Method, device, electronic device and storage medium for predicting wireless coverage

Abstract

Embodiments of the present application provide a method and device for predicting wireless coverage, electronic equipment and storage medium. The method comprises: obtaining input data of a target area, and obtaining an input feature vector of the target area according to the input data; inputting the input feature vector into a pre-trained wireless coverage prediction model to obtain a predicted coverage result of the target area output by the wireless coverage prediction model; wherein the wireless coverage prediction model is trained based on an automatic hyperparameter tuning gradient boosting decision tree (GBDT) model. In the embodiments of the present application, the automatic hyperparameter tuning gradient boosting decision tree model is used to predict the wireless coverage of the target area, which reduces the complexity of predicting the wireless coverage of the actual antenna, saves a large amount of computing resources, and realizes efficient and accurate wireless coverage prediction.

Description

Technical Field

[0001] This application relates to the fields of wireless communication technology and computer technology, and in particular to a method, apparatus, electronic device and storage medium for predicting wireless coverage. Background Technology

[0002] With the development of communication technology, the number of base station devices has exploded to meet the ever-increasing traffic demand, posing a significant challenge to the deployment of base stations in actual mobile networks. Network planning problems can often be reduced to a hybrid shape programming problem, the complexity of which increases dramatically with the problem size. As a core subproblem of network planning, the efficiency of wireless coverage prediction directly affects the network planning outcome. To ensure that mobile wireless networks can dynamically adjust cell antenna parameters based on real-time coverage conditions, efficient and accurate wireless coverage prediction is required.

[0003] Due to the anisotropy of actual antennas, accurately determining antenna coverage in a given area traditionally requires establishing complex wireless coverage models and calculating the field strength at various locations within the target area, necessitating substantial computational resources. Field strength prediction is a crucial issue in mobile wireless networks. In practical mobile communication system antenna deployment, large-scale field measurements of the target area are impractical due to the high cost and conditions required for on-site road surveys. Therefore, a reasonable field strength prediction model is necessary. Currently, the main field strength prediction models include statistical models and ray tracing models; however, both suffer from excessive complexity when used to predict the wireless coverage of 5G and other mobile wireless networks.

[0004] Therefore, how to achieve efficient and accurate wireless coverage prediction with low complexity has become an urgent technical problem to be solved. Summary of the Invention

[0005] This application provides a method, apparatus, electronic device, and storage medium for predicting wireless coverage, in order to solve the problem of excessive complexity when predicting the actual wireless coverage of antennas, and to achieve efficient and accurate wireless coverage prediction with lower complexity.

[0006] In a first aspect, embodiments of this application provide a method for predicting wireless coverage, including:

[0007] Obtain input data for the target region, and obtain the input feature vector of the target region based on the input data;

[0008] The input feature vector is input into a pre-trained wireless coverage prediction model to obtain the predicted coverage result of the target area output by the wireless coverage prediction model; wherein the wireless coverage prediction model is trained based on an automated hyperparameter tuning gradient boosting decision tree (GBDT) model.

[0009] Optionally, according to the method for predicting wireless coverage in this application, the training process of the automated hyperparameter tuning gradient boosting decision tree (GBDT) model includes:

[0010] Obtain an input sample that includes multiple features, the input sample being used to predict wireless coverage;

[0011] The GBDT ensemble tree is trained based on the input samples, wherein in the process of generating the subtrees of the GBDT ensemble tree, the local optimal split point is found by treating each feature, which includes multiple feature values, as a black box.

[0012] Optionally, according to the method for predicting wireless coverage in embodiments of this application, the step of using each feature, which includes multiple feature values, as a black box to find locally optimal segmentation points includes:

[0013] In the black box of each feature, the cluster center obtained using the mean-shift clustering algorithm will be used as the threshold;

[0014] The loss value is calculated based on each feature and its corresponding threshold, and the threshold corresponding to the minimum loss value is taken as the local optimal split point of the feature.

[0015] Optionally, according to the method for predicting wireless coverage in embodiments of this application, the process of generating a subtree of the GBDT integration tree includes:

[0016] Based on the local optimal split point and its corresponding loss value, the loss value of the subtree is calculated.

[0017] Optionally, according to the method for predicting wireless coverage according to embodiments of this application, training a GBDT ensemble tree based on the input samples includes:

[0018] Set a set of geometric sequences as the update step size parameters for the subtree;

[0019] The hyperparameters other than the update step size parameter are automatically tuned.

[0020] Optionally, in the method for predicting wireless coverage according to embodiments of this application, the automatic tuning of hyperparameters other than the update step size parameter includes:

[0021] A distribution function is set for each of the remaining hyperparameters;

[0022] Based on the remaining hyperparameters and their corresponding distribution functions, the random hyperparameters of the subtree are generated.

[0023] Optionally, in the method for predicting wireless coverage according to embodiments of this application, the input data includes:

[0024] The target area includes terrain information, building information, existing wireless coverage information, and antenna azimuth information.

[0025] Optionally, in the method for predicting wireless coverage according to embodiments of this application, obtaining the input feature vector of the target area based on the input data includes:

[0026] The target area is divided into multiple grids with predetermined areas;

[0027] The input data is assigned to the nearest grid cell according to the proximity principle, and an input feature vector corresponding to each grid cell is constructed.

[0028] Optionally, in the method for predicting wireless coverage according to embodiments of this application, the predicted coverage result of the target area includes at least one of the following:

[0029] Coverage of each grid cell in the target area;

[0030] The overall coverage of the target area.

[0031] Secondly, embodiments of this application also provide an electronic device, including a memory, a transceiver, and a processor;

[0032] The memory is used to store computer programs; the transceiver is used to send and receive data under the control of the processor; the processor is used to read the computer programs in the memory and perform the following operations:

[0033] Obtain input data for the target region, and obtain the input feature vector of the target region based on the input data;

[0034] The input feature vector is input into a pre-trained wireless coverage prediction model to obtain the predicted coverage result of the target area output by the wireless coverage prediction model; wherein the wireless coverage prediction model is trained based on an automated hyperparameter tuning gradient boosting decision tree GBDT model.

