Electrically tunable antenna list generation method and system based on NLP

Abstract

The invention relates to the technical field of wireless network optimization, and particularly discloses an NLP-based electrically tunable antenna list generation method and system, and the method comprises the steps: obtaining a data set; performing data cleaning and standardization processing on the data set to obtain a standardized data set; training the permutation language model according to the standardized data set to obtain a first feature extraction model; according to the similarity between the actual feature code output by the first feature extraction model and the target feature code, adjusting the first feature extraction model to obtain a second feature extraction model; constructing a relation list by using the antenna feature code extracted by the second feature extraction model and the corresponding parameter set; and utilizing a coverage algorithm to match according to the subunit frequency bands in the relation list to obtain a corresponding target cell, and correspondingly updating the target cell into the relation list to obtain an electrically tunable antenna list.

Description

Technical Field

[0001] The present invention relates to the technical field of wireless network optimization, and in particular to a method and system for generating an electrically tuned antenna list based on NLP. Background Art

[0002] Most of the manufacturers of multi-band electrically-adjustable antenna equipment currently used by operators have the problem that the connected antenna equipment subunits cannot report frequency band information. When the electrically-adjustable antenna has antenna subunits of different frequency bands, the following two methods are currently used:

[0003] 1) Extract antenna frequency band feature codes based on expert experience;

[0004] 2) Manage the electrically-adjusted antenna by manually compiling statistical tables on a regular basis;

[0005] For method 1), due to the wide variety of antenna types, it is impossible to manually establish a standard for the frequency band information corresponding to all serial number extraction feature codes, and there is no combination rule for the serial number feature codes of new models of electrically adjustable antenna equipment. Relying solely on manual sorting has certain limitations. At the same time, due to the large number of antenna frequency band feature code combinations, it is impossible to fully cover all types of antenna equipment by relying solely on experience. The efficiency problems generated simultaneously during the manual identification process make it impossible to timely and accurately support related work on wireless network structure optimization;

[0006] For method 2), there is a lack of unified antenna information maintenance and management standards. When there are multiple co-frequency cells in the same sector, there is a lack of standards for judging the actual physical co-electric antenna cells. At the same time, there is a lack of management standards for repeated antenna configurations, information configuration errors, etc. There is a lack of a standardized and normalized electric antenna management process. When the electric antenna and cell information are very different, it is easy to make misjudgments and the accuracy is not high.

[0007] In summary, there are many antenna manufacturers and the models are relatively complex. In addition, the serial number and frequency band characteristics of antennas from different manufacturers or different models of the same manufacturer are completely inconsistent, making it difficult to match the electronically adjustable antenna subunits with the cells. The existing manual matching method has low accuracy and cannot effectively support large-scale network structure cell electronic tilt analysis and strategy optimization. Summary of the invention

[0008] In view of the above problems, the purpose of the present invention is to provide a method and system for generating a list of electrically adjustable antennas based on NLP, and to train a model that can identify antenna feature codes based on NLP, thereby replacing the traditional method of relying on experience to judge the corresponding frequency band of the antenna serial number, thereby improving data accuracy while reducing costs and improving benefits.

[0009] The present invention provides a method for generating a list of electrically adjustable antennas based on NLP, comprising:

[0010] Acquire a data set that can normally report subunit frequency band information; the data set includes a parameter set and a target feature code; the parameter set includes one or more of the electrically adjustable antenna manufacturer, model, subunit ID, subunit frequency band, and subunit serial number;

[0011] Performing data cleaning and standardization processing on the data set to obtain a standardized data set;

[0012] The permutation language model is trained according to the standardized data set to obtain a first feature extraction model;

[0013] According to the similarity between the actual feature code output by the first feature extraction model and the target feature code, the first feature extraction model is adjusted to obtain a second feature extraction model;

[0014] constructing a relationship list using the antenna feature codes extracted by the second feature extraction model and the corresponding parameter set;

[0015] By using a coverage algorithm, the corresponding target cell is obtained according to the sub-unit frequency band matching in the relationship list, and the target cell is updated to the relationship list accordingly, so as to obtain a list of electrically tilted antennas that can reflect the corresponding relationship between the cell and the sub-unit of the electrically tilted antenna under each frequency band.

