CN116470975A - Modeling method of SVR-based adaptive propagation prediction model - Google Patents
Modeling method of SVR-based adaptive propagation prediction model Download PDFInfo
- Publication number
- CN116470975A CN116470975A CN202310361214.0A CN202310361214A CN116470975A CN 116470975 A CN116470975 A CN 116470975A CN 202310361214 A CN202310361214 A CN 202310361214A CN 116470975 A CN116470975 A CN 116470975A
- Authority
- CN
- China
- Prior art keywords
- model
- svr
- data
- parameters
- super
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 26
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 15
- 230000005540 biological transmission Effects 0.000 claims abstract description 24
- 238000012360 testing method Methods 0.000 claims abstract description 20
- 238000012549 training Methods 0.000 claims abstract description 16
- 230000006870 function Effects 0.000 claims abstract description 14
- 238000012795 verification Methods 0.000 claims abstract description 14
- 238000012545 processing Methods 0.000 claims abstract description 12
- 238000010606 normalization Methods 0.000 claims abstract description 8
- 238000005457 optimization Methods 0.000 claims description 7
- 238000004458 analytical method Methods 0.000 claims description 6
- 238000005259 measurement Methods 0.000 claims description 3
- 238000011161 development Methods 0.000 abstract description 2
- 238000010801 machine learning Methods 0.000 abstract description 2
- 238000004364 calculation method Methods 0.000 description 5
- 238000004891 communication Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013524 data verification Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000005684 electric field Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000013179 statistical model Methods 0.000 description 1
- 238000012876 topography Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/391—Modelling the propagation channel
- H04B17/3913—Predictive models, e.g. based on neural network models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/27—Regression, e.g. linear or logistic regression
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Software Systems (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Electromagnetism (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Monitoring And Testing Of Transmission In General (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a modeling method of an adaptive propagation prediction model based on SVR, which comprises the following steps: reading characteristic data and expected output data required by modeling and carrying out normalization processing; selecting a verification strategy of 6-fold cross verification based on the SVR model, and randomly dividing data into 6 mutually exclusive subsets in a data dividing stage; randomly selecting a union set of 5 subsets each time as a training set, taking the remaining subset as a test set, and performing 6 times of training and testing; selecting an optimal kernel function and super parameters, and establishing an SVR model; and testing the model, outputting a transmission loss regression predicted value, comparing with the actual transmission loss, and evaluating the model. The SVR-based adaptive propagation prediction model provides ideas for the improvement of the propagation prediction precision of the frequency modulation broadcast signal wave, the combination and development of a prediction method and machine learning and the like.
Description
Technical Field
The invention relates to the technical field of wireless communication, in particular to a modeling method of a self-adaptive propagation prediction model based on SVR.
Background
The basis of the wireless communication system is the research of the propagation characteristics of a wireless channel and the establishment of a channel model. In the radio wave propagation prediction, propagation prediction methods can be classified into three types.
The method is characterized in that the method utilizes the accumulated data and statistical curve of the standardized propagation characteristic test, interpolates or extrapolates according to the equipment frequency parameter, finally obtains the propagation prediction result of a certain area through a series of corrections of topography, environmental characteristics and the like, and generally provides a chart or fitting formula.
And secondly, a deterministic prediction method is obtained by a method for carrying out strict electromagnetic calculation according to an electric wave propagation fluctuation theory, and the method has accurate prediction precision, but complex calculation and needs detailed environmental parameters.
And thirdly, a semi-empirical semi-deterministic prediction method combines the first two methods, replaces partial complex electromagnetic calculation with a statistical result, and improves the operation speed within the range of acceptable calculation accuracy.
Currently available channel propagation models are mostly built based on specific foreign areas or typical features, and the application range is limited. Therefore, on the basis of the existing propagation prediction model, it is urgently required to establish a propagation prediction, coverage characteristic analysis and network planning method suitable for specific areas in China, or put forward a practical rule for specific applications.
Disclosure of Invention
The invention aims at solving the technical defects existing in the prior art and provides a modeling method of an adaptive propagation prediction model based on SVR.