[0035] Optionally, in the electronic device according to embodiments of this application, the training process of the automated hyperparameter tuning gradient boosting decision tree (GBDT) model includes:

[0036] Obtain an input sample that includes multiple features, the input sample being used to predict wireless coverage;

[0037] The GBDT ensemble tree is trained based on the input samples, wherein in the process of generating the subtrees of the GBDT ensemble tree, the local optimal split point is found by treating each feature, which includes multiple feature values, as a black box.

[0038] Optionally, in the electronic device according to embodiments of this application, the step of using each feature, which includes multiple feature values, as a black box to find the local optimal segmentation point includes:

[0039] In the black box of each feature, the cluster center obtained using the mean-shift clustering algorithm will be used as the threshold;

[0040] The loss value is calculated based on each feature and its corresponding threshold, and the threshold corresponding to the minimum loss value is taken as the local optimal split point of the feature.

[0041] Optionally, in the electronic device according to embodiments of this application, the process of generating a subtree of the GBDT integration tree includes:

[0042] Based on the local optimal split point and its corresponding loss value, the loss value of the subtree is calculated.

[0043] Optionally, in the electronic device according to embodiments of this application, training the GBDT ensemble tree based on the input samples includes:

[0044] Set a set of geometric sequences as the update step size parameters for the subtree;

[0045] The hyperparameters other than the update step size parameter are automatically tuned.

[0046] Optionally, in the electronic device according to embodiments of this application, the automatic tuning of hyperparameters other than the update step size parameter includes:

[0047] A distribution function is set for each of the remaining hyperparameters;

[0048] Based on the remaining hyperparameters and their corresponding distribution functions, the random hyperparameters of the subtree are generated.

[0049] Optionally, in the electronic device according to embodiments of this application, the input data includes:

[0050] The target area includes terrain information, building information, existing wireless coverage information, and antenna azimuth information.

[0051] Optionally, in the electronic device according to embodiments of this application, the step of obtaining the input feature vector of the target region based on the input data is as follows:

[0052] The target area is divided into multiple grids with predetermined areas;

[0053] The input data is assigned to the nearest grid cell according to the proximity principle, and an input feature vector corresponding to each grid cell is constructed.

[0054] Optionally, in the electronic device according to embodiments of this application, the predicted coverage result of the target area includes at least one of the following:

[0055] Coverage of each grid cell in the target area;

[0056] The overall coverage of the target area.

[0057] Thirdly, embodiments of this application also provide an apparatus for predicting wireless coverage, comprising:

[0058] An input data acquisition unit is used to acquire input data of a target region and obtain an input feature vector of the target region based on the input data.

[0059] The wireless coverage prediction unit is used to input the input data into a pre-trained wireless coverage prediction model to obtain the predicted coverage result of the target area output by the wireless coverage prediction model; wherein the wireless coverage prediction model is trained based on the Gradient Boosting Decision Tree (GBDT) model with automated hyperparameter tuning.

[0060] Fourthly, embodiments of this application also provide a processor-readable storage medium storing a computer program for causing the processor to perform the method for predicting wireless coverage as described in the first aspect above.

[0061] The method, apparatus, electronic device, and storage medium for predicting wireless coverage provided in this application use a pre-trained automated hyperparameter tuning gradient boosting decision tree (GBDT) model to predict the wireless coverage of a target area. This reduces the complexity of predicting the actual wireless coverage of antennas, saves a lot of computing resources, and achieves efficient and accurate wireless coverage prediction. Attached Figure Description

[0062] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 This is one of the flowcharts illustrating the method for predicting wireless coverage provided in the embodiments of this application;

[0064] Figure 2 This is a second schematic flowchart of the method for predicting wireless coverage provided in the embodiments of this application;

[0065] Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application;

[0066] Figure 4 This is a schematic diagram of the structure of the device for predicting wireless coverage provided in the embodiments of this application. Detailed Implementation

[0067] In the embodiments of this application, the term "and / or" describes the relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following associated objects have an "or" relationship.

[0068] Furthermore, the terms "first," "second," etc., are used for descriptive purposes only to distinguish features with similarity, and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature. In the embodiments of this application, the term "multiple" refers to two or more, and other quantifiers are similar.

[0069] In this application, "determining B based on A" means that factor A must be considered when determining B. It is not limited to "B can be determined based solely on A," but should also include: "determining B based on A and C," "determining B based on A, C, and E," "determining C based on A, and further determining B based on C," etc. It can also include using A as a condition for determining B, for example, "when A satisfies the first condition, B is determined using the first method"; or "when A satisfies the second condition, B is determined," or "when A satisfies the third condition, B is determined based on the first parameter," etc. Of course, it can also be a condition where A is a factor in determining B, for example, "when A satisfies the first condition, C is determined using the first method, and B is further determined based on C," etc.

[0070] The technical solutions provided in this application can be applied to various systems, especially 5G systems. For example, applicable systems include Global System for Mobile Communication (GSM), Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA) General Packet Radio Service (GPRS), Long Term Evolution (LTE), LTE Frequency Division Duplex (FDD), LTE Time Division Duplex (TDD), Long Term Evolution Advanced (LTE-A), Universal Mobile Telecommunication System (UMTS), Worldwide Interoperability for Microwave Access (WiMAX), and 5G New Radio (NR). All of these systems include terminal equipment and network equipment. The systems may also include a core network component, such as Evolved Packet System (EPS) and 5G systems (5GS).

[0071] The terminal devices involved in the embodiments of this application can be devices that provide voice and / or data connectivity to users, handheld devices with wireless connectivity, or other processing devices connected to a wireless modem. The names of the terminal devices may differ in different systems; for example, in a 5G system, a terminal device can be called User Equipment (UE). Wireless terminal devices can communicate with one or more core networks (CNs) via a Radio Access Network (RAN). Wireless terminal devices can be mobile terminal devices, such as mobile phones (or "cellular" phones) and computers with mobile terminal devices, for example, portable, pocket-sized, handheld, computer-embedded, or vehicle-mounted mobile devices that exchange voice and / or data with the RAN. Examples include Personal Communication Service (PCS) phones, cordless phones, Session Initiated Protocol (SIP) phones, Wireless Local Loop (WLL) stations, and Personal Digital Assistants (PDAs). Wireless terminal equipment can also be referred to as a system, subscriber unit, subscriber station, mobile station, mobile station, remote station, access point, remote terminal, access terminal, user terminal, user agent, or user device, but is not limited to these terms in the embodiments of this application.