[0016] In a possible implementation, the performing data cleaning and standardization on the data set to obtain a standardized data set includes:

[0017] Deleting abnormal data in the data set and / or supplementing blank data in the data set to obtain a cleaned data set;

[0018] According to the character length in the cleaned data set, non-standard format data in the cleaned data set is eliminated to obtain a standardized data set.

[0019] In a possible implementation, the training of the permutation language model according to the standardized data set to obtain the first feature extraction model includes:

[0020] Performing cyclic segmentation on the sub-unit strings in the standardized data set by using the arrangement language model to determine the character features with the highest word frequency in the sub-unit strings;

[0021] The arrangement language model is cyclically adjusted according to the character features to obtain a first feature extraction model.

[0022] In a possible implementation, adjusting the first feature extraction model according to the similarity between the actual feature code output by the first feature extraction model and the target feature code to obtain the second feature extraction model includes:

[0023] The first feature extraction model using different structures outputs actual feature codes;

[0024] Fusion of the plurality of actual feature codes to obtain an actual fused feature code;

[0025] According to the similarity between the actual fusion feature code and the target feature code, the first feature extraction model is adjusted to obtain a second feature extraction model.

[0026] In a possible implementation, the using of the coverage algorithm to obtain a corresponding target cell according to the sub-unit frequency band matching in the relationship list, and updating the target cell to the relationship list accordingly, so as to obtain a list of electrically tilted antennas that can reflect the corresponding relationship between the cell and the sub-unit of the electrically tilted antenna in each frequency band, includes:

[0027] A set of candidate cells with the same geographical location and the same coverage direction is obtained by matching the RRU power supply association relationship of the antenna and / or the longitude, latitude and / or azimuth in the engineering parameter data;

[0028] By using a covering algorithm, target cells corresponding to each sub-unit frequency band are extracted from the candidate cell set according to the sub-unit frequency bands in the relationship list.

[0029] In a possible implementation, obtaining a set of candidate cells with the same geographical location and the same coverage direction by matching the RRU power supply association relationship of the antenna and / or the longitude, latitude and / or azimuth in the engineering parameter data includes:

[0030] When the number of cells in the candidate cell set is greater than 1, cells in the cell set that have no RRU power supply association relationship are removed to obtain a first cell set.

[0031] In a possible implementation, the method further includes:

[0032] When multiple co-frequency cells in the candidate cell set do not have a power supply association relationship with the RRU, an abnormality list for manual verification is generated according to the multiple co-frequency cells.

[0033] In a possible implementation, the method further includes:

[0034] When the change amount of the TA value of a cell in the first cell set is less than a preset change amount, the cell is removed from the first cell set to obtain a candidate cell set.

[0035] In a possible implementation manner, when a change in the TA value of a cell in the first cell set is less than a preset change, the cell is removed from the first cell set, and before obtaining the candidate cell set, the method further includes:

[0036] Acquire cells in the first cell set whose electronic inclination angle is less than or equal to a first preset angle to obtain cells in a first inclination range;

[0037] Adjusting the inclination angles of the cells in the first inclination angle range to a second preset angle;

[0038] Acquire cells in the first cell set whose electronic inclination angle is greater than a first preset angle to obtain cells in a second inclination range;

[0039] Adjusting the inclination angles of the cells in the second inclination angle range to a third preset angle;

[0040] The third preset angle is smaller than the first preset angle, and the first preset angle is smaller than the second preset angle.

