Method, device and application of indoor electromagnetic propagation prediction model
By preprocessing and optimizing training based on perceptron networks, a high-precision indoor electromagnetic propagation prediction model was established, which solved the problem of loss of scene features and feature information in existing models and achieved high-precision prediction applicable across different scenarios.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- XIDIAN UNIV
- Filing Date
- 2022-10-24
- Publication Date
- 2026-04-10
AI Technical Summary
Existing indoor electromagnetic propagation prediction models lose information such as scene features, frequency features, and antenna features when considering the spatial coordinates of the receiving point, resulting in poor prediction performance.
By constructing an indoor electromagnetic propagation prediction model based on a perceptron network, a training dataset is established using simulation data from a ray tracing model. Data preprocessing is performed, such as removing bad samples, amplifying input features, and normalization. A four-layer perceptron network structure is used for training, and the connection weights are optimized using the Sigmoid function and regularization terms.
It improves the accuracy of indoor electromagnetic propagation prediction, solves the problem of cross-scenario applicability, and achieves high-precision electromagnetic wave propagation prediction.
Smart Images
Figure CN116029195B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application belongs to the technical field of constructing an electromagnetic propagation prediction model, and particularly relates to a method, system, device and terminal for constructing an indoor electromagnetic propagation prediction model. BACKGROUND
[0002] At present, the indoor electromagnetic propagation prediction model mainly includes a deterministic model and a model based on data statistics. The deterministic model usually has the problems of large calculation amount and low efficiency. The traditional statistical model based on data construction is limited to a specific scene, and cannot be used in another scene. The application establishes a new indoor electromagnetic propagation prediction model based on measured data and a perception machine network.
[0003] The perception machine network is an artificial neural network that simulates the calculation and reproduction of the human brain neural system in some simplified sense. The artificial neural network can be divided into two categories of feedforward network (also known as multilayer perception machine network) and feedback network according to its model structure. The former can be regarded as a large-scale nonlinear mapping system in mathematics, and the latter is a large-scale nonlinear dynamic system. According to the learning method, the artificial neural network can be divided into three categories of supervised learning, unsupervised and semi-supervised learning. According to the working method, the artificial neural network can be divided into two categories of certainty and randomness. According to the time characteristics, the artificial neural network can be divided into two categories of continuous type and discrete type, etc.
[0004] A typical perception machine network has the following three parts:
[0005] Architecture: The architecture specifies the variables in the network and their topological relationship. For example, the variables in the perception machine network can be the weights of neuron connections and the activities of neurons.
[0006] Activation Rule: Most perception machine network models have a short-time-scale dynamic rule to define how a neuron changes its activity according to the activities of other neurons. Generally, the activation function depends on the weights in the network (i.e., the parameters of the network).
[0007] Learning Rule: The learning rule specifies how the weights in the network are adjusted over time. This is generally considered as a long-time-scale dynamic rule. Generally, the learning rule depends on the activities of neurons. It can also depend on the target values provided by the supervisor and the current weight values.
[0008] However, for the wave prediction model established by the perception machine network, only considering the three-dimensional input data of the spatial coordinates of the receiving point loses a lot of information, and the perception machine network may not achieve good training effect (underfitting), so a new indoor electromagnetic propagation prediction model construction method needs to be designed.
[0009] Through the above analysis, the problems and defects of the prior art are that: for the traditional wave prediction model established based on the neural network, only considering the three-dimensional input data of the spatial coordinates of the receiving point loses a lot of information, for example, the scene features, frequency features, antenna features and the like are not used, so that the calculation effect cannot be good. SUMMARY
[0010] In view of the problems existing in the prior art, the present application provides a construction method, system, device and terminal of an indoor electromagnetic propagation prediction model, and particularly relates to a method, system, medium, device and terminal for establishing an indoor electromagnetic propagation prediction model based on a perception machine network.
[0011] The present application is implemented in the following manner: a construction method of an indoor electromagnetic propagation prediction model, the construction method of the indoor electromagnetic propagation prediction model comprising: constructing a receiving power data set for an indoor environment, and preprocessing the receiving power data set, including normalization, translation and the like; training a perception machine network by using the preprocessed receiving power data set, finding the connection weight corresponding to a smaller cost function through an optimization algorithm, and taking the obtained connection weight as the connection parameter value of the perception machine network for prediction simulation comparison and verification.
