization information by applying the characteristic information, so as to support the allocation of carrier frequency resources, thereby effectively solving the network utilization prediction problem caused by long-period prediction difficulty and insufficient model generalization capability. By determining the uplink and downlink carrier frequency increment and decrement of the target base station based on the network utilization information of the target base station in the current period and the network utilization information of the target base station in the next period, and determining the total amount of carrier frequency increment and decrement of the target base station based on the uplink and downlink carrier frequency increment and decrement and the carrier frequency increment caused by service migration, the total amount of carrier frequency increment and decrement can be applied to carrier frequency resource allocation, an intelligent carrier frequency allocation method for service development and bearing efficiency change is realized, the influence of uplink and downlink service change and service migration caused by 4G frequency reduction is quantitatively evaluated, and the accuracy of carrier frequency resource allocation can be effectively improved, and the problems of investment waste or insufficient user protection can be avoided.

[0198] Based on any of the above embodiments, the feature determination module 810 includes: a region information acquisition unit configured to acquire a region population quantity and a region area of a region where the target base station is located; a population scale determination unit configured to determine a coverage population scale of the target base station based on a single-station coverage area of the target base station, the region population quantity, and the region area, wherein the coverage population characteristics include the single-station coverage area and the coverage population scale; and a network quality division unit configured to select sampling points within the single-station coverage area, and perform network quality division on the coverage population scale based on channel signal strength and uplink and downlink rates of the sampling points to obtain the network quality division result.

[0199] Based on any of the above embodiments, the network quality division unit is specifically configured to: calculate network quality division thresholds of the sampling points based on network modulation order distribution characteristics of the sampling points in the current period; and perform network quality division on the coverage population scale based on the network quality division thresholds of the sampling points, and the channel signal strength and uplink and downlink rates of the sampling points to obtain the network quality division result.

[0200] Based on any of the above embodiments, the feature determination module 810 further includes a data processing unit, which is configured to: remove outliers from the network quality division result, and perform data aggregation on the network quality division result after removing the outliers, the coverage population characteristics, and network service traffic data of the target base station to obtain the characteristic information of the target base station.

[0201] Based on any of the above embodiments, the prediction evaluation module 820 is specifically configured to: input the feature information into the network utilization rate prediction model corresponding to the target base station to obtain network utilization rate information of the target base station in the next time period output by the network utilization rate prediction model; and the network utilization rate prediction model is obtained by training based on sample feature information and sample network utilization rate information of the target base station and model accuracy evaluation.

[0202] Based on any of the above embodiments, the objective function of the network utilization rate prediction model includes a training loss function based on a relative error square root and a regularization loss function, and correspondingly, the prediction evaluation module 820 includes a model training unit, which includes: a data acquisition sub-unit configured to acquire a training data set and a test data set of the target base station, the training data set including sample feature information and sample network utilization rate information, and the test data set including test feature information and test network utilization rate information; an information prediction sub-unit configured to input the sample feature information into an initial network utilization rate prediction model to obtain first predicted network utilization rate information output by the initial network utilization rate prediction model; a parameter iteration sub-unit configured to perform parameter iteration on the initial network utilization rate prediction model based on the first predicted network utilization rate information and the sample network utilization rate information to obtain a pre-trained network utilization rate prediction model; and a model evaluation sub-unit configured to apply the test feature information and the test network utilization rate information to perform model accuracy evaluation on the pre-trained network utilization rate prediction model to obtain the network utilization rate prediction model.

[0203] Based on any of the above embodiments, the model evaluation sub-unit is specifically configured to: input the test feature information into the pre-trained network utilization rate prediction model to obtain second predicted network utilization rate information output by the pre-trained network utilization rate prediction model; compare the second predicted network utilization rate information with the test network utilization rate information, and determine a model accuracy evaluation result of the pre-trained network utilization rate prediction model based on the comparison result; in the case that the model accuracy evaluation result is inaccurate, rearrange the features of the test feature information and remove interference features to obtain new test feature information; and apply the new test feature information to re-evaluate the model accuracy of the pre-trained network utilization rate prediction model until the model accuracy evaluation result is accurate, and obtain the network utilization rate prediction model.