[0041] The present invention also provides a system for generating a list of electrically adjustable antennas based on NLP, which is used to implement any of the above-mentioned methods for generating a list of electrically adjustable antennas, and is characterized by comprising:

[0042] An acquisition module is used to acquire a data set that can normally report subunit frequency band information; the data set includes a parameter set and a target feature code; the parameter set includes one or more of the electric tilt antenna manufacturer, model, subunit ID, subunit frequency band, and subunit serial number;

[0043] A data processing module, used for performing data cleaning and standardization on the data set to obtain a standardized data set;

[0044] A training module, used to train the permutation language model according to the standardized data set to obtain a first feature extraction model; an adjustment module, used to adjust the first feature extraction model according to the similarity between the actual feature code output by the first feature extraction model and the target feature code to obtain a second feature extraction model;

[0045] A relationship list construction module, used to construct a relationship list using the antenna feature code extracted by the second feature extraction model and the corresponding parameter set;

[0046] The electrically tilted antenna list generation module is used to use a coverage algorithm to obtain corresponding target cells according to the sub-unit frequency band matching in the relationship list, and update the target cells to the relationship list accordingly, so as to obtain an electrically tilted antenna list that can reflect the corresponding relationship between the cells and the sub-units of the electrically tilted antennas in each frequency band.

[0047] The NLP-based electrically tilted antenna list generation method and system provided by the present invention trains a model that can identify antenna feature codes based on NLP, replaces the traditional method of relying on experience to judge the corresponding frequency band of the antenna serial number, and realizes the association between each frequency band subunit of the electrically tilted antenna and the corresponding frequency band cell. At the same time, a data identification and judgment solution is proposed to address the problem of inaccurate coverage of abnormal associated cells, so as to get rid of the influence of human factors and cover as much information as possible; abnormal data is identified and proposed in a larger range, and the accuracy of abnormal discovery is improved, so as to comprehensively and accurately realize the association relationship between the cell and the antenna subunit of the corresponding frequency band under the multi-frequency electrically tilted antenna, thereby reducing costs and improving benefits while improving data accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A schematic diagram of a flow chart of a method for generating a list of electrically adjustable antennas provided in an embodiment of the present invention;

[0049] Figure 2 A first schematic diagram of segmentation provided by an embodiment of the present invention;

[0050] Figure 3 A second schematic diagram of segmentation provided by an embodiment of the present invention;

[0051] Figure 4 A third schematic diagram of segmentation provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The following detailed description of the embodiments of the present invention is further described in detail in conjunction with the accompanying drawings and examples. The detailed description of the following embodiments and the accompanying drawings are used to exemplarily illustrate the principles of the present invention, but cannot be used to limit the scope of the present invention, that is, the present invention is not limited to the preferred embodiments described, and the scope of the present invention is defined by the claims.

[0053] In the description of the present invention, it should be noted that, unless otherwise specified, “plurality” means two or more than two; the terms “first”, “second”, etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance; for ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances.

[0054] Figure 1 A flow chart of a method for generating a list of electrically adjustable antennas provided in an embodiment of the present invention is shown in FIG. Figure 1 As shown, the method for generating a list of electrically-adjusted antennas based on NLP of the present invention comprises:

[0055] Step S1, obtaining a data set that can normally report subunit frequency band information;

[0056] The data set includes a parameter set and a target feature code; the parameter set includes one or more of the electrically adjustable antenna manufacturer, model, subunit ID, subunit frequency band, and subunit serial number. Table 1 below is a partial example of the data set.

[0057] Table 1

[0058]

[0059] Step S2, performing data cleaning and standardization on the data set to obtain a standardized data set;

[0060] In a possible implementation, abnormal data in the data set is deleted and / or blank data in the data set is supplemented to obtain a cleaned data set;

[0061] According to the character length in the cleaned data set, non-standard format data in the cleaned data set is removed to obtain a standardized data set.

[0062] The cleaned data set can more accurately show the correspondence between the subunit serial number and the subunit frequency band, which facilitates the generation of frequency band feature data. Standardized data sets can improve the reliability of data. Since the data in the data set is affected by force majeure factors such as the health of the antenna operation and the completeness of data collection, in order to eliminate the dimensional differences between different features and make different features comparable, according to the character length in the cleaned data set, the non-standard format data in the cleaned data set is eliminated to obtain a standardized data set.

[0063] Step S3, training the permutation language model according to the standardized data set to obtain a first feature extraction model;

[0064] In one possible implementation, the sub-unit string numbers in the standardized data set are cyclically segmented and segmented through a permutation language model to determine the character features with the highest word frequency in the sub-unit string numbers; the permutation language model is cyclically adjusted according to the character features to obtain a first feature extraction model.