[0012] Further, the construction method of the indoor electromagnetic propagation prediction model comprises the following steps:
[0013] Step one: establishing an original training data set; this step is the basis of the whole process, and the data set limits the upper limit of the model, so it is necessary to establish a good data set.
[0014] Step two: preprocessing the original training data set; this step is to reduce some invalid data, for example, some extreme values affecting the output, and normalization is also a necessary process for solving.
[0015] Step three: selecting the perception machine network structure and the number of neurons according to the sample features and the number; this step is to adjust the optimal perception machine network structure.
[0016] Step four: bringing the training set into the perception machine network for training, outputting the final connection weight of the perception machine network, and detecting and verifying the perception machine network. This step can obtain the accuracy of the model and the simulation result.
[0017] Further, the training data set in step one is obtained by ray tracing model simulation, ray tracing model in the deterministic radio wave prediction model is used for simulation, and the original training data set is obtained by using simulation data.
[0018] Further, the training data pre-processing in step two includes:
[0019] (1) Eliminate bad sample information: there is a value of -inf or far less than the actual signal receiver minimum receiving lower limit in the ray tracing simulation data, the power value of the sample point is set to a smaller value;
[0020] (2) Amplify sample input features: increase the high-order items of the original input as new input features into the training set, the ray tracing model original output result is a group of receiving antenna position corresponding to a receiving power; The mapping relationship is as follows:
[0021]
[0022] The mapping relationship after adding high-order items to the data is as follows:
[0023]
[0024] Wherein, x, y, z represent the coordinates of the space receiving point, p represents the highest order number of the input feature, and Power represents the ideal receiving power value of the receiving point;
[0025] (3) Normalization processing: when p is large enough, the input feature value increases exponentially, and the input feature is normalized to make the input feature amplified.
[0026] Further, the perception machine network structure in step three is: a four-layer perception machine network structure of (3*p) x 20 x 20 x 1; Wherein, 3 in the input layer represents the number of original input parameter features, and p represents the number of high-order items added; The number of neurons in the hidden layer is 20; The output layer has only one neuron to calculate the final power prediction result.
[0027] Further, the construction method of the indoor electromagnetic propagation prediction model further includes:
[0028] According to the sample data and sample features of the data set, a suitable perception machine network is established, the number of input layer neurons is consistent with the number of sample input features, the number of neurons in the hidden layer is determined according to the experience value, and the output layer only requires one neuron to output the final power result.
[0029] The neuron node activation function is mainly Sigmod function, and the expression is as follows:
[0030]
[0031] Wherein, z is the output value of neuron, which is transmitted to the next layer of neurons after experiencing the activation function; the forward propagation cost function represents the error between the prediction result of the perceptron network and the ideal result, and the mean square deviation of the observed value and the true value is used as the cost function, and the expression is:
[0032]
[0033] Wherein, Θ represents the connection weight of each neuron of the perceptron network, m represents the number of training samples, y and y' represent the ideal output and the actual output of the perceptron network respectively; λ represents the regularization term, which is used to prevent the perceptron network from overfitting.
[0034] Another object of the present application is to provide a construction system of an indoor electromagnetic propagation prediction model, which applies the construction method of the indoor electromagnetic propagation prediction model.
[0035] The data set establishment module is configured to establish an original training data set.
[0036] The data preprocessing module is configured to preprocess the original training data set.
[0037] The perceptron network determination module is configured to select the structure of the perceptron network and the number of neurons according to the sample characteristics and the number.
[0038] The network training module is configured to bring the training set into the perceptron network for training, output the final connection weight of the perceptron network, and detect and verify the perceptron network.
[0039] Another object of the present application is to provide a computer device, which comprises a memory and a processor, the memory stores a computer program, and the computer program is executed by the processor to make the processor execute the steps of the construction method of the indoor electromagnetic propagation prediction model.
[0040] Another object of the present application is to provide a computer readable storage medium, which stores a computer program, and the computer program is executed by the processor to make the processor execute the steps of the construction method of the indoor electromagnetic propagation prediction model.
[0041] Another object of the present application is to provide an information data processing terminal, which is used to realize the construction system of the indoor electromagnetic propagation prediction model.