[0204] Based on any of the above embodiments, the carrier frequency allocation module 830 comprises: a necessity evaluation unit configured to evaluate the necessity of increasing or decreasing the carrier frequency for the area covered by the target base station based on the network utilization information of the current period of the target base station and the network utilization information of the next period of the target base station; and an increase / decrease amount determination unit configured to determine the uplink and downlink carrier frequency increase / decrease amount of the target base station based on the evaluation result and the average limit carrying capacity correlation model.

[0205] Based on any of the above embodiments, the necessity evaluation unit is specifically configured to: determine the uplink and downlink traffic growth coefficient based on the uplink and downlink traffic in the network utilization information of the current period and the uplink and downlink traffic in the network utilization information of the next period; determine the uplink and downlink utilization rate growth coefficient based on the uplink and downlink channel utilization rate in the network utilization information of the current period and the uplink and downlink channel utilization rate in the network utilization information of the next period; determine the uplink and downlink carrier frequency resource carrying efficiency growth coefficient based on the uplink and downlink traffic growth coefficient and the uplink and downlink utilization rate growth coefficient; and evaluate the necessity of increasing or decreasing the uplink and downlink carrier frequency for the area covered by the target base station based on the uplink and downlink traffic growth coefficient, the uplink and downlink utilization rate growth coefficient, and the uplink and downlink carrier frequency resource carrying efficiency growth coefficient.

[0206] FIG. 9 shows a schematic diagram of the physical structure of an electronic device, as shown in FIG. 9, which can include a processor 910, a communications interface 920, a memory 930, and a communications bus 940, wherein the processor 910, the communications interface 920, and the memory 930 complete mutual communication through the communications bus 940. The processor 910 can invoke the logical instructions in the memory 930 to execute the carrier frequency resource allocation method, which comprises: obtaining the coverage population characteristics of a target base station and the network quality division result of the population covered by the target base station, and determining the characteristic information of the target base station based on the coverage population characteristics and the network quality division result; predicting the network utilization information of the next period of the target base station based on the characteristic information of the target base station; determining the uplink and downlink carrier frequency increase / decrease amount of the target base station based on the network utilization information of the current period of the target base station and the network utilization information of the next period of the target base station; determining the total carrier frequency increase / decrease amount of the target base station based on the uplink and downlink carrier frequency increase / decrease amount and the carrier frequency increase amount caused by service migration, and applying the total carrier frequency increase / decrease amount for carrier frequency resource allocation.

[0207] In addition, the logic instructions in the memory 930 described above can be implemented in the form of software functional units and sold or used as independent products, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present disclosure essentially or contribute to the related art, or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present disclosure. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0208] In another aspect, the present disclosure also provides a computer program product, which comprises a computer program, the computer program can be stored on a non-transitory computer readable storage medium, and the computer program can be executed by a processor to enable a computer to perform the carrier frequency resource allocation method provided by the above-mentioned methods. The method comprises: obtaining the coverage population characteristics of a target base station and the network quality division result of the population covered by the target base station, and determining the characteristic information of the target base station based on the coverage population characteristics and the network quality division result; predicting the network utilization information of the target base station in the next period based on the characteristic information of the target base station; determining the uplink and downlink carrier frequency increment of the target base station based on the network utilization information of the target base station in the current period and the network utilization information of the target base station in the next period; determining the total amount of carrier frequency increment and reduction of the target base station based on the uplink and downlink carrier frequency increment and the carrier frequency increment caused by service migration, and applying the total amount of carrier frequency increment and reduction for carrier frequency resource allocation.

[0209] In yet another aspect, the disclosure also provides a non-transitory computer readable storage medium having stored thereon a computer program, which, when executed by a processor, implements the carrier frequency resource allocation method provided by the above method, and the method comprises: obtaining a coverage population characteristic of a target base station and a network quality division result of a population covered by the target base station, and determining characteristic information of the target base station based on the coverage population characteristic and the network quality division result; predicting network utilization information of a next period of the target base station based on the characteristic information of the target base station; determining an uplink and downlink carrier frequency increment of the target base station based on network utilization information of a current period of the target base station and the network utilization information of the next period of the target base station; determining a total amount of carrier frequency increment and reduction of the target base station based on the uplink and downlink carrier frequency increment and a carrier frequency increment caused by service migration, and performing carrier frequency resource allocation by using the total amount of carrier frequency increment and reduction.