[0065] Among them, the permutation language model is a model derived from natural language processing (NLP).

[0066] In a possible implementation, the target training work is completed by introducing a permutation language model. The subunit string number X is used as the first input of the permutation language model. The subunit string number X is composed of several characters in sequence, such as X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X22, X33, X44, X55, X66, X77, X88, X9, X12, X13, X14, X15, X16, X17, X18, X19, X20, X21, X34, X45, X56, X67, X78, X89, X9, X10, X11, X12, X13, X14, X15 4。。。。。。 Xn is composed of numbers and letters in sequence, X1, X2, X3, X 4。。。。。。 Xn corresponds to position 1, position 2, position 3, position 4, ... position n respectively.

[0067] The antenna signature Ti can be a single character or a combination of multiple characters. Assume that the antenna signature Ti to be predicted is X3, which is at position 3. In order to be able to see the character X4 at position 4 at position 1 or position 2, when X3 is at position 3, randomly arrange and combine the characters in the string, and select a part of the various possibilities after random arrangement and combination as the second input X of the arrangement language model.

[0068] At the same time, because the permutation language model XLNet adopts the Attention mask mechanism, the current input string is X, and the feature code Ti to be predicted is the i-th character. The first 1 to i-1 characters will not change in the input part. Through the Attention mask, i-1 characters are randomly selected from the input of X, that is, the Ti string, and placed at the Ti position. The input of other words is hidden through the Attention mask, and the antenna feature code can be output.

[0069] Figure 2 The first schematic diagram of segmented word segmentation provided by the embodiment of the present invention, the word segmentation order is 3→2→4→1, Figure 3 The second schematic diagram of the segmentation word segmentation provided by the embodiment of the present invention, the word segmentation order is 1→4→2→3. As shown in the figure, it is predicted that X3 is not a string at other positions, and can only be predicted based on the previous hidden state mem. Figure 3 , it can be predicted that X3 is a string that can be found at other positions.

[0070] Figure 4 The third schematic diagram of segmented word segmentation provided by an embodiment of the present invention is to input the serial number of the electrically adjustable antenna subunit as an arrangement language model, and extract the characteristic code characters in the serial number of the antenna model through the frequency of the characters contained in the antenna serial number.

[0071] The antenna serial number "M2898K31DWV77RTb3" is segmented into several character combinations using the permutation language model XLNet, namely the feature code "M2, 28, 89, 98······b3", and the feature code with the highest frequency is output according to the antenna model. At the same time, the frequency data is integrated to further enhance the accuracy of the feature code. The feature code of this type of antenna model input this time is combined with the historical frequency data. The feature code 98 appears the most, and the feature code of this frequency band is output as 98. Other feature codes are hidden through the Attention mask.

[0072] The hidden state can be thought of as the "memory" or "information" of the model. At each time step, the model computes a new hidden state based on the current input and the hidden state from the previous time step. This "memory" capability enables the capture of long-term dependencies in time series and plays an important role in understanding and predicting the next element in a sequence.

[0073] Step S4, adjusting the first feature extraction model according to the similarity between the actual feature code output by the first feature extraction model and the target feature code to obtain a second feature extraction model;

[0074] In a possible implementation, a first feature extraction model with a different structure is used to output an actual feature code; multiple actual feature codes are fused to obtain an actual fused feature code; and the first feature extraction model is adjusted according to the similarity between the actual fused feature code and the target feature code to obtain a second feature extraction model.

[0075] This step evaluates the first feature extraction model, determines the accuracy of the actual feature code output by the first feature extraction model, and then makes some fine-tuning to the first feature extraction model to improve data accuracy. By integrating and adjusting multiple models, the robustness and generalization ability of the second feature extraction model can be improved. At the same time, the training data of the model is also transformed to generate new training samples. In the calculation of antenna serial number similarity, data enhancement can further improve the generalization ability of the model.

[0076] Step S5, constructing a relationship list using the antenna feature code extracted by the second feature extraction model and the corresponding parameter set;

[0077] The relationship list can realize the output of antenna subunit serial numbers and corresponding subunit frequency bands reported by multiple manufacturers.

[0078] Step S6, using the coverage algorithm, obtain the corresponding target cell according to the sub-unit frequency band matching in the relationship list, and update the target cell to the relationship list accordingly, to obtain a list of electrically tilted antennas that can reflect the corresponding relationship between the cell and the sub-unit of the electrically tilted antenna under each frequency band.

[0079] In a possible implementation, step S61, obtaining a set of candidate cells with the same geographical location and the same coverage direction through matching of the RRU power supply association relationship of the antenna and / or the longitude, latitude and / or azimuth in the engineering parameter data;

[0080] In an example, a set of candidate cells is generated based on cells within 10 meters from the cell corresponding to the currently powered RRU and with an azimuth angle less than ±30°.

[0081] In one example, the cell corresponding to the antenna-powered RRU uses the longitude and latitude + azimuth in the engineering parameter data to match other frequency band cells in the same geographical location and the same coverage direction.

[0082] In one example, the same geographical location means that the difference in longitude and latitude of the two cells is less than 10 meters. The same coverage direction means that the absolute value of the difference in azimuth angle of the two cells is less than 30°.

[0083] In a possible implementation, S61 includes:

[0084] When multiple cells with the same frequency in the candidate cell set have no RRU power supply association relationship, an exception list for manual verification is generated based on the multiple cells with the same frequency in the candidate cell set, and does not participate in the subsequent generation of the electrical adjustment antenna list.

[0085] When the number of cells in the candidate cell set is greater than 1, the cells without RRU power supply association in the candidate cell set are eliminated to obtain the first cell set; that is, when there are multiple co-frequency cells in the coverage data due to longitude, latitude or azimuth, the cells with RRU power supply association are retained first to ensure the accuracy of the data.

[0086] Acquire cells whose electronic inclination angles are less than or equal to the first preset angle in the first cell set to obtain cells in a first inclination range; adjust the inclination angles of the cells in the first inclination range to a second preset angle;

[0087] Obtain cells in the first cell set whose electronic inclination angle is greater than the first preset angle to obtain cells in the second inclination range; adjust the inclination angles of the cells in the second inclination range to a third preset angle; the third preset angle is smaller than the first preset angle, and the first preset angle is smaller than the second preset angle.

[0088] In one example, the inclination angles of cells with electronic inclination angles less than or equal to 6° are uniformly adjusted to 12° to observe whether the TA value of the cell changes. When the inclination angle of the cell increases significantly, it means that the cell coverage distance becomes closer and the TA value of the cell will shrink significantly.

[0089] Adjust the inclination angle of cells with electronic inclination angle greater than 6° to 2° and observe whether the TA value of the cell changes. When the inclination angle becomes significantly smaller, it means that the coverage distance of the cell becomes longer and the TA value of the cell will be significantly expanded.

[0090] In this way, it is possible to determine whether the cell is physically connected to the corresponding electrically tilted antenna by the change in the TA value of the cell, thereby avoiding incorrect management of cells that are not actually connected to the electrically tilted antenna.

[0091] When the variation of the TA value of a cell in the first cell set is greater than or equal to a preset variation, it indicates that the cell is physically connected to the corresponding electrically-adjusted antenna.

[0092] When the change in the TA value of a cell in the first cell set is less than a preset change, it indicates that there is no physical connection between the cell and the corresponding electrically-adjusted antenna, and the cell is removed from the first cell set to obtain a candidate cell set.

[0093] Step S62: extracting a target cell corresponding to the sub-unit frequency band from the candidate cell set according to the sub-unit frequency band in the relationship list.