[0042] In combination with the above technical solutions and the technical problems solved, the technical solution to be protected by the present application has the following advantages and positive effects:
[0043] First, in view of the technical problems existing in the prior art and the difficulty of solving the problems, the technical scheme to be protected by the application and the results and data in the research and development process are combined closely, and the technical problems solved by the technical scheme of the application and some creative technical effects brought after the problems are solved are analyzed in detail and profoundly. The specific description is as follows:
[0044] The perception machine network is an artificial neural network which imitates the calculation of the human brain neural system in some simplified sense. The application uses a multi-layer perception machine network to realize power coverage prediction of an indoor environment, so as to establish a radio wave propagation prediction model. Considering the artificial cost of obtaining indoor measured data, the application uses a large amount of data simulated by a ray tracing model to establish an initial training data set. After a series of data preprocessing operations such as removing, amplifying sample features and normalizing, the data set is put into the perception machine network for training and prediction. After inspection, the perception machine network after final training performs well on the cross-validation set and the test set, and can be used to establish the radio wave propagation prediction model.
[0045] In order to reasonably establish a high-precision indoor radio wave prediction model, the application provides a method for establishing an indoor electromagnetic propagation prediction model based on a perception machine network. The perception machine network is trained by using the preprocessed data in the application, and the connection weight value corresponding to a smaller cost function is found through some optimization algorithm. At this time, the obtained connection weight value is used as the connection parameter value of the perception machine network for prediction simulation comparison and verification. The feasibility of the method for establishing the indoor radio wave prediction model by the perception machine network is verified in the embodiments.
[0046] Secondly, from the perspective of the product or as a whole, the technical effects and advantages of the technical scheme to be protected by the application are described as follows:
[0047] According to the comparison simulation output prediction power result data set of the ray tracing model, the data set is processed by feature amplification, so that the perception machine network can better fit the data. The application uses the perception machine network to establish an indoor radio wave prediction model with high precision, and the prediction accuracy is high.
[0048] Thirdly, as the creative auxiliary evidence of the claims of the application, it is also reflected in the following important aspects:
[0049] The technical scheme of the application solves the technical problem that people have been eager to solve but have failed to succeed: the technical scheme of the application solves the problem that the wireless channel prediction model based on the neural network cannot be used across scenes, and the prediction accuracy is improved due to the extraction of frequency features, material features and scene features. BRIEF DESCRIPTION OF DRAWINGS
[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments of the present application will be briefly introduced as follows. Obviously, the drawings described below only represent some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without any creative effort on the basis of these drawings.
[0051] Figure 1 is a flow chart of the method for constructing the indoor electromagnetic propagation prediction model provided by the embodiments of the present application;
[0052] Figure 2 is a main flow chart of the method for constructing the indoor electromagnetic propagation prediction model provided by the embodiments of the present application;
[0053] Figure 3 is a schematic diagram of feature selection provided by the embodiments of the present application;
[0054] Figure 4 is a schematic diagram of the structure of the perception machine network provided by the embodiments of the present application;
[0055] Figure 5 is a result diagram of the perception machine network prediction fitting data provided by the embodiments of the present application;
[0056] Figure 6 is a prediction coverage power diagram of the ray tracing model provided by the embodiments of the present application;
[0057] Figure 7 is a prediction coverage power diagram of the perception machine network provided by the embodiments of the present application. DETAILED DESCRIPTION
[0058] In order to make the objects, technical solutions and advantages of the present application more clear, the present application will be further described in detail below with reference to the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not used to limit the present application.
[0059] In order to solve the problems in the prior art, the present application provides a method, system, device and terminal for constructing an indoor electromagnetic propagation prediction model, which will be described in detail below with reference to the drawings.
[0060] I. Explanation of the embodiments. In order to enable those skilled in the art to fully understand how the present application is specifically implemented, this part is an explanation of the embodiments of the technical solutions of the claims.
[0061] As shown in Figure 1 , the method for constructing the indoor electromagnetic propagation prediction model provided by the embodiments of the present application comprises the following steps:
[0062] S101, establishing an original training data set;
[0063] S102, preprocessing the original training data set;
[0064] S103, selecting the perceptron network structure and the number of neurons according to the sample characteristics and the number;
[0065] S104, bringing the training set into the perceptron network for training, outputting the final connection weight of the perceptron network, and detecting and verifying the perceptron network.