[0072] The network device involved in this application embodiment can be a base station, which may include multiple cells providing services to terminals. Depending on the specific application, a base station may also be called an access point, or a device in an access network that communicates with a wireless terminal device through one or more sectors on the air interface, or other names. The network device can be used to exchange received air frames with Internet Protocol (IP) packets, acting as a router between the wireless terminal device and the rest of the access network, where the rest of the access network may include an Internet Protocol (IP) communication network. The network device can also coordinate the attribute management of the air interface. For example, the network equipment involved in the embodiments of this application can be a base transceiver station (BTS) in a Global System for Mobile communications (GSM) or Code Division Multiple Access (CDMA), a NodeB in a Wide-band Code Division Multiple Access (WCDMA) system, an evolved Node B (eNB or e-NodeB) in a long term evolution (LTE) system, a 5G base station (gNB) in a next generation system, a Home evolved Node B (HeNB), a relay node, a femto, a pico, etc., and is not limited in the embodiments of this application. In some network structures, the network equipment may include centralized unit (CU) nodes and distributed unit (DU) nodes, and the centralized unit and distributed unit may be geographically separated.

[0073] Network devices and terminal devices can each use one or more antennas for multiple-input multiple-output (MIMO) transmission. MIMO transmission can be single-user MIMO (SU-MIMO) or multiple-user MIMO (MU-MIMO). Depending on the configuration and number of antenna combinations, MIMO transmission can be 2D-MIMO, 3D-MIMO, FD-MIMO, or massive-MIMO, and can also be diversity transmission, precoding transmission, or beamforming transmission, etc.

[0074] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0075] When predicting the wireless coverage of mobile wireless networks, a reasonable field strength prediction model needs to be established. Currently, the main field strength prediction models include statistical models and ray tracing models. However, both of these models suffer from excessive complexity when used to predict the wireless coverage of mobile wireless networks.

[0076] Statistical models utilize empirical field strength test information and employ statistical methods to fit equations relating field strength to parameters such as transmitter-receiver distance, base station height, and signal frequency. Due to their simplicity, statistical models are currently the most commonly used field strength prediction model. However, because some factors affecting field strength in real-world environments are difficult to represent effectively using simple mathematical forms, the model's accuracy is not guaranteed. In practical engineering, the target area is typically divided into several sub-regions, and statistical models are used to predict the field strength for each sub-region separately to improve prediction accuracy. The more detailed the sub-region division, the higher the field strength prediction accuracy, but the greater the model complexity, and its accuracy has an upper limit.

[0077] Ray tracing models utilize electromagnetic wave propagation theory, considering phenomena such as reflection and refraction, to analyze the actual path of electromagnetic waves from the transmitter to the receiver. By superimposing multiple paths, the field strength information at the receiver is calculated. Using ray tracing models and forward and reverse algorithms, accurate prediction of field strength in a small three-dimensional area is achieved. Compared to statistical models, this model more closely reflects the electromagnetic wave propagation patterns in real-world scenarios, achieving high prediction accuracy. However, for each point in the target area, the ray tracing model requires calculating the path from the transmitter to that point. Since electromagnetic wave propagation paths in dense urban networks are typically very complex, the complexity of this model far exceeds that of statistical models, making it difficult to apply to large-scale coverage areas.

[0078] Traditional solution methods determine the quality of antenna parameters by accurately predicting field strength. Although they are highly accurate, they require a lot of computing resources and cannot meet the actual needs of 5G and post-5G mobile networks to dynamically adjust radio frequency parameters through real-time coverage prediction.

[0079] This application provides a method for predicting wireless coverage, which solves the problem of excessive complexity when predicting the actual wireless coverage of antennas, and achieves efficient and accurate wireless coverage prediction with lower complexity.

[0080] Figure 1 This is one of the flowcharts illustrating the method for predicting wireless coverage provided in this application embodiment, see below. Figure 1 The method for predicting wireless coverage provided in this application includes:

[0081] Step 101: Obtain the input data of the target region, and obtain the input feature vector of the target region based on the input data;

[0082] Step 102: Input the input feature vector into the pre-trained wireless coverage prediction model to obtain the predicted coverage result of the target area output by the wireless coverage prediction model; wherein the wireless coverage prediction model is trained based on the Gradient Boosting Decision Tree (GBDT) model with automated hyperparameter tuning.

[0083] Specifically, in this embodiment, input data for the target area is first acquired. This input data includes antenna parameters of base stations and other transceiver points in the target area, wireless environmental information affecting electromagnetic wave propagation, etc. By performing necessary processing and preprocessing on the input data, an input feature vector suitable for a pre-trained wireless coverage prediction model is obtained. Then, this input feature vector is input into the wireless coverage prediction model to obtain the wireless coverage prediction result.

[0084] The wireless coverage prediction model is trained based on an automated hyperparameter tuning GBDT model. GBDT is a decision tree ensemble algorithm where input feature vectors are constructed into trees. After successfully building an initial subtree, the new subtree corrects the errors of existing trees in the model. Tree construction stops when there is no further improvement in the model's classification results. The sum of the scores of all leaf nodes corresponds to the score for evaluating the classification category, and the category with the highest score is selected as the final classification category. Using the GBDT model to predict the feature vectors of the target area allows the fusion model to more fully learn the true relationship between data and labels, achieving more efficient prediction results. In this embodiment, based on the input data of the wireless coverage model and the desired wireless coverage prediction function, the wireless coverage prediction problem of the target area is abstracted into a regression problem with multiple samples, and the wireless coverage prediction is achieved using an automated hyperparameter tuning GBDT model.

[0085] The method for predicting wireless coverage provided in this application employs an automated hyperparameter tuning GBDT model to predict the coverage of specified antenna parameters in a target area. It eliminates the need for complex field strength prediction algorithms, thus saving significant computational resources. This provides a new approach and method for dynamic planning and optimization of mobile communication networks, achieving efficient and accurate wireless coverage prediction. The method provided in this application significantly reduces computational load while maintaining accuracy, offering a fundamental reference for dynamic network planning such as 5G.