The technical scheme adopted for realizing the purpose of the invention is as follows:
a modeling method of an adaptive propagation prediction model based on SVR comprises the following steps:
step A: reading characteristic data and expected output data required by modeling and carrying out normalization processing on a characteristic data set;
and (B) step (B): selecting a verification strategy of 6-fold cross verification based on the SVR model, and randomly dividing data into 6 mutually exclusive subsets in a data dividing stage; randomly selecting a union set of 5 subsets each time as a training set, taking the remaining subset as a test set, and performing 6 times of training and testing;
step C: selecting an optimal kernel function and super parameters, and establishing an SVR model;
step D: and testing the model, outputting a transmission loss regression predicted value, comparing with the actual transmission loss, and evaluating the model.
In the step A, the characteristic data set comprises receiving and transmitting distance, longitude and latitude of a receiving point, terrain roughness and frequency characteristic information; the expected output data is the transmission loss data of the wireless channel actually measured from the transmitting position to the receiving position;
after the data is read, the normalization processing is performed on all the feature data to enable the feature data with different dimensions to have the same measurement scale, so that model operation is convenient, and the feature data is enabled to be between 0 and 1.
In the step C, the kernel function is selected to be widely applied and can be mapped to a Gaussian kernel function with infinite dimension; and performing traversal optimization on the key super parameters C and G of the model, and determining the model based on the traversal optimization results of the key super parameters C and G.
Wherein, when traversing and optimizing the key super parameters C and G of the model, the search ranges of C and G are respectively 2 -5 -2 15 And 2 -15 -2 10 Taking Root Mean Square Error (RMSE) predicted by SVR model as an objective function; substituting different combinations of the above super-parameters into the SVR model, predicting the transmission loss of the channel, calculating the root mean square error, and finally searching the key super-parameters C and G with minimum root mean square error.
The specific steps of traversing and optimizing the key super parameters C and G of the model are as follows:
c21: selecting a first set of C and G hyper-parameters;
c22: constructing an SVR model according to the current C and G super parameters;
c23: recording the RMSE predicted by the SVR model training set under the current C and G super parameters;
c24: judging whether the traversal is finished;
if not, then
And C25: updating the combination of the C and G super parameters;
repeating steps C22-C25 until all combinations of the C and G super parameters are traversed;
c26: finding the RMSE predicted by the minimum SVR model training set;
c27: outputting the C and G super parameter combination corresponding to the RMSE predicted by the SVR model training set with the minimum.
The step D comprises the steps of model test and model verification;
during model test, a characteristic data set of a test set is put into a model to carry out regression prediction of transmission loss;
and during model verification, comparing the transmission loss regression predicted value with actual wireless channel transmission loss data, and performing processing analysis on errors, correlation coefficients and prediction distribution.
The SVR-based adaptive propagation prediction model established by the invention provides ideas for the improvement of the propagation prediction precision of the frequency modulation broadcast signal electric wave, the combination and development of a prediction method and machine learning and the like, and lays a foundation for the study of the system planning, the frequency band management and the like of the frequency modulation broadcast signal.
Drawings
FIG. 1 is a flow chart of a method for modeling an adaptive propagation prediction model based on SVR according to an embodiment of the present invention.
FIG. 2 is a flow chart of selecting optimal kernel functions and hyper-parameters to build SVR models in accordance with an embodiment of the invention.
Detailed Description
The invention is described in further detail below with reference to the drawings and the specific examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
The invention relates to a large-scale channel modeling method, wherein a large-scale channel model is used for predicting the change of electric field intensity at a receiving end and is the average statistics of a channel in a quasi-static area range.
Under the condition of measuring data of different time and place percentages, a calculation model is induced after analysis by a mathematical statistical method, and the model is called a regional statistical model and has the characteristics of simple modeling method and good applicability to a modeling region.