[0066] As a preferred embodiment, as shown in the figure, the method for constructing the indoor electromagnetic propagation prediction model provided by the embodiment of the application specifically comprises the following steps: Figure 2
[0067] (1) establishing a suitable training data set, the original purpose of the radio wave propagation prediction model is that after the positions of the transmitting and receiving antennas and the corresponding predicted environment are given, the model can output the predicted power or path loss, etc., when the perceptron network is used to establish the radio wave propagation prediction model, the input and output structures of the training data should be fully considered, in the simplest case, the input data is the spatial coordinates of the receiving point, and the output data can be the predicted power value of the point, so that the training data with the structure is used to train the output of the perceptron network;
[0068] (2) preprocessing the training data, removing bad sample points therefrom and amplifying the sample characteristics of the input variables, and finally normalizing the input of the training set, so as to increase the fitting degree of the perceptron network;
[0069] (3) selecting the perceptron network structure and the number of neurons according to the sample characteristics and the number;
[0070] (4) bringing the training set into the perceptron network for training, outputting the final connection weight of the perceptron network, and detecting and verifying the perceptron network.
[0071] The acquisition mode of the original data set in step (1) provided by the embodiment of the application is mainly obtained by ray tracing model simulation, the ray tracing model in the deterministic radio wave prediction model is used for simulation, and the original training data set is established by using the simulation data.
[0072] The training data preprocessing step in step (2) provided by the embodiment of the application is:
[0073] Removing bad sample information: there are some power values of -inf or values far smaller than the minimum receiving lower limit of the actual signal receiver in the ray tracing simulation data, and the power values of these sample points are set to a smaller value;
[0074] Amplification sample input features: namely, increase the high order term of the original input as a new input feature into the training set, such as the ray tracing model original output result is a group of receiving antenna position corresponding a received power; The mapping relationship is as follows:
[0075]
[0076] The mapping relationship after adding high order term to data is as follows:
[0077]
[0078] Wherein x, y, z represent the coordinates of the space receiving point, p represents the highest order term number of the input feature, and power represents the ideal received power value of the receiving point;
[0079] Normalization processing: when p is large enough, the input feature value will increase exponentially, and here it needs to be normalized to make the input feature amplified readable.
[0080] The perception machine network structure in step (3) provided by the embodiment of the application is a four-layer perception machine network structure of (3*p) x 20 x 20 x 1; wherein 3 in the input layer represents the number of original input parameter features, and p represents the number of added high order terms; the number of neurons in the hidden layer is 20; and the output layer has only one neuron to calculate the final power prediction result.
[0081] The embodiment of the application provides a received power data set preprocessing method for an indoor environment. For the electric wave prediction model established by the perception machine network, only considering the three-dimensional input data of the space coordinates of the receiving point loses a lot of information, and the perception machine network may not achieve good training effect (underfitting). Therefore, on the basis of the existing input data, the high order term processing can be carried out, the input features of the training data are amplified, so that the perception machine network can better fit the data to achieve the effect of the electric wave propagation model prediction.
[0082] The method for establishing an indoor electric wave prediction model by using the perception machine network provided by the embodiment of the application comprises the following steps: establishing a suitable perception machine network according to the sample data and sample features of the data set; generally, the number of neurons in the input layer is consistent with the number of sample input features, the number of neurons in the hidden layer is determined according to the experience value, and the output layer only requires one neuron to output the final power result.
[0083] The neuron node activation function is mainly the Sigmod function, and the expression is as follows:
[0084]
[0085] Wherein, z is the output value of neuron, which is transmitted to the next layer of neuron after experiencing the activation function. The forward propagation cost function represents the error between the prediction result of the perceptron network and the ideal result, and the square mean deviation between the observation value and the true value is used as the cost function, and the expression is:
[0086]
[0087] Wherein, Θ represents the connection weight of each neuron of the perceptron network, m represents the number of training samples, y and y' represent the ideal output and the actual output of the perceptron network respectively, and λ represents the regularization term, which is used to prevent the perceptron network from overfitting.
[0088] The system for constructing the indoor electromagnetic propagation prediction model provided by the embodiment of the application comprises:
[0089] A data set establishing module is configured to establish an original training data set.
[0090] A data preprocessing module is configured to preprocess the original training data set.
[0091] A perceptron network determining module is configured to select a perceptron network structure and a number of neurons according to sample features and quantities.
[0092] A network training module is configured to bring the training set into the perceptron network for training, output final connection weights of the perceptron network, and detect and verify the perceptron network.
[0093] II. Application Embodiment. In order to prove the creativity and technical value of the technical scheme of the application, this part is an application embodiment of the technical scheme of the claim on a specific product or related technology.