[0086] Optionally, according to the method for predicting wireless coverage in this application, the training process of the automated hyperparameter tuning gradient boosting decision tree (GBDT) model includes:

[0087] Obtain an input sample that includes multiple features, the input sample being used to predict wireless coverage;

[0088] The GBDT ensemble tree is trained based on the input samples, wherein in the process of generating the subtrees of the GBDT ensemble tree, the local optimal split point is found by treating each feature, which includes multiple feature values, as a black box.

[0089] Specifically, the first step is to obtain input samples for predicting wireless coverage. Each sample has 20 features, namely terrain information, building information, existing coverage information, and azimuth information of the target area. These features form a feature vector, which is then input into the base model of GBDT for training.

[0090] Then, a GBDT ensemble tree is trained based on the feature vectors. GBDT regression trains multiple subtrees for the prediction problem. In this embodiment, a set of subtrees is trained on the input data of the input samples used to predict wireless coverage. The root node of subtree 0 (i.e., the first subtree) is the average of the predicted values ​​of all sample data. Next, the second subtree, subtree 1, is built. Specifically, this involves first calculating the residuals for all samples, and then building subtree 1 along the direction of the fastest negative gradient descent of the loss function. This subtree building process is iterated until the model error meets the requirements, thus obtaining a complete GBDT ensemble tree, which can be used to predict the wireless coverage of the target area. GBDT is a boosting method, the main idea of ​​which is that the new model in each iteration is built by building the previous model along the direction of the fastest negative gradient descent of the loss function.

[0091] During the subtree construction process, each time a node splits, all split points for each feature must be analyzed to find the local optimum split point. Then, these local optimum split points are compared to obtain the optimal split variable and the corresponding optimal split point. Generally, each feature value is substituted into the calculation, and the local optimum split point is found by traversing each feature value. However, this process consumes a lot of computing resources and time, and has high time complexity. Therefore, the embodiments of this application improve the process of finding the local optimum split point. Instead of substituting each feature value into the calculation, each feature is regarded as a black box, and all feature values ​​of each feature are considered to be put into the black box. For each black box, several different thresholds are set according to the distribution of feature values ​​to calculate the benefit after splitting, i.e., the benefit after splitting. Splitting means dividing these black boxes into two groups to ensure that the loss value in each black box is minimized. Then, the black boxes are merged, and the benefit is the value related to the negative gradient of the loss function. A black box is a concept in the field of computer science. Its characteristics are similar to an opaque black box. When operating on a black box, one only focuses on its input and output, without paying attention to or being able to know its internal implementation and operation process.

[0092] Optionally, according to the method for predicting wireless coverage in embodiments of this application, the step of using each feature, which includes multiple feature values, as a black box to find locally optimal segmentation points includes:

[0093] In the black box of each feature, the cluster center obtained using the mean-shift clustering algorithm will be used as the threshold;

[0094] The loss value is calculated based on each feature and its corresponding threshold, and the threshold corresponding to the minimum loss value is taken as the local optimal split point of the feature.

[0095] Specifically, in this embodiment, taking the generation process of subtree 1 as an example, the generation process of subsequent subtrees is similar to that of subtree 1. The specific calculation formula for generating subtree 1 includes the following formula:

[0096] [s1,s2,…,s k ]→[c1,c2,…,c m (1)

[0097]

[0098] In equation (1) above, s represents all values ​​of each feature, k represents the number of values ​​of the current feature, c represents the threshold of the feature, m represents the number of segmentation thresholds, and → represents the mean-shift clustering algorithm. This algorithm is used to find dense regions of data points based on a sliding window. It is a centroid-based algorithm, which specifically locates the center point of each group of data by updating the candidate center points to the mean of the points within the sliding window. Then, similar windows are removed from these candidate windows, and finally, a set of center points and corresponding groups are formed. In the embodiment of this application, the cluster center of each feature is obtained according to the mean-shift clustering algorithm, and the cluster center is used as the threshold. In equation (2) above, L j Let (j, c) be the loss for all thresholds c for feature j, where (j, c) represents feature j and its corresponding threshold c, and x is the sample data. Subtree Left and Subtree Right The calculation results correspond to the left subtree space and the right subtree space, respectively. j represents the nth feature and yi represents the value assigned to the corresponding R space. Each leaf node corresponds to a space. The input space is divided using this key-value pair (j, c). Then, the Subtree within the divided space is taken as its output value, i.e., the predicted value of the leaf node.

[0099] This application improves the process of finding the local optimal split point by adopting the mean-shift clustering algorithm to obtain the cluster center of each feature and using the cluster center as a threshold, thereby reducing the time complexity and saving a lot of computing resources and time.

[0100] Optionally, according to the method for predicting wireless coverage in embodiments of this application, the process of generating a subtree of the GBDT integration tree includes:

[0101] Based on the local optimal split point and its corresponding loss value, the loss value of the subtree is calculated.

[0102] Specifically, in this embodiment of the application, taking the specific generation process of subtree 1 as an example, the generation process of subsequent subtrees is similar to that of subtree 1. The specific calculation formula for generating subtree 1 also includes the following formula:

[0103] L = min L j (j=1,…,o) (3)

[0104] In equation (3) above, L is the loss function for the spanning tree, L j Let $\frac{ ...

[0105] This application embodiment adopts the above-described specific method for generating subtrees. By improving the process of generating subtrees of the GBDT ensemble tree, it completes the two-step calculation of tree generation and pruning in one step, which greatly reduces the amount of computation. Thus, while maintaining a similar accuracy, it significantly speeds up the algorithm's computation time and significantly improves training efficiency.

[0106] Optionally, according to the method for predicting wireless coverage according to embodiments of this application, training a GBDT ensemble tree based on the input samples includes:

[0107] Set a set of geometric sequences as the update step size parameters for the subtree;

[0108] The hyperparameters other than the update step size parameter are automatically tuned.