In order to improve the prediction precision and robustness of regional frequency modulation broadcast signal electric wave propagation, the inventor establishes a self-adaptive propagation prediction model suitable for multiple frequency bands and multiple scenes based on SVR, determines super parameters of an SVR algorithm through traversal optimization, expands a data set by cross verification, improves model generalization capability, plays an important role in improving the accuracy of coverage characteristic analysis of a broadcast television system, and lays a foundation for research in aspects of system planning, frequency management and the like of the broadcast television system.
As shown in fig. 1, modeling based on an SVR adaptive propagation prediction model according to an embodiment of the present invention is implemented by the following technical scheme:
step A: data reading and processing
And reading the characteristic data and expected output data and carrying out normalization processing on the characteristic data.
And (B) step (B): data partitioning
A verification strategy of 6-fold cross verification is selected based on the SVR model of the invention. The data is then divided into 6 mutually exclusive subsets of similar size during the data division phase. The union of 5 subsets is randomly selected as the training set each time, and the remaining subset is the test set, namely 6 times of training and testing are performed.
Step C: modeling
And selecting the optimal kernel function and super parameters, and finally establishing an SVR model.
Step D: and (5) evaluating a model.
And outputting a transmission loss regression predicted value, and comparing the transmission loss regression predicted value with the actual transmission loss, so that the follow-up data verification and analysis are convenient.
Further, the step a of the embodiment of the present invention specifically includes:
a1: reading data
In the data reading link, the data is divided into two main categories, namely a characteristic data set which consists of characteristic information such as receiving and transmitting distance, longitude and latitude of a receiving point, terrain roughness, frequency and the like; and secondly, the corresponding expected output data, namely the transmission loss data of the wireless channel actually measured from the transmitting position to the receiving position.
A2: normalization processing
After the data is read, in order to enable the feature data with different dimensions to have the same measurement scale, the model operation is convenient, normalization processing is required to be carried out on all the feature data, and the feature data is between 0 and 1.
Further, step C of the embodiment of the present invention specifically includes:
c1: model determination
The SVR model is selected, and the kernel function is widely applied and can be mapped to Gaussian kernel functions with infinite dimensions in view of nonlinear correlation of modeling characteristic data.
C2: super parameter optimizing
In order to ensure the precision of the model establishment based on SVR, the key super parameters C and G of the model are subjected to traversal optimization technology based on the established training set, and the search range of C and G is respectively 2 -5 -2 15 And 2 -15 -2 10 Taking Root Mean Square Error (RMSE) of SVR model prediction as an objective function; substituting different combinations of the above super-parameters into an SVR model, predicting the transmission loss of a channel, calculating the root mean square error, and finally searching the C and G super-parameters with the minimum root mean square error; see fig. 2.
Further, step C2 of the embodiment of the present invention specifically includes, as shown in fig. 2:
c21: selecting a first set of C and G hyper-parameters;
c22: constructing an SVR model according to the current C and G super parameters;
c23: recording the RMSE predicted by the SVR model training set under the current super-parameter combination;
c24: judging whether the traversal is finished;
if not, then
And C25: updating the combination of the C and G super parameters;
repeating steps C22-C25; until all combinations of C and G super parameters are traversed.
C26: find the smallest RMSE;
c27: outputting the C and G super parameter combination corresponding to the minimum RMSE.
And C3: model determination
Updating the C and G super parameters of the SVR model to the optimal combined super parameters found in the step C2, and determining the SVR model.
Further, step D of the embodiment of the present invention specifically includes:
d1: model testing
And putting the characteristic data of the test set into a model, and carrying out regression prediction of the transmission loss.
D2: model verification
And comparing the transmission loss regression predicted value with actual wireless channel transmission loss data, and performing processing analysis such as error, correlation coefficient, prediction distribution and the like.
While the fundamental and principal features of the invention and advantages of the invention have been shown and described, it will be apparent to those skilled in the art that the invention is not limited to the details of the foregoing exemplary embodiments, but may be embodied in other specific forms without departing from the spirit or essential characteristics thereof;
the present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.
Furthermore, it should be understood that although the present disclosure describes embodiments, not every embodiment is provided with a separate embodiment, and that this description is provided for clarity only, and that the disclosure is not limited to the embodiments described in detail below, and that the embodiments described in the examples may be combined as appropriate to form other embodiments that will be apparent to those skilled in the art.