[0094] When deploying an indoor wireless signal source, the selection of the deployment position is time-consuming and labor-consuming by using a deterministic model, and needs to measure data by using a traditional statistical model, and only a small amount of deterministic model simulation is needed by using the present application, and the signal energy distribution can be quickly obtained by training, so that the deployment position and the transmission power are optimized.
[0095] III. Evidence of the effects of the embodiments. The embodiments of the application have achieved some positive effects in the research and development or use process, and indeed have great advantages compared with the prior art. The following content is described in combination with the data and graphs of the test process.
[0096] The embodiment of the application mainly realizes the signal power prediction in the indoor space by using the perceptron network, obtains an original data set by using a ray tracing model with high precision in the industry, and then uses the original data set as the training of the perceptron network, so that the trained perceptron network model achieves the effect of indoor signal prediction, such as Figure 2The main flow chart of establishing a prediction model by using a perception machine network is shown.
[0097] The original data set preprocessing scheme for the indoor ray tracing model simulation in the embodiment of the present application includes the following steps:
[0098] 1) Establishment of the data set
[0099] The prediction power result of the ray tracing simulation is used as the main data set source to preliminarily verify the neural network model in the embodiment of the present application.
[0100] The power data obtained by the surface simulation of the indoor environment of an apartment in Dujiangyan is used as the initial data set, which contains 5444 data points and includes the position of the information receiving antenna and the predicted power. The data needs to be preprocessed before being used for the neural network, such as eliminating bad samples, increasing the sample feature dimension, etc. Here, the value of the received power in the sample is replaced (replaced by a smaller value), and then 60% of the 5444 samples obtained by simulation are randomly selected as training samples, 20% as cross-validation samples, and another 20% as test samples. The feature selection schematic diagram provided by the embodiment of the present application is shown in Figure 3 .
[0101] Secondly, considering that the data set input only contains the position information of the receiving point, the neural input layer dimension is low and the data features are not obvious, the original input data is processed by increasing the high-order term, such as:
[0102]
[0103] For example, the original input is a coordinate point (1, 2, 3), and the processed input is (1, 1, 1, 2, 4, 8, 3, 9, 27). When the order is too large, the data needs to be normalized to prevent data overflow.
[0104] 2) Specific construction of the neural network
[0105] The neuron node activation function is mainly the sigmod function, and the expression is as follows:
[0106]
[0107] The forward propagation cost function represents the error between the prediction result of the neural network and the ideal result. Here, the square mean deviation of the observed value and the true value is used as the cost function, and the expression is:
[0108]
[0109] For example, Figure 4As shown, in the perceptron network, except for the output layer neuron Y' which directly outputs the sum of the previous layer (i.e., the last layer neuron is directly activated without the restriction of an activation function), the outputs of other neurons are the Sigmoid function values of the sum of the previous layer. Ignoring paranoid neurons, the neural network is a four-layer structure (3*p)×20×20×1, where 3 in the input layer represents the number of original input parameter features, and p represents the number of added higher-order terms; the hidden layers have two layers, each with 20 neurons; and the output layer has only one neuron that calculates the final power prediction result.
[0110] 3) Fitting effect of neural network
[0111] Table 1 shows the final performance of the neural network on the dataset after training. The dataset has a total of 5444 points, of which 60% are used to train the neural network, 20% are used as a cross-validation set to examine the model's fitting effect, and the remaining 20% are used as a test set to examine the error of the final model.
[0112] Table 1 Training Status
[0113]
[0114] like Figure 5 As shown, the embodiment of the present invention utilizes the original data from RT simulation and the final neural network prediction fitting data. For this sample, the standard deviation between the neural network prediction result and the original data result is 3.9516 dBm, and the average error is -0.025972 dBm.
[0115] The RT simulation prediction coverage power provided in this embodiment of the invention is as follows: Figure 6 As shown, the RT simulation coverage provided in this embodiment of the invention Figure Four The layer neural network fitting results predict the coverage power as follows: Figure 7 As shown.
[0116] It should be noted that embodiments of the present application can be realized by hardware, software, or a combination of software and hardware. The hardware portion can be realized by a special logic; the software portion can be stored in a memory and executed by a proper instruction execution system, such as a microprocessor or a specially designed hardware. A person of ordinary skill in the art can understand that the above-mentioned apparatus and method can be realized by computer executable instructions and / or included in processor control codes, such as a carrier medium, such as a magnetic disk, CD or DVD-ROM, a programmable memory, such as a read-only memory (firmware), or a data carrier, such as an optical or electronic signal carrier. The apparatus of the present application and its modules can be realized by a hardware circuit, such as a very large scale integrated circuit or a gate array, a semiconductor, such as a logic chip, a transistor, or a programmable hardware device, such as a field programmable gate array, a programmable logic device, or the like, by software executed by various types of processors, or by a combination of the above-mentioned hardware circuit and software, such as firmware.