[0094] The present invention also provides a system for generating a list of electrically adjustable antennas based on NLP, which is used to implement any of the above-mentioned methods for generating a list of electrically adjustable antennas, comprising:

[0095] The acquisition module is used to obtain a data set that can normally report subunit frequency band information; the data set includes the manufacturer, model, subunit ID, subunit frequency band, subunit serial number and target feature code of the electrically adjustable antenna;

[0096] A data processing module is used to clean and standardize the data set to obtain a standardized data set;

[0097] A training module, used for training the permutation language model according to the standardized data set to obtain a first feature extraction model;

[0098] An adjustment module, used to adjust the first feature extraction model according to the similarity between the actual feature code output by the first feature extraction model and the target feature code, so as to obtain a second feature extraction model;

[0099] A relationship list building module, used to build a relationship list using the antenna feature code extracted by the second feature extraction model, and the corresponding electrically adjustable antenna manufacturer, model, subunit ID, subunit frequency band and subunit serial number;

[0100] The electric tilt antenna list generation module is used to use the coverage algorithm to obtain the corresponding target cell according to the sub-unit frequency band matching in the relationship list, and update the target cell to the relationship list accordingly, so as to obtain the electric tilt antenna list that can reflect the corresponding relationship between the cell and the sub-unit of the electric tilt antenna under each frequency band.

[0101] The NLP-based method and system for generating a list of electrically-adjusted antennas provided by the present invention solves the problem of difficulty in associating existing electrically-adjusted antenna subunits with cells. The present invention mainly utilizes the NLP-XLNet advanced text reasoning function and uses the correspondence between the serial number and the frequency band reported by the existing equipment as a training data set for extracting the antenna frequency band serial number feature code.

[0102] Based on NLP, a model that can identify antenna feature codes is trained to replace the traditional method of relying on experience to determine the corresponding frequency band of the antenna serial number, so as to associate each frequency band sub-unit of the electrically adjustable antenna with the corresponding frequency band cell.

[0103] At the same time, a data identification and judgment solution was proposed to address the problem of inaccurate coverage of abnormally associated cells, in order to get rid of the influence of human factors and cover as much information as possible.

[0104] Abnormal data can be identified and presented on a larger scale, which improves the accuracy of detecting abnormalities, so as to fully and accurately realize the association between the cell and the antenna subunit of the corresponding frequency band under the multi-frequency electrically adjustable antenna, thereby reducing costs and improving benefits while improving data accuracy.

[0105] Through the present invention, the tilt angle of the electrically-adjustable antenna subunit in different frequency bands can be effectively managed, and the relevant information data can effectively support the analysis and optimization of the 4G and 5G wireless network structures. This method can provide a comprehensive reference for the formulation of wireless network structure optimization strategies, help better evaluate the effects of different optimization strategies, further optimize and improve the structure optimization scheme, and improve the performance and efficiency of the wireless network. Increase the positive benefits brought by network structure optimization.

[0106] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for generating a list of electrically adjustable antennas based on NLP, characterized in that: include: Acquire a data set that can normally report subunit frequency band information; the data set includes a parameter set and a target feature code; the parameter set includes one or more of the electrically adjustable antenna manufacturer, model, subunit ID, subunit frequency band, and subunit serial number; Performing data cleaning and standardization processing on the data set to obtain a standardized data set; The permutation language model is trained according to the standardized data set to obtain a first feature extraction model; According to the similarity between the actual feature code output by the first feature extraction model and the target feature code, the first feature extraction model is adjusted to obtain a second feature extraction model; constructing a relationship list using the antenna feature codes extracted by the second feature extraction model and the corresponding parameter set; By using a coverage algorithm, the corresponding target cell is obtained according to the sub-unit frequency band matching in the relationship list, and the target cell is updated to the relationship list accordingly, so as to obtain a list of electrically tilted antennas that can reflect the corresponding relationship between the cell and the sub-unit of the electrically tilted antenna under each frequency band.

2. The method for generating a list of electrically-adjustable antennas according to claim 1, wherein: The data set is cleaned and standardized to obtain a standardized data set, including: Deleting abnormal data in the data set and / or supplementing blank data in the data set to obtain a cleaned data set; According to the character length in the cleaned data set, non-standard format data in the cleaned data set is eliminated to obtain a standardized data set.

3. The method for generating a list of electrically-adjustable antennas according to claim 1, wherein: The training of the permutation language model according to the standardized data set to obtain the first feature extraction model comprises: Performing cyclic segmentation on the sub-unit strings in the standardized data set by using the arrangement language model to determine the character features with the highest word frequency in the sub-unit strings; The arrangement language model is cyclically adjusted according to the character features to obtain a first feature extraction model.