Claims (6)
1. The modeling method of the SVR-based adaptive propagation prediction model is characterized by comprising the following steps:
step A: reading the characteristic data and expected output data required by modeling and carrying out normalization processing on the characteristic data;
and (B) step (B): selecting a verification strategy of 6-fold cross verification based on the SVR model, and randomly dividing data into 6 mutually exclusive subsets in a data dividing stage; randomly selecting a union set of 5 subsets each time as a training set, taking the remaining subset as a test set, and performing 6 times of training and testing;
step C: selecting an optimal kernel function and super parameters, and establishing an SVR model;
step D: and testing the model, outputting a transmission loss regression predicted value, comparing with the actual transmission loss, and evaluating the model.
2. The method for modeling an adaptive propagation prediction model based on SVR according to claim 1, wherein in step a, the feature data set includes a transceiving distance, a longitude and latitude of a receiving point, a terrain roughness, and frequency feature information; the expected output data is the transmission loss data of the wireless channel actually measured from the transmitting position to the receiving position;
after the data is read, the normalization processing is performed on all the feature data to enable the feature data with different dimensions to have the same measurement scale, so that model operation is convenient, and the feature data is enabled to be between 0 and 1.
3. A method of modeling an adaptive propagation prediction model based on SVR as claimed in claim 3 wherein in step C, the kernel function is selected from a broad range of gaussian kernel functions that can be mapped to infinite dimensions; and performing traversal optimization on the key super parameters C and G of the model, and determining the model based on the traversal optimization results of the key super parameters C and G.
4. The modeling method of SVR-based adaptive propagation prediction model according to claim 3, wherein the search ranges of C and G are 2 when the key super parameters C and G of the model are subjected to traversal optimization -5 -2 15 And 2 -15 -2 10 Taking Root Mean Square Error (RMSE) predicted by SVR model as an objective function; substituting different combinations of the above super-parameters into the SVR model, predicting the transmission loss of the channel, calculating the root mean square error, and finally searching the key super-parameters C and G with minimum root mean square error.
5. The modeling method of an adaptive propagation prediction model based on SVR according to claim 4, wherein the specific steps of the key hyper-parameters C and G of the model are as follows:
c21: selecting a first set of C and G hyper-parameters;
c22: constructing an SVR model according to the current C and G super parameters;
c23: recording the RMSE predicted by the SVR model training set under the current C and G super parameter combination;
c24: judging whether the traversal is finished;
if not, then
And C25: updating the combination of the C and G super parameters;
repeating steps C22-C25 until all combinations of the C and G super parameters are traversed;
c26: finding the RMSE predicted by the minimum SVR model training set;
c27: outputting the C and G super parameter combination corresponding to the RMSE predicted by the SVR model training set with the minimum.