[0117] The above description is merely a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any modification, equivalent replacement, and improvement within the technical range disclosed by the present application, and within the spirit and principle of the present application, should be covered within the protection scope of the present application.
Claims
1. A method for constructing an indoor electromagnetic propagation prediction model, characterized in that, The method for constructing the indoor electromagnetic propagation prediction model comprises the following steps: The method for constructing the indoor electromagnetic propagation prediction model comprises the following steps: Step one, establishing an original training data set; Step two, preprocessing the original training data set; Step three, selecting a perceptron network structure and a number of neurons according to sample features and quantities; Step four, inputting the training set into the perceptron network for training, outputting final connection weights of the perceptron network, and detecting and verifying the perceptron network; The training data set in the step one is obtained by simulation of a ray tracing model, the ray tracing model in a deterministic radio wave prediction model is used for simulation, and the original training data set is obtained by using simulation data; The training data preprocessing in the step two comprises: (1) eliminating bad sample information: the power value of a sample point is set to a smaller value, because the power is-inf or lower than the minimum receiving lower limit of an actual signal receiver in the ray tracing simulation data; (2) amplifying sample input features: increasing high-order items of original inputs as new input features into the training set, the ray tracing model outputs a group of receiving antenna positions corresponding to a received power; the mapping relationship is as follows: ； After adding high-order items, the mapping relationship is as follows: ； Wherein, x, y and z represent coordinates of a space receiving point, p represents a highest order item number of input features, and Power represents an ideal received power value of the receiving point; (3) normalization processing: with the increase of p, the input feature value increases exponentially, and the input feature is normalized to make the input feature read after amplification.
2. The method of claim 1, wherein, The perception machine network structure in the step three is: The four-layer perception machine network structure of the formula (1) ; wherein 3 in the input layer represents the original input parameter characteristic number, p represents the number of increased high-order terms; the number of neurons in the hidden layer is 20; the output layer has only one neuron to calculate the final power prediction result.
3. The method of claim 1, wherein the electromagnetic propagation model is constructed by: The method for constructing the indoor electromagnetic propagation prediction model further comprises: A suitable perceptron network is established according to sample data and sample features of the data set, the number of input layer neurons is consistent with the number of sample input features, the number of neurons in the hidden layer is determined according to an experience value, and the output layer only requires one neuron to output a final power result; The neuron node activation function is mainly a Sigmod function, and the expression is as follows: ； Wherein, z is an output value of a neuron, which is transmitted to a neuron of a next layer after experiencing the activation function; the forward propagation cost function represents an error between a prediction result of the perceptron network and an ideal result, and the square mean of an observation value and a true value deviation is used as the cost function, and the expression is as follows: ； wherein, represents the connection weight of each neuron in the perceptron network, m represents the number of training samples, y and y' represent the ideal output and the actual output of the perceptron network, respectively; represents a regularization term, which is used to prevent overfitting of the perceptron network.
4. A system for constructing an indoor electromagnetic propagation prediction model using the method for constructing an indoor electromagnetic propagation prediction model according to any one of claims 1 to 3, characterized in that, The system for constructing the indoor electromagnetic propagation prediction model comprises: A data set establishing module for establishing an original training data set; A data preprocessing module for preprocessing the original training data set; A perceptron network determining module for selecting a perceptron network structure and a number of neurons according to sample features and quantities; The network training module is used for bringing the training set into the perception machine network for training, outputting the final connection weight of the perception machine network, and detecting and verifying the perception machine network.
5. A computer device, comprising: The computer device comprises a memory and a processor, the memory stores a computer program, and the computer program is executed by the processor to make the processor execute the steps of the construction method of the indoor electromagnetic propagation prediction model in any one of claims 1-3. 6.A computer readable storage medium storing a computer program, the computer program being executed by a processor to make the processor execute the steps of the construction method of the indoor electromagnetic propagation prediction model in any one of claims 1-3.
7. An information data processing terminal, characterized by The information data processing terminal is used to implement the construction system of the indoor electromagnetic propagation prediction model in claim 4.