[0109] Specifically, after the current tree is built, update the GBDT model prediction value Y. n (x), as shown in equation (4):

[0110]

[0111] In equation (4) above, n represents the nth tree; x represents the input feature vector of the raster sample data; Y n (x) represents the predicted coverage value of x; P is the update step size, which is the distance of each update. Since the previously built trees have a greater effect on reducing the loss, a set of geometric sequences is set as the update step size parameter of the current tree to highlight the role of the previously built trees. Other hyperparameters are automatically tuned. In the context of machine learning, hyperparameters are parameters whose values ​​are set before the learning process begins, rather than parameter data obtained through training. Usually, it is necessary to optimize the hyperparameters and select a set of optimal hyperparameters for the model to improve the learning performance and effect. Here is an explanation: hyperparameters include the maximum number of features considered when partitioning, the maximum depth of the decision tree, the minimum number of samples in the leaf node, the maximum number of leaf nodes, the maximum number of trees built, etc. Equation (4) will be further explained below.

[0112] Optionally, in the method for predicting wireless coverage according to embodiments of this application, the automatic tuning of hyperparameters other than the update step size parameter includes:

[0113] A distribution function is set for each of the remaining hyperparameters;

[0114] Based on the remaining hyperparameters and their corresponding distribution functions, the random hyperparameters of the subtree are generated.

[0115] Specifically, a distribution function is defined for each of the remaining hyperparameters to generate random hyperparameters. This has a significant advantage for some insensitive hyperparameters, allowing for fewer parameter combinations compared to the overall parameter space. Probability distribution search also allows for fine-grained repeated search optimization. In equation (4) above, γn is the current predicted value of the leaf node; R j,n The spaces are segmented based on features. Each leaf node corresponds to a space, with the first index representing the leaf node number and the second index representing the tree number. 'I' indicates that if x belongs to this space, it is 1; otherwise, it is 0. 'J' represents the tree number. n This indicates the number of leaf nodes.

[0116] In this embodiment, the generation of subtree 1 is achieved by employing the specific subtree generation method described above. The predicted value of subtree 1 is closer to the true sample label than the predicted value of subtree 0. Next, subtree 2 is built. The process of building subtree 2 is equivalent to repeating the subtree 1 generation process described above, iteratively generating subtrees until the error of the predicted value output by the model meets the requirements, thereby obtaining the complete GBDT ensemble tree.

[0117] Optionally, in the method for predicting wireless coverage according to embodiments of this application, the input data includes:

[0118] The target area includes terrain information, building information, existing wireless coverage information, and antenna azimuth information.

[0119] Specifically, to confirm the propagation path from the transmitter to the receiver, it is necessary to know the terrain and building information of the target area. Considering that in a real environment, an area is often served by multiple base stations and multiple antennas working together, the existing wireless coverage of the area must also be used as input. Since antennas are anisotropic, the antenna azimuth angle also affects coverage. Terrain information data mainly includes the latitude and longitude, altitude, etc. of the site. Building information data mainly refers to building map data, including building altitude data, ground feature data, street vector data, etc. Ground features refer to fixed objects on the earth's surface, including naturally formed and man-made objects, generally referring to relatively fixed objects on the earth's surface. Existing wireless coverage and antenna azimuth angle information data mainly include equipment hardware parameters, antenna feeder parameters, tower height, etc.

[0120] Optionally, in the method for predicting wireless coverage according to embodiments of this application, obtaining the input feature vector of the target area based on the input data includes:

[0121] The target area is divided into multiple grids with predetermined areas;

[0122] The input data is assigned to the nearest grid cell according to the proximity principle, and an input feature vector corresponding to each grid cell is constructed.

[0123] Specifically, the embodiments of this application can be applied to densely deployed cell scenarios, where antenna density is very high and the radiation range of each antenna is relatively small. Assuming the antennas only affect an area within a 300-400m radius, to accurately describe the area information, the target area is divided into several 5m×5m grids. Geographic data rasterization involves uniformly and regularly dividing the entire map into multiple squares of the same size. Feature data directly related to geographic data is assigned to the nearest grid according to the proximity principle. Statistical characteristics within each grid area are characterized using terrain information, building information, and existing coverage information. Since existing mobile communication networks have a large number of base stations, to control computational load and improve simulation speed, the calculation of the coverage area of ​​each base station is usually controlled within a square area centered on the base station and encompassing several grids. Half the side length of this square area is called the computational radius of the base station. The calculation radius is determined based on the coverage scenario. For example, the coverage radius of an F-band LTE base station in an urban area is generally 300-400m, so the calculation radius can be taken as 600-800m; the coverage radius of an F-band LTE base station in a rural area is generally around 1000m, so the calculation radius can be taken as around 2000m. In summary, the terrain information, building information, existing coverage information, and orientation information of each grid cell in the target area are used to form the input feature vector.

[0124] Optionally, in the method for predicting wireless coverage according to embodiments of this application, the predicted coverage result of the target area includes at least one of the following:

[0125] Coverage of each grid cell in the target area;

[0126] The overall coverage of the target area.

[0127] Specifically, the wireless coverage prediction model provided in this application provides an output of real numbers from 0 to 1, used to represent the coverage status of each grid cell, where 0 indicates no coverage and 1 indicates complete coverage. For the output, the wireless coverage problem is concerned with the overall coverage effect; therefore, the overall coverage rate can be used as the output. The overall target area includes multiple 5m×5m grid cells, and the coverage rate of each grid cell is a real number from 0 to 1. The overall coverage rate can then be obtained by averaging the coverage rates of all grid cells.

[0128] Figure 2This is a second flowchart illustrating the method for predicting wireless coverage provided in this application embodiment. (Refer to...) Figure 2 The wireless coverage prediction model training process provided in this application embodiment includes:

[0129] Step 201: Input sample data;

[0130] Step 202: Find the optimal split point for node splitting, where each feature including multiple feature values ​​is treated as a black box to find the local optimal split point;

[0131] Step 203: Generate subtrees and calculate the loss function of the subtrees;

[0132] Step 204: Iteratively generate subtrees along the direction of the fastest negative gradient descent of the loss function until the training of the GBDT model is completed.

[0133] Specifically, the terrain information, building information, existing coverage information, and orientation information of each grid cell in the sample data are combined into a feature vector, which is then input into the GBDT model for training. The GBDT model takes 20 feature vectors as input and outputs continuous variables from 0 to 1. The cluster center for each feature is obtained using the mean-shift clustering algorithm, and this cluster center is used as a threshold. A loss value is calculated for each feature and its corresponding threshold, and the threshold corresponding to the minimum loss value is taken as the local optimum split point for that feature. The loss value of the subtree is calculated according to the loss function described in the above embodiment, and a subtree is built along the direction of the fastest negative gradient descent of the loss function. This subtree building process is iterated until the model error meets the requirements, thus obtaining a complete GBDT ensemble tree. This GBDT ensemble tree can be used to predict the wireless coverage of a target area.

[0134] In actual testing, the method for predicting wireless coverage provided in this application can achieve the following effect: it can clearly and specifically show the coverage of each grid in the target area and output an accurate overall coverage rate. In one test, the predicted coverage rate was 98.23% and the actual coverage rate was 98.18%. It can be seen that the predicted coverage rate obtained by the method for predicting wireless coverage provided in this application can characterize the effective coverage rate of the actual antenna.

[0135] Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application, such as... Figure 3 As shown, the electronic device includes a memory 320, a transceiver 300, and a processor 310, wherein:

[0136] The memory 320 is used to store computer programs; the transceiver 300 is used to send and receive data under the control of the processor 310; the processor 310 is used to read the computer program in the memory 320 and perform the following operations:

[0137] Obtain input data for the target region, and obtain the input feature vector of the target region based on the input data;

[0138] The input feature vector is input into a pre-trained wireless coverage prediction model to obtain the predicted coverage result of the target area output by the wireless coverage prediction model; wherein the wireless coverage prediction model is trained based on an automated hyperparameter tuning gradient boosting decision tree GBDT model.

[0139] Specifically, transceiver 300 is used to receive and send data under the control of processor 310.

[0140] Among them, Figure 3 In this context, the bus architecture can include any number of interconnected buses and bridges, specifically linking various circuits together, represented by one or more processors (processor 310) and memory (memory 320). The bus architecture can also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. Bus interface 330 provides an interface. Transceiver 300 can be multiple elements, including transmitters and receivers, providing units for communicating with various other devices over transmission media, including wireless channels, wired channels, optical fibers, etc. Processor 310 is responsible for managing the bus architecture and general processing, and memory 320 can store data used by processor 310 during operation.

[0141] The processor 310 can be a central processing unit (CPU), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a complex programmable logic device (CPLD). The processor can also adopt a multi-core architecture.

[0142] Optionally, in the electronic device according to embodiments of this application, the training process of the automated hyperparameter tuning gradient boosting decision tree (GBDT) model includes:

[0143] Obtain an input sample that includes multiple features, the input sample being used to predict wireless coverage;

[0144] The GBDT ensemble tree is trained based on the input samples, wherein in the process of generating the subtrees of the GBDT ensemble tree, the local optimal split point is found by treating each feature, which includes multiple feature values, as a black box.

[0145] Optionally, in the electronic device according to embodiments of this application, the step of using each feature, which includes multiple feature values, as a black box to find the local optimal segmentation point includes:

[0146] In the black box of each feature, the cluster center obtained using the mean-shift clustering algorithm will be used as the threshold;

[0147] The loss value is calculated based on each feature and its corresponding threshold, and the threshold corresponding to the minimum loss value is taken as the local optimal split point of the feature.

[0148] Optionally, in the electronic device according to embodiments of this application, the process of generating a subtree of the GBDT integration tree includes:

[0149] Based on the local optimal split point and its corresponding loss value, the loss value of the subtree is calculated.

[0150] Optionally, in the electronic device according to embodiments of this application, training the GBDT ensemble tree based on the input samples includes:

[0151] Set a set of geometric sequences as the update step size parameters for the subtree;

[0152] The hyperparameters other than the update step size parameter are automatically tuned.

[0153] Optionally, in the electronic device according to embodiments of this application, the automatic tuning of hyperparameters other than the update step size parameter includes:

[0154] A distribution function is set for each of the remaining hyperparameters;

[0155] Based on the remaining hyperparameters and their corresponding distribution functions, the random hyperparameters of the subtree are generated.

[0156] Optionally, in the electronic device according to embodiments of this application, the input data includes:

[0157] The target area includes terrain information, building information, existing wireless coverage information, and antenna azimuth information.

[0158] Optionally, in the electronic device according to embodiments of this application, the step of obtaining the input feature vector of the target region based on the input data is as follows:

[0159] The target area is divided into multiple grids with predetermined areas;

[0160] The input data is assigned to the nearest grid cell according to the proximity principle, and an input feature vector corresponding to each grid cell is constructed.

[0161] Optionally, in the electronic device according to embodiments of this application, the predicted coverage result of the target area includes at least one of the following:

[0162] Coverage of each grid cell in the target area;

[0163] The overall coverage of the target area.

[0164] It should be noted that the electronic device provided in this application embodiment can implement all the method steps implemented by the method embodiment with the execution subject being an electronic device, and can achieve the same technical effect. Here, the parts that are the same as those in the method embodiment and the beneficial effects will not be described in detail.

[0165] This application provides a method and apparatus for predicting wireless coverage, which solves the problem of excessive complexity when predicting the actual wireless coverage of antennas, and achieves efficient and accurate wireless coverage prediction with lower complexity.

[0166] The method and apparatus are based on the same concept of the application. Since the methods and apparatus solve problems in similar ways, the implementation of the apparatus and methods can refer to each other, and the repeated parts will not be described again.

[0167] Figure 4 This is a schematic diagram of the structure of the device for predicting wireless coverage provided in the embodiments of this application, with reference to... Figure 4 The device for predicting wireless coverage provided in this application includes:

[0168] The input data acquisition unit 410 is used to acquire input data of the target region and obtain the input feature vector of the target region based on the input data;

[0169] The wireless coverage prediction unit 420 is used to input the input data into a pre-trained wireless coverage prediction model to obtain the predicted coverage result of the target area output by the wireless coverage prediction model; wherein the wireless coverage prediction model is trained based on the Gradient Boosting Decision Tree (GBDT) model with automated hyperparameter tuning.

[0170] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a processor-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0171] Optionally, in the apparatus for predicting wireless coverage according to embodiments of this application, the training process of the automated hyperparameter tuning gradient boosting decision tree (GBDT) model includes:

[0172] Obtain an input sample that includes multiple features, the input sample being used to predict wireless coverage;

[0173] The GBDT ensemble tree is trained based on the input samples, wherein in the process of generating the subtrees of the GBDT ensemble tree, the local optimal split point is found by treating each feature, which includes multiple feature values, as a black box.

[0174] Optionally, in the apparatus for predicting wireless coverage according to embodiments of this application, the step of using each feature, which includes multiple feature values, as a black box to find locally optimal segmentation points includes:

[0175] In the black box of each feature, the cluster center obtained using the mean-shift clustering algorithm will be used as the threshold;

[0176] The loss value is calculated based on each feature and its corresponding threshold, and the threshold corresponding to the minimum loss value is taken as the local optimal split point of the feature.

[0177] Optionally, in the apparatus for predicting wireless coverage according to embodiments of this application, the process of generating a subtree of the GBDT integration tree includes:

[0178] Based on the local optimal split point and its corresponding loss value, the loss value of the subtree is calculated.

[0179] Optionally, in the apparatus for predicting wireless coverage according to embodiments of this application, training the GBDT ensemble tree based on the input samples includes:

[0180] Set a set of geometric sequences as the update step size parameters for the subtree;

[0181] The hyperparameters other than the update step size parameter are automatically tuned.

[0182] Optionally, in the apparatus for predicting wireless coverage according to embodiments of this application, the automatic tuning of hyperparameters other than the update step size parameter includes:

[0183] A distribution function is set for each of the remaining hyperparameters;

[0184] Based on the remaining hyperparameters and their corresponding distribution functions, the random hyperparameters of the subtree are generated.

[0185] Optionally, in the apparatus for predicting wireless coverage according to embodiments of this application, the input data includes:

[0186] The target area includes terrain information, building information, existing wireless coverage information, and antenna azimuth information.

[0187] Optionally, in the apparatus for predicting wireless coverage according to embodiments of this application, the step of obtaining the input feature vector of the target area based on the input data is as follows:

[0188] The target area is divided into multiple grids with predetermined areas;

[0189] The input data is assigned to the nearest grid cell according to the proximity principle, and an input feature vector corresponding to each grid cell is constructed.

[0190] Optionally, in the apparatus for predicting wireless coverage according to embodiments of this application, the predicted coverage result of the target area includes at least one of the following:

[0191] Coverage of each grid cell in the target area;

[0192] The overall coverage of the target area.

[0193] It should be noted that the apparatus provided in this embodiment of the invention can implement all the method steps implemented in the above method embodiment and can achieve the same technical effect. Therefore, the parts and beneficial effects that are the same as those in the method embodiment will not be described in detail here.

[0194] On the other hand, embodiments of this application also provide a processor-readable storage medium storing a computer program for causing the processor to execute the method for predicting wireless coverage provided in the above embodiments, including: acquiring input data of a target area and obtaining an input feature vector of the target area based on the input data; inputting the input feature vector into a pre-trained wireless coverage prediction model to obtain a predicted coverage result of the target area output by the wireless coverage prediction model; wherein the wireless coverage prediction model is trained based on an automated hyperparameter tuning gradient boosting decision tree (GBDT) model.

[0195] The processor-readable storage medium can be any available medium or data storage device that the processor can access, including but not limited to magnetic memory (e.g., floppy disk, hard disk, magnetic tape, magneto-optical disk (MO)), optical memory (e.g., CD, DVD, BD, HVD), and semiconductor memory (e.g., ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)).

[0196] Optionally, according to the processor-readable storage medium of the embodiments of this application, the training process of the automated hyperparameter tuning gradient boosting decision tree GBDT model includes:

[0197] Obtain an input sample that includes multiple features, the input sample being used to predict wireless coverage;

[0198] The GBDT ensemble tree is trained based on the input samples, wherein in the process of generating the subtrees of the GBDT ensemble tree, the local optimal split point is found by treating each feature, which includes multiple feature values, as a black box.

[0199] Optionally, according to the processor-readable storage medium of the embodiments of this application, the step of using each feature, which includes multiple feature values, as a black box to find the local optimal segmentation point includes:

[0200] In the black box of each feature, the cluster center obtained using the mean-shift clustering algorithm will be used as the threshold;

[0201] The loss value is calculated based on each feature and its corresponding threshold, and the threshold corresponding to the minimum loss value is taken as the local optimal split point of the feature.

[0202] Optionally, according to the processor-readable storage medium of the embodiments of this application, the process of generating a subtree of the GBDT integration tree includes:

[0203] Based on the local optimal split point and its corresponding loss value, the loss value of the subtree is calculated.

[0204] Optionally, according to the processor-readable storage medium of the embodiments of this application, training the GBDT ensemble tree based on the input samples includes:

[0205] Set a set of geometric sequences as the update step size parameters for the subtree;

[0206] The hyperparameters other than the update step size parameter are automatically tuned.

[0207] Optionally, according to the processor-readable storage medium of the embodiments of this application, the automatic tuning of hyperparameters other than the update step size parameter includes:

[0208] A distribution function is set for each of the remaining hyperparameters;

[0209] Based on the remaining hyperparameters and their corresponding distribution functions, the random hyperparameters of the subtree are generated.

[0210] Optionally, according to the processor-readable storage medium of the present application embodiments, the input data includes:

[0211] The target area includes terrain information, building information, existing wireless coverage information, and antenna azimuth information.

[0212] Optionally, according to the processor-readable storage medium of the present application embodiment, the step of obtaining the input feature vector of the target region based on the input data is as follows:

[0213] The target area is divided into multiple grids with predetermined areas;

[0214] The input data is assigned to the nearest grid cell according to the proximity principle, and an input feature vector corresponding to each grid cell is constructed.

[0215] Optionally, according to the processor-readable storage medium of the embodiments of this application, the predicted coverage result of the target region includes at least one of the following:

[0216] Coverage of each grid cell in the target area;

[0217] The overall coverage of the target area.