4. The method for generating a list of electrically-adjustable antennas according to claim 1, wherein: The adjusting the first feature extraction model according to the similarity between the actual feature code output by the first feature extraction model and the target feature code to obtain the second feature extraction model includes: The first feature extraction model using different structures outputs actual feature codes; Fusion of the plurality of actual feature codes to obtain an actual fused feature code; According to the similarity between the actual fusion feature code and the target feature code, the first feature extraction model is adjusted to obtain a second feature extraction model.

5. The method for generating a list of electrically-adjustable antennas according to claim 1, wherein: The use of the coverage algorithm to obtain a corresponding target cell according to the sub-unit frequency band matching in the relationship list, and updating the target cell to the relationship list accordingly, to obtain a list of electrically tilted antennas that can reflect the corresponding relationship between the cells and the sub-units of the electrically tilted antennas in each frequency band includes: A set of candidate cells with the same geographical location and the same coverage direction is obtained by matching the RRU power supply association relationship of the antenna and / or the longitude, latitude and / or azimuth in the engineering parameter data; By using a covering algorithm, target cells corresponding to each sub-unit frequency band are extracted from the candidate cell set according to the sub-unit frequency bands in the relationship list.

6. The method for generating a list of electrically-adjustable antennas according to claim 5, characterized in that: The candidate cell set with the same geographical location and the same coverage direction obtained by matching the RRU power supply association relationship of the antenna and / or the longitude, latitude and / or azimuth in the engineering parameter data includes: When the number of cells in the candidate cell set is greater than 1, cells in the cell set that have no RRU power supply association relationship are removed to obtain a first cell set.

7. The method for generating a list of electrically-adjustable antennas according to claim 6, wherein: Also includes: When multiple co-frequency cells in the candidate cell set do not have a power supply association relationship with the RRU, an abnormality list for manual verification is generated according to the multiple co-frequency cells.

8. The method for generating a list of electrically-adjustable antennas according to claim 6, wherein: Also includes: When the change amount of the TA value of a cell in the first cell set is less than a preset change amount, the cell is removed from the first cell set to obtain a candidate cell set.

9. The method for generating a list of electrically-adjustable antennas according to claim 8, wherein: When the change amount of the TA value of a cell in the first cell set is less than a preset change amount, the cell is removed from the first cell set, and before obtaining the candidate cell set, the method further includes: Acquire cells in the first cell set whose electronic inclination angle is less than or equal to a first preset angle to obtain cells in a first inclination range; Adjusting the inclination angles of the cells in the first inclination angle range to a second preset angle; Acquire cells in the first cell set whose electronic inclination angle is greater than a first preset angle to obtain cells in a second inclination range; Adjusting the inclination angles of the cells in the second inclination angle range to a third preset angle; The third preset angle is smaller than the first preset angle, and the first preset angle is smaller than the second preset angle.

10. A system for generating a list of electrically adjustable antennas based on NLP, used to implement the method for generating a list of electrically adjustable antennas as claimed in any one of claims 1 to 9, characterized in that: include: An acquisition module is used to acquire a data set that can normally report subunit frequency band information; the data set includes a parameter set and a target feature code; the parameter set includes one or more of the electric tilt antenna manufacturer, model, subunit ID, subunit frequency band, and subunit serial number; A data processing module, used for performing data cleaning and standardization on the data set to obtain a standardized data set; A training module, used for training the permutation language model according to the standardized data set to obtain a first feature extraction model; an adjustment module, configured to adjust the first feature extraction model according to the similarity between the actual feature code output by the first feature extraction model and the target feature code, so as to obtain a second feature extraction model; A relationship list construction module, used to construct a relationship list using the antenna feature code extracted by the second feature extraction model and the corresponding parameter set; The electric tilt antenna list generation module is used to use a coverage algorithm to obtain corresponding target cells according to the sub-unit frequency band matching in the relationship list, and update the target cells to the relationship list accordingly, so as to obtain an electric tilt antenna list that can reflect the corresponding relationship between the cells and the sub-units of the electric tilt antennas in each frequency band.

Images