[0210] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, i.e., they can be located in one place or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0211] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software plus necessary universal hardware platforms, and of course, can also be realized by hardware. Based on such understanding, the above technical solutions, essentially or in terms of related art, can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0212] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the disclosure, rather than limit them; although the disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the disclosure.

Claims

1. A method for allocating carrier frequency resources, comprising: Obtaining coverage population characteristics of a target base station and a network quality classification result of the population covered by the target base station, and determining characteristic information of the target base station based on the coverage population characteristics and the network quality classification result; Based on the characteristic information of the target base station, predicting the network utilization information of the target base station in the next time period; Determining an increase or decrease in uplink and downlink carrier frequencies of the target base station based on network utilization information of the target base station in a current period and network utilization information of the target base station in a next period; as well as Based on the uplink and downlink carrier frequency increase and decrease and the carrier frequency increment caused by service migration, the total carrier frequency increase and decrease of the target base station is determined, and the total carrier frequency increase and decrease is applied to perform carrier frequency resource allocation.

2. The carrier frequency resource allocation method according to claim 1, wherein: The obtaining of the coverage population characteristics of the target base station and the network quality division result of the population covered by the target base station includes: Obtaining the population and area of ​​the area where the target base station is located; Determining the population size covered by the target base station based on the single-station coverage area of ​​the target base station, the population size of the region, and the area of ​​the region, wherein the covered population characteristics include the single-station coverage area and the covered population size; and A plurality of sampling points are selected within the coverage area of ​​the single station, and network quality division is performed on the covered population size based on the channel signal strength and uplink and downlink rates of each sampling point to obtain the network quality division result.

3. The carrier frequency resource allocation method according to claim 2, wherein: The performing of network quality classification on the covered population size based on the channel signal strength and uplink and downlink rates of each sampling point to obtain the network quality classification result includes: Calculating a network quality classification threshold for each sampling point based on a network modulation order distribution characteristic of each sampling point in a current time period; and Based on the network quality classification threshold of each sampling point, and the channel signal strength and uplink and downlink rates of each sampling point, network quality classification is performed on the covered population size to obtain the network quality classification result.

4. The carrier frequency resource allocation method according to claim 1, wherein: The determining, based on the covered population characteristics and the network quality classification result, characteristic information of the target base station includes: Outliers are removed from the network quality division results, and data is aggregated on the network quality division results after outliers are removed, the coverage population characteristics, and the network service flow data of the target base station to obtain characteristic information of the target base station.

5. The carrier frequency resource allocation method according to claim 1, wherein: The predicting, based on the characteristic information of the target base station, network utilization information of the target base station in the next time period includes: Inputting the characteristic information into a network utilization prediction model corresponding to the target base station to obtain network utilization information of the target base station in the next time period output by the network utilization prediction model; The network utilization prediction model is based on the sample feature information and sample network utilization information of the target base station and is obtained through training after model accuracy evaluation.

6. The carrier frequency resource allocation method according to claim 5, wherein: The objective function of the network utilization prediction model includes a training loss function based on the square root of relative error and a regularization loss function. The training steps of the network utilization prediction model include: Acquire a training data set and a test data set of the target base station, wherein the training data set includes sample feature information and sample network utilization information, and the test data set includes test feature information and test network utilization information; Inputting the sample feature information into an initial network utilization prediction model to obtain first predicted network utilization information output by the initial network utilization prediction model; Based on the first predicted network utilization information and the sample network utilization information, performing parameter iteration on the initial network utilization prediction model to obtain a pre-trained network utilization prediction model; The test feature information and the test network utilization information are applied to perform a model accuracy evaluation on the pre-trained network utilization prediction model to obtain the network utilization prediction model.

7. The carrier frequency resource allocation method according to claim 6, wherein: The applying the test feature information and the test network utilization information to perform a model accuracy evaluation on the pre-trained network utilization prediction model to obtain the network utilization prediction model includes: Inputting the test feature information into the pre-trained network utilization prediction model to obtain second predicted network utilization information output by the pre-trained network utilization prediction model; Comparing the second predicted network utilization information with the test network utilization information, and determining a model accuracy evaluation result of the pre-trained network utilization prediction model based on the comparison result; When the model accuracy evaluation result is inaccurate, rearrange the test feature information and remove interference features to obtain new test feature information; Applying the new test feature information, re-evaluating the model accuracy of the pre-trained network utilization prediction model until the model accuracy evaluation result is accurate, thereby obtaining the network utilization prediction model.

8. The carrier frequency resource allocation method according to any one of claims 1 to 7, wherein: The determining the increase or decrease amount of the uplink and downlink carrier frequencies of the target base station based on the network utilization information of the target base station in the current time period and the network utilization information of the target base station in the next time period includes: Based on the network utilization information of the target base station in the current period and the network utilization information of the target base station in the next period, evaluating the necessity of increasing or decreasing the carrier frequency for the area covered by the target base station; and Based on the evaluation result and the average limit carrying capacity cross-correlation model, the increase or decrease amount of the uplink and downlink carrier frequencies of the target base station is determined.

9. The carrier frequency resource allocation method according to claim 8, wherein: The performing the necessity evaluation of increasing or decreasing the carrier frequency for the area covered by the target base station based on the network utilization information of the target base station in the current time period and the network utilization information of the target base station in the next time period includes: Determining an uplink and downlink traffic growth coefficient based on the uplink and downlink traffic in the network utilization information of the current period and the uplink and downlink traffic in the network utilization information of the next period; Determining an uplink and downlink utilization rate growth coefficient based on the uplink and downlink channel utilization rates in the network utilization rate information of the current period and the uplink and downlink channel utilization rates in the network utilization rate information of the next period; Determining an uplink and downlink carrier frequency resource carrying efficiency growth coefficient based on the uplink and downlink traffic growth coefficient and the uplink and downlink utilization growth coefficient; Based on the uplink and downlink traffic growth coefficient, the uplink and downlink utilization growth coefficient and the uplink and downlink carrier frequency resource bearing efficiency growth coefficient, the necessity of increasing or decreasing the uplink and downlink carrier frequencies is evaluated for the area covered by the target base station.

10. The carrier frequency resource allocation method according to any one of claims 1 to 9, wherein the carrier frequency increment caused by the service migration is determined based on the increment of 4G frequency reduction demand for 5G capacity expansion, and the increment of 4G frequency reduction demand for 5G capacity expansion is calculated in an equivalent conversion manner using the carrier frequency scale, carrier bandwidth and wireless utilization rate of each frequency band of 4G.

11. A carrier frequency resource allocation system, comprising: A feature determination module is configured to obtain characteristics of a population covered by a target base station and a network quality classification result of the population covered by the target base station, and determine feature information of the target base station based on the characteristics of the population covered and the network quality classification result; A prediction and evaluation module, configured to predict network utilization information of the target base station in the next time period based on the characteristic information of the target base station; The carrier frequency allocation module is used to determine the increase or decrease in the uplink and downlink carrier frequencies of the target base station based on the network utilization information of the target base station in the current time period and the network utilization information of the target base station in the next time period, determine the total increase or decrease in the carrier frequency of the target base station based on the increase or decrease in the uplink and downlink carrier frequencies and the carrier frequency increment caused by service migration, and apply the total increase or decrease in the carrier frequency to perform carrier frequency resource allocation.

12. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the carrier frequency resource allocation method according to any one of claims 1 to 10 is implemented.

13. A non-transitory computer-readable storage medium having a computer program stored thereon, wherein: When the computer program is executed by a processor, the carrier frequency resource allocation method according to any one of claims 1 to 10 is implemented.

14. A computer program product comprising a computer program, wherein When the computer program is executed by a processor, the carrier frequency resource allocation method according to any one of claims 1 to 10 is implemented.

Images