[0218] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0219] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-executable instructions. These computer-executable instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0220] These processor-executable instructions may also be stored in a processor-readable memory that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the processor-readable memory produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0221] These processors can execute instructions that can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0222] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for predicting wireless coverage, characterized in that, include: Obtain input data for the target region, and obtain the input feature vector of the target region based on the input data; The input feature vector is input into a pre-trained wireless coverage prediction model to obtain the predicted coverage result of the target area output by the wireless coverage prediction model; wherein the wireless coverage prediction model is trained based on an automated hyperparameter tuning gradient boosting decision tree GBDT model. The training process of the automated hyperparameter tuning gradient boosting decision tree (GBDT) model includes: Obtain an input sample that includes multiple features, the input sample being used to predict wireless coverage; The GBDT ensemble tree is trained based on the input samples, wherein in the process of generating the subtrees of the GBDT ensemble tree, local optimal split points are found by treating each feature, which includes multiple feature values, as a black box; the step of treating each feature, which includes multiple feature values, as a black box to find local optimal split points includes: In the black box of each feature, the cluster center obtained using the mean-shift clustering algorithm will be used as the threshold; The loss value is calculated based on each feature and its corresponding threshold, and the threshold corresponding to the minimum loss value is taken as the local optimal split point of the feature.

2. The method for predicting wireless coverage according to claim 1, characterized in that, The process of generating a subtree of the GBDT ensemble tree includes: Based on the local optimal split point and its corresponding loss value, the loss value of the subtree is calculated.

3. The method for predicting wireless coverage according to claim 1, characterized in that, The training of the GBDT ensemble tree based on the input samples includes: Set a set of geometric sequences as the update step size parameters for the subtree; The hyperparameters other than the update step size parameter are automatically tuned.

4. The method for predicting wireless coverage according to claim 3, characterized in that, The automated tuning of hyperparameters other than the update step size parameter includes: A distribution function is set for each of the remaining hyperparameters; Based on the remaining hyperparameters and their corresponding distribution functions, the random hyperparameters of the subtree are generated.

5. The method for predicting wireless coverage according to claim 1, characterized in that, The input data includes: The target area includes terrain information, building information, existing wireless coverage information, and antenna azimuth information.

6. The method for predicting wireless coverage according to claim 5, characterized in that, The step of obtaining the input feature vector of the target region based on the input data includes: The target area is divided into multiple grids with predetermined areas; The input data is assigned to the nearest grid cell according to the proximity principle, and an input feature vector corresponding to each grid cell is constructed.

7. The method for predicting wireless coverage according to claim 6, characterized in that, The predicted coverage result of the target area includes at least one of the following: Coverage of each grid cell in the target area; The overall coverage of the target area.

8. An electronic device, characterized in that, Includes memory, transceiver, and processor; The memory is used to store computer programs; the transceiver is used to send and receive data under the control of the processor; the processor is used to read the computer programs in the memory and perform the following operations: Obtain input data for the target region, and obtain the input feature vector of the target region based on the input data; The input feature vector is input into a pre-trained wireless coverage prediction model to obtain the predicted coverage result of the target area output by the wireless coverage prediction model; wherein the wireless coverage prediction model is trained based on an automated hyperparameter tuning gradient boosting decision tree GBDT model. The training process of the automated hyperparameter tuning gradient boosting decision tree (GBDT) model includes: Obtain an input sample that includes multiple features, the input sample being used to predict wireless coverage; The GBDT ensemble tree is trained based on the input samples, wherein in the process of generating the subtrees of the GBDT ensemble tree, local optimal split points are found by treating each feature, which includes multiple feature values, as a black box; the step of treating each feature, which includes multiple feature values, as a black box to find local optimal split points includes: In the black box of each feature, the cluster center obtained using the mean-shift clustering algorithm will be used as the threshold; The loss value is calculated based on each feature and its corresponding threshold, and the threshold corresponding to the minimum loss value is taken as the local optimal split point of the feature.

9. The electronic device according to claim 8, characterized in that, The process of generating a subtree of the GBDT ensemble tree includes: Based on the local optimal split point and its corresponding loss value, the loss value of the subtree is calculated.

10. The electronic device according to claim 8, characterized in that, The training of the GBDT ensemble tree based on the input samples includes: Set a set of geometric sequences as the update step size parameters for the subtree; The hyperparameters other than the update step size parameter are automatically tuned.

11. The electronic device according to claim 10, characterized in that, The automated tuning of hyperparameters other than the update step size parameter includes: A distribution function is set for each of the remaining hyperparameters; Based on the remaining hyperparameters and their corresponding distribution functions, the random hyperparameters of the subtree are generated.

12. The electronic device according to claim 8, characterized in that, The input data includes: The target area includes terrain information, building information, existing wireless coverage information, and antenna azimuth information.

13. The electronic device according to claim 12, characterized in that, The input feature vector of the target region is obtained based on the input data: The target area is divided into multiple grids with predetermined areas; The input data is assigned to the nearest grid cell according to the proximity principle, and an input feature vector corresponding to each grid cell is constructed.

14. The electronic device according to claim 13, characterized in that, The predicted coverage result of the target area includes at least one of the following: Coverage of each grid cell in the target area; The overall coverage of the target area.

15. A device for predicting wireless coverage, characterized in that, include: An input data acquisition unit is used to acquire input data of a target region and obtain an input feature vector of the target region based on the input data. A wireless coverage prediction unit is used to input the input data into a pre-trained wireless coverage prediction model to obtain the predicted coverage result of the target area output by the wireless coverage prediction model. The wireless coverage prediction model mentioned above is trained based on an automated hyperparameter tuning gradient boosting decision tree GBDT model. The training process of the automated hyperparameter tuning gradient boosting decision tree (GBDT) model includes: Obtain an input sample that includes multiple features, the input sample being used to predict wireless coverage; The GBDT ensemble tree is trained based on the input samples, wherein in the process of generating the subtrees of the GBDT ensemble tree, local optimal split points are found by treating each feature, which includes multiple feature values, as a black box; the step of treating each feature, which includes multiple feature values, as a black box to find local optimal split points includes: In the black box of each feature, the cluster center obtained using the mean-shift clustering algorithm will be used as the threshold; The loss value is calculated based on each feature and its corresponding threshold, and the threshold corresponding to the minimum loss value is taken as the local optimal split point of the feature.

16. A processor-readable storage medium, characterized in that, The processor-readable storage medium stores a computer program for causing the processor to perform the method according to any one of claims 1 to 7.

Images