6. The method for modeling an adaptive propagation prediction model based on SVR according to claim 5, wherein in step D, the steps of model testing and model verification are included;
during model test, a characteristic data set of a test set is put into a model to carry out regression prediction of transmission loss;
and during model verification, comparing the transmission loss regression predicted value with actual wireless channel transmission loss data, and performing processing analysis on errors, correlation coefficients and prediction distribution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310361214.0A CN116470975B (en) | 2023-04-06 | 2023-04-06 | Modeling method of SVR-based adaptive propagation prediction model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310361214.0A CN116470975B (en) | 2023-04-06 | 2023-04-06 | Modeling method of SVR-based adaptive propagation prediction model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116470975A true CN116470975A (en) | 2023-07-21 |
| CN116470975B CN116470975B (en) | 2024-07-12 |
Family
ID=87174531
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310361214.0A Active CN116470975B (en) | 2023-04-06 | 2023-04-06 | Modeling method of SVR-based adaptive propagation prediction model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116470975B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101277278A (en) * | 2008-05-21 | 2008-10-01 | 清华大学 | Wireless channel prediction technique aiming at railcar traffic |
| US20140257768A1 (en) * | 2013-03-11 | 2014-09-11 | Toyota Motor Engineering & Manufacturing North America, Inc. | Svr dynamic system modeling with delayed output measurements |
| CN111628837A (en) * | 2019-02-27 | 2020-09-04 | 中国移动通信有限公司研究院 | Channel modeling method and device |
| CN111817805A (en) * | 2020-07-08 | 2020-10-23 | 中国人民解放军国防科技大学 | Method, device and medium for adjusting channel propagation model parameters |
| CN113055111A (en) * | 2021-03-04 | 2021-06-29 | 武汉科技大学 | Channel modeling method and system based on Bayesian optimization |
-
2023
- 2023-04-06 CN CN202310361214.0A patent/CN116470975B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101277278A (en) * | 2008-05-21 | 2008-10-01 | 清华大学 | Wireless channel prediction technique aiming at railcar traffic |
| US20140257768A1 (en) * | 2013-03-11 | 2014-09-11 | Toyota Motor Engineering & Manufacturing North America, Inc. | Svr dynamic system modeling with delayed output measurements |
| CN111628837A (en) * | 2019-02-27 | 2020-09-04 | 中国移动通信有限公司研究院 | Channel modeling method and device |
| CN111817805A (en) * | 2020-07-08 | 2020-10-23 | 中国人民解放军国防科技大学 | Method, device and medium for adjusting channel propagation model parameters |
| CN113055111A (en) * | 2021-03-04 | 2021-06-29 | 武汉科技大学 | Channel modeling method and system based on Bayesian optimization |
Non-Patent Citations (1)
| Title |
|---|
| 杨心凯;袁伟娜;: "基于SVR插值的FBMC系统时变信道估计", 华东理工大学学报(自然科学版), no. 05, 30 October 2018 (2018-10-30) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116470975B (en) | 2024-07-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2020119372A1 (en) | Field intensity prediction method and apparatus, and device and storage medium | |
| CN110213003B (en) | Wireless channel large-scale fading modeling method and device | |
| US11206555B2 (en) | Method for implementing antenna azimuth correction based on user data | |
| JP2002057614A (en) | Method and device for evaluating rf propagation in radio communication system | |
| CN112543470B (en) | Terminal positioning method and system based on machine learning | |
| Redondi | Radio map interpolation using graph signal processing | |
| CN111817805B (en) | Method, device and medium for adjusting channel propagation model parameters | |
| CN110636516B (en) | Method and device for determining signal propagation model | |
| Bernardin et al. | Cell radius inaccuracy: a new measure of coverage reliability | |
| CN116470975B (en) | Modeling method of SVR-based adaptive propagation prediction model | |
| CN108574927B (en) | Mobile terminal positioning method and device | |
| CN116027118B (en) | Electromagnetic environment monitoring method and system applied to weather radar station | |
| CN104821854A (en) | Multidimensional spectrum sensing method for multiple main users based on random sets | |
| CN108519579B (en) | WiFi fingerprint positioning method for analyzing optimal AP based on interval overlapping degree | |
| CN109005552B (en) | Method for accurately evaluating wireless network based on LTE MR data | |
| CN104635206A (en) | Wireless positioning method and device | |
| CN114884597B (en) | Offshore radio wave propagation prediction method based on neural network | |
| Sukemi et al. | Path Loss Prediction Accuracy Based on Random Forest Algorithm in Palembang City Area | |
| CN114157374B (en) | Cell wireless signal strength prediction method and device and computer readable storage medium | |
| CN114422947A (en) | Method, device, equipment, storage medium and program for positioning 5G measurement report | |
| CN114980315A (en) | High-precision positioning method and system for mobile intelligent terminal | |
| CN108419247A (en) | A kind of radio communication base station cell correlation calculations method, electronic equipment and storage medium | |
| Heidari et al. | Performance evaluation of indoor geolocation systems using PROPSim hardware and ray tracing software | |
| CN117521564B (en) | Optimized design method and system for microwave circuit | |
| CN117335912B (en) | Spectrum intelligent sensing method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |