CN110912627B - Data-driven cell received power prediction method - Google Patents
Data-driven cell received power prediction method Download PDFInfo
- Publication number
- CN110912627B CN110912627B CN201911347724.2A CN201911347724A CN110912627B CN 110912627 B CN110912627 B CN 110912627B CN 201911347724 A CN201911347724 A CN 201911347724A CN 110912627 B CN110912627 B CN 110912627B
- Authority
- CN
- China
- Prior art keywords
- data
- parameters
- cell
- transmitter
- map
- 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.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/309—Measuring or estimating channel quality parameters
- H04B17/318—Received signal strength
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/373—Predicting channel quality or other radio frequency [RF] parameters
-
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W24/00—Supervisory, monitoring or testing arrangements
- H04W24/06—Testing, supervising or monitoring using simulated traffic
Abstract
The invention discloses a data-driven cell received power prediction method, which comprises the steps of preprocessing collected cell data, dividing the data into engineering parameters and map parameters according to functions, reserving all the engineering parameters, designing a characteristic link distance, an actual height, a signal line downward inclination angle and a target grid and signal line vertical distance on the basis of the engineering parameters, carrying out characteristic screening on the map parameters by calculating variance and Pearson correlation coefficients, merging the expanded engineering parameters with the screened map parameters, and then inputting the data into a cascade model XGS + LR for training to obtain a prediction model; the wireless signal coverage strength in the new environment is preferably predicted by using a prediction model, so that the wireless network construction cost is greatly reduced, and the network construction efficiency is improved.
Description
Technical Field
The invention belongs to the technical field of wireless communication, and particularly relates to a data-driven cell received power prediction method.
Background
Wireless communication is one of the most active directions in the information technology field today. With the vigorous development of the 5G technology, the influence of the technology on the human social progress is more and more paid attention by governments of various countries. In the process of building a 5G network, in consideration of increasing the coverage area of wireless signals as much as possible and reducing the station arrangement cost of base stations, reasonable wireless network planning becomes an important problem to be solved urgently. In the implementation of radio planning, Radio Signal Received Power (RSRP) prediction plays a very important role: by reasonably predicting the wireless signal propagation characteristics of the communication coverage area, the result of wireless signal receiving power prediction provides experimental basis for improving the existing network planning and user service quality on one hand, and provides directional guidance for the next generation of 5G wireless network planning on the other hand, and is one of important references for formulating a base station construction strategy.
The research methods for wireless signal received power prediction mainly include physical methods and artificial intelligence methods. The physical method is based on a wireless communication empirical model, and corresponding power is obtained through calculation according to observable data such as the position of a base station, the position of a receiving point, the height of a transmitter, the signal transmitting angle and the like. The physical method involves few features, and cannot consider the influence of some actual influence conditions, such as terrain, building size, weather conditions and electromagnetic noise generated by non-natural world, so that the measured result and the actual measured result are different from each other. The artificial intelligence method is a prediction method combining intelligent algorithms such as a machine learning algorithm and an artificial neural network, and has been widely used in recent years. The prediction method applying the neural network comprises an input layer, a hidden layer and an output layer, an end-to-end structure is used for outputting a predicted value, the prediction precision of the prediction method is premised on massive training data, and the model is poor in interpretability. The prediction method applying the machine learning algorithm is mainly based on supervised learning, can describe the highly complex relationship between input data and output data, and is suitable for processing a large amount of samples and nonlinear data. In the machine learning algorithm, XGboost is an algorithm generated by effectively combining a boosting thought and a tree model, Logistic Regression (LR) is a generalized linear model, and the effect of '1 +1> 2' can be generated by cascading the XGboost and the logistic regression.
Facebook in 2014 proposes that the problem of predicting CTR by using the click rate of the user is solved through a GBDT and LR cascading model, and the thesis finds that GBDT is a good characteristic combination mode, can enhance the expression capacity of LR, and verifies that the effect generated by the cascading model is superior to that of a single model. XGboost is an improved GBDT, the generalization performance is stronger, an authorized file CN 109886349A-a user classification method based on multi-model fusion is used for classifying users, the technical content is that a data set containing user characteristics is derived, the characteristics of the data set and the derived characteristics are input into the XGboost model to obtain a leaf node number set, and are input into an LR model through One-hot coding to realize the prediction of user classification, and the XGboost + LR cascade method solves the problems of low precision, complex iteration, incapability of well processing data sparseness and the like of the traditional model.
Compared with the traditional experience model, the acquired historical data is utilized and the machine learning technology is combined, so that manpower and material resources are saved. The prediction of the wireless signal power can be realized by adding a proper machine learning model into the original data, however, the power prediction scene is usually complex, and more factors need to be considered, so that the feature engineering is performed by combining professional knowledge on the basis of the original data, the upper limit of model expression is improved, and the method is worthy of deep research.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a cell received power prediction method based on data driving, which is used for predicting the average power of a wireless communication receiving point.
To achieve the above object, the present invention provides a data-driven cell received power prediction method, comprising:
(1) acquiring multiple groups of historical measured data of the cell, wherein each group of historical measured data comprises data with labels, engineering parameters and map parameters; then preprocessing the acquired data;
wherein the engineering parameters include: position coordinates (Cell X, Cell Y) of the transmitter, Height of the transmitter relative to the ground, horizontal direction angle A of the Cell transmitter, vertical electrical downtilt angle theta of the transmitterEDVertical mechanical down tilt theta of transmitterMDThe center frequency of a cell transmitter and the transmitting power of the cell transmitter;
the map parameters include: building height of a grid where a Cell site is located, Cell elevation of the grid where the Cell site is located, type index of a ground object of the grid where the Cell site is located, grid position coordinates (X, Y), building height of a grid position, elevation of the grid position, and index of the ground object of the grid position;
the tag data is the average power RSRP of the wireless communication receiving points;
(2) calculating the effective height h of the cell transmitterb;
hb=Height+Cell Altitude-Altitude
(3) Calculating the link distance d between the transmitter and the grid;
wherein Δ d is the grid distance;
(4) calculating the actual downward inclination angle theta of the signal lineSL;
θSL=θED+θMD
(5) Calculating the vertical distance delta h between the target grid and the signal lineue;
Δhue=cos(θSL)hb-sin(θSL)d
(6) Screening map parameters by using the variance and the Pearson correlation coefficient;
(6.1) calculating the variance D of the jth map parameterj;
Wherein x isijThe data number of the ith data in the jth map parameter is represented, wherein i is 1,2, …, n and n is the jth map parameter; x is the number ofjIs the mean value of the jth map parameter;
(6.2) sorting the variances of each map parameter in a descending order, and then eliminating the map parameters with the variances lower than a preset threshold value sigma;
(6.3) calculating a Pearson correlation coefficient P between the jth map parameter and the tag dataj;
Wherein σjIs the standard deviation of the jth map parameter, yiData i, σ, being label data RSRPyIs the standard deviation of the tag data RSRP;
(6.4) sorting according to the calculation result and rejecting PearsonThe correlation coefficient being above the threshold P*The map parameters of (a);
(7) the engineering parameters, the label data, the screened map parameters and the h obtained by calculationb、d、θSL、ΔhueStoring the data in a training data set, and carrying out standardization and discretization processing;
(8) in the training data set, a certain group of engineering parameters, the screened map parameters and the h obtained by calculationb、d、θSL、ΔhueInputting the data serving as input data into the XGboost + LR cascade model, outputting the data serving as corresponding prediction label data, and repeatedly training to obtain a prediction model;
(9) and if the engineering parameters and the map parameters of the new environment are known, processing according to the method in the steps (2) to (6), and inputting the processed engineering parameters and the map parameters into a prediction model, so that the predicted value of the RSRP under the new environment is predicted.
The invention aims to realize the following steps:
the invention relates to a data-driven cell received power prediction method, which comprises the steps of preprocessing collected cell data, dividing the data into engineering parameters and map parameters according to functions, reserving all the engineering parameters, designing a characteristic link distance, an actual height, a signal line downward inclination angle and a target grid and signal line vertical distance on the basis of the engineering parameters, carrying out characteristic screening on the map parameters by calculating variance and Pearson correlation coefficients, merging expanded engineering parameters with screened map parameters, and then inputting the data into a cascade model XGboost + LR for training to obtain a prediction model; the wireless signal coverage strength in the new environment is preferably predicted by using a prediction model, so that the wireless network construction cost is greatly reduced, and the network construction efficiency is improved.
Drawings
FIG. 1 is a flow chart of a data-driven cell received power prediction method according to the present invention;
FIG. 2 is a schematic view of rasterization;
fig. 3 is a schematic diagram of the geometry of the target grid and the transmitter.
Detailed Description
The following description of the embodiments of the present invention is provided in order to better understand the present invention for those skilled in the art with reference to the accompanying drawings. It is to be expressly noted that in the following description, a detailed description of known functions and designs will be omitted when it may obscure the subject matter of the present invention.
Examples
Fig. 1 is a flow chart of a data-driven cell received power prediction method according to the present invention.
In this embodiment, as shown in fig. 1, a data-driven cell received power prediction method according to the present invention includes the following steps:
s1, acquiring multiple groups of historical measured data of the cell, wherein each group of historical measured data comprises data with labels, engineering parameters and map parameters; then preprocessing the acquired data;
the historical measured data of the residential area is collected according to the grid position, and the data is collected in multiple aspects as much as possible for the convenience of follow-up research in the data collection process. A set of data is preprocessed: including extracting distortion data, filling missing values, removing outliers and noise data
Wherein, as shown in table 1, the engineering parameters include: position coordinates (Cell X, Cell Y) of the transmitter, Height of the transmitter relative to the ground, horizontal direction angle A of the Cell transmitter, vertical electrical downtilt angle theta of the transmitterEDVertical mechanical down tilt theta of transmitterMDThe center frequency of a cell transmitter and the transmitting power of the cell transmitter;
the map parameters include: building height of a grid where a Cell site is located, Cell elevation of the grid where the Cell site is located, type index of a ground object of the grid where the Cell site is located, grid position coordinates (X, Y), building height of a grid position, elevation of the grid position, and index of the ground object of the grid position;
the tag data is the average power RSRP of the wireless communication receiving points;
TABLE 1
In the present embodiment, the concept of grids is shown in fig. 2, where one grid has an area of 25 square meters, and one cell may be divided into a plurality of grids, and each grid position corresponds to a set of measurement data, including engineering parameters, map parameters, and tag data; the geometry of the target grid with respect to the transmitter is shown in fig. 3.
S2, calculating the effective height h of the cell transmitterbThe Height of the Cell transmitter relative to the ground, the Altitude Cell availability of the grid where the Cell site is located, and the Altitude availability of the grid location in the measurement data are expressed as follows:
hb=Height+Cell Altitude-Altitude
s3, calculating the link distance d between the transmitter and the grid, which is represented by grid position X coordinate (X), transmitter grid position X coordinate (Cell X), grid position Y coordinate (Y), and transmitter grid position Y coordinate (Cell Y) in the data information:
wherein Δ d is the grid distance;
s4, calculating the actual downward inclination angle theta of the signal wireSLThe signal line is a connection line from the transmitter to the receiving point, and the actual downward inclination angle is the vertical electrical downward inclination angle theta of the cell transmitterED(Electrical downlink) and cell transmitter vertical mechanical Downtilt angle θMD(Mechanical download) sum of both:
θSL=θED+θMD
s5, calculating the vertical distance delta h between the target grid and the signal wireue;
Δhue=cos(θSL)hb-sin(θSL)d
In this embodiment, as known from the classic wireless communication model Cost-Hata231, the characteristics obtained by S2-S5 contain rich information, and have a strong nonlinear relationship with tag data, and engineering parameters can be enhanced by adding new characteristics.
S6, screening map parameters by using the variance and the Pearson correlation coefficient;
s6.1, calculating the variance D of the jth map parameterj;
Wherein x isijThe data number of the ith data in the jth map parameter is represented, wherein i is 1,2, …, n and n is the jth map parameter;is the mean value of the jth map parameter;
s6.2, sorting the variances corresponding to the features in a descending order according to the variances shown in the table 2, and then eliminating map parameters with the variances lower than a preset threshold value sigma which is 0.1;
TABLE 2S6.3 calculation of Pearson correlation coefficient P between jth map parameter and tag dataj;
Wherein σjIs the standard deviation of the jth map parameter, yiData i, σ, being label data RSRPyIs the standard deviation of the label data RSRP.
S6.4, sorting in descending order according to the calculation result and eliminating the Pearson correlation coefficient higher than the threshold value P as shown in Table 3*0.5 map parameter.
TABLE 3
S7, calculating the engineering parameters, the label data, the screened map parameters and the calculated hb、d、θSL、ΔhueStoring the data in a training data set, and carrying out standardization and discretization processing;
s8, in the training data set, using a certain set of engineering parameters, the screened map parameters and the calculated hb、d、θSL、ΔhueInputting the data serving as input data into the XGboost + LR cascade model, outputting the data serving as corresponding prediction label data, and repeatedly training to obtain a prediction model;
in the embodiment, training data are firstly input into an independent XGboost model, and a training tuning method is a grid search + cross validation GridSearchCV method; training the XGboost model after tuning is completed, traversing each tree in the XGboost model, and calculating leaf node vectors; discretizing the obtained leaf node vector, and encoding by using One-hot-encoding; inputting the processed discrete data into an LR model for training and tuning, wherein the training and tuning method is a grid search + cross validation GridSearchCV method; and the vector of the leaf node of the XGboost is input into the LR, and is a prediction model of XGboost and LR cascade.
And S9, if the engineering parameters and the map parameters of the new environment are known, processing according to the method of the steps S2-S6, and inputting the processed parameters into a prediction model, so that the predicted value of the RSRP in the new environment is predicted.
Although illustrative embodiments of the present invention have been described above to facilitate the understanding of the present invention by those skilled in the art, it should be understood that the present invention is not limited to the scope of the embodiments, and various changes may be made apparent to those skilled in the art as long as they are within the spirit and scope of the present invention as defined and defined by the appended claims, and all matters of the invention which utilize the inventive concepts are protected.
Claims (2)
1. A data-driven cell received power prediction method is characterized by comprising the following steps:
(1) acquiring multiple groups of historical measured data of the cell, wherein each group of historical measured data comprises tag data, engineering parameters and map parameters; then preprocessing the acquired data;
(2) calculating the effective height h of the cell transmitterb;
hb=Height+Cell Altitude-Altitude
Wherein, Height is the Height of the transmitter relative to the ground, Cell elevation is the grid Altitude of the Cell site, and elevation is the grid position Altitude;
(3) calculating the link distance d between the transmitter and the grid;
wherein, Δ d is the grid distance, (X, Y) is the grid position coordinate, and (Cell X, Cell Y) is the transmitter position coordinate;
(4) calculating the actual downward inclination angle theta of the signal lineSL;
θSL=θED+θMD
Wherein, thetaEDAt a vertical electrical downtilt angle of the transmitter, thetaMDVertical mechanical down tilt for the transmitter;
(5) calculating the vertical distance delta h between the target grid and the signal lineue;
Δhue=cos(θSL)hb-sin(θSL)d
(6) Screening map parameters by using the variance and the Pearson correlation coefficient;
(6.1) calculating the jth map parameter itselfVariance D ofj;
Wherein x isijThe data number of the ith data in the jth map parameter is represented, wherein i is 1,2, …, n and n is the jth map parameter;is the mean value of the jth map parameter;
(6.2) sorting the variances of each map parameter in a descending order, and then eliminating the map parameters with the variances lower than a preset threshold value sigma;
(6.3) calculating a Pearson correlation coefficient P between the jth map parameter and the tag dataj;
Wherein σjIs the standard deviation of the jth map parameter, yiThe ith data of the tag data RSRP,is the mean, σ, of the label data RSRPyIs the standard deviation of the tag data RSRP;
(6.4) sorting according to the calculation result, and eliminating the Pearson correlation coefficient higher than the threshold value P*The map parameters of (a);
(7) the engineering parameters, the label data, the screened map parameters and the h obtained by calculationb、d、θSL、ΔhueStoring the data in a training data set, and carrying out standardization and discretization processing;
(8) in the training data set, a certain group of engineering parameters, the screened map parameters and the h obtained by calculationb、d、θSL、ΔhueAs input data, the data is input to XGboost + LR cascade moduleThe output of the model is corresponding prediction label data, and a prediction model is obtained through repeated training;
(9) and if the engineering parameters and the map parameters of the new environment are known, processing according to the method in the steps (2) to (6), and inputting the processed engineering parameters and the map parameters into a prediction model, so that the predicted value of the RSRP under the new environment is predicted.
2. The method of claim 1, wherein the engineering parameters comprise: position coordinates (CellX, CellY) of the transmitter, Height (Height) of the transmitter relative to the ground, horizontal direction angle A of the cell transmitter, and vertical electrical downtilt angle theta of the transmitterEDVertical mechanical down tilt theta of transmitterMDThe center frequency of a cell transmitter and the transmitting power of the cell transmitter;
the map parameters include: building height of a grid where a Cell site is located, Cell elevation of the grid where the Cell site is located, type index of a ground object of the grid where the Cell site is located, grid position coordinates (X, Y), building height of a grid position, elevation of the grid position, and index of the ground object of the grid position;
the label data is wireless communication receiving point average power RSRP.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911347724.2A CN110912627B (en) | 2019-12-24 | 2019-12-24 | Data-driven cell received power prediction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911347724.2A CN110912627B (en) | 2019-12-24 | 2019-12-24 | Data-driven cell received power prediction method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110912627A CN110912627A (en) | 2020-03-24 |
CN110912627B true CN110912627B (en) | 2021-05-14 |
Family
ID=69827380
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911347724.2A Active CN110912627B (en) | 2019-12-24 | 2019-12-24 | Data-driven cell received power prediction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110912627B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112367129B (en) * | 2020-07-27 | 2022-02-11 | 南京邮电大学 | 5G reference signal received power prediction method based on geographic information |
CN114095856B (en) * | 2020-07-31 | 2022-09-23 | 中国电信股份有限公司 | Processing method and processing device for saving energy of base station |
CN112118596B (en) * | 2020-09-22 | 2022-12-06 | 东南大学 | Short-range wireless signal strength prediction method based on path sequence regression |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106548210B (en) * | 2016-10-31 | 2021-02-05 | 腾讯科技(深圳)有限公司 | Credit user classification method and device based on machine learning model training |
CN109874146B (en) * | 2017-12-05 | 2020-12-25 | 华为技术有限公司 | Method and device for predicting path loss |
CN109996251B (en) * | 2017-12-29 | 2022-08-16 | 中兴通讯股份有限公司 | Method for signal coverage prediction |
CN110418354A (en) * | 2019-08-06 | 2019-11-05 | 北京邮电大学 | It is a kind of that propagation model wireless network planning method is exempted from based on machine learning |
-
2019
- 2019-12-24 CN CN201911347724.2A patent/CN110912627B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN110912627A (en) | 2020-03-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110912627B (en) | Data-driven cell received power prediction method | |
Ahmadien et al. | Predicting path loss distribution of an area from satellite images using deep learning | |
CN110197218B (en) | Thunderstorm strong wind grade prediction classification method based on multi-source convolution neural network | |
Ates et al. | Path loss exponent and shadowing factor prediction from satellite images using deep learning | |
CN110418354A (en) | It is a kind of that propagation model wireless network planning method is exempted from based on machine learning | |
CN108920481B (en) | Road network reconstruction method and system based on mobile phone positioning data | |
CN104105106B (en) | The automatic classifying identification method of wireless communication networks smart antenna covering scene | |
CN103778477B (en) | Monitor station cloth station method and apparatus | |
CN104185275B (en) | A kind of indoor orientation method based on WLAN | |
CN114362367A (en) | Cloud edge cooperation-oriented power transmission line monitoring system and method, and cloud edge cooperation-oriented power transmission line identification system and method | |
CN111405585B (en) | Neighbor relation prediction method based on convolutional neural network | |
Hayashi et al. | A study on the variety and size of input data for radio propagation prediction using a deep neural network | |
CN112285807B (en) | Meteorological information prediction method and device | |
WO2022021602A1 (en) | 5g reference signal received power prediction method based on geographic information | |
CN115619963A (en) | City building entity modeling method based on content perception | |
CN114828026A (en) | Base station planning method, device, equipment, storage medium and program product | |
CN114611388B (en) | Wireless channel characteristic screening method based on artificial intelligence | |
CN110366188B (en) | Interference measurement point deployment method, interference measurement path planning method and system | |
Nagao et al. | A study on urban structure map extraction for radio propagation prediction using XGBoost | |
EP3567757A1 (en) | Method and device for selecting reception point and transmission point in wireless communication system | |
CN106658533B (en) | Airspace flow characteristic extraction method considering user distribution | |
CN113936166A (en) | Hail echo identification method and system based on Doppler weather radar data | |
CN109286900B (en) | Wi-Fi sample data optimization method | |
Rahman et al. | Ensemble classifier based modulation recognition for beyond 5G massive MIMO (mMIMO) communication | |
CN103581667B (en) | A kind of intelligent city based on technology of wireless sensing network safe precaution method |
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 |