Base station load prediction method and device, storage medium and electronic equipment

Abstract

The invention discloses a base station load prediction method and device, a storage medium and electronic equipment, and relates to the technical field of artificial intelligence, and the method comprises the steps: carrying out the analysis of load prediction query information through employing a large language model under the condition of obtaining a load prediction request, analyzing the intention of a user, and obtaining the feature data of a target activity. Then, inputting the characteristic data of the target activity into the prediction model to carry out load prediction, and obtaining a minute-level load prediction result of the base station cell in the communication guarantee area in a target activity period; the prediction model is obtained based on the related information of the held activity, the minute-level load data of the base station cell related to the held activity during the holding period and the base station feature data, so that the minute-level load result of the base station cell can be accurately predicted.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a base station load prediction method, device, storage medium, and electronic device. Background Art

[0002] In today's digital age, network communications have become the cornerstone of social operations and economic development. From mobile phone calls and online shopping in daily life to remote control and intelligent logistics in industrial production, remote medical consultations, and online courses in education, communication networks carry massive amounts of information. Their performance directly impacts operational efficiency and user experience in various fields. Therefore, how to secure and optimize communication networks is a technical issue worthy of research.

[0003] In the field of network optimization and security, load forecasting is carried out for base station cells. Flexible adjustment of the capacity of base station cells based on the load forecast results can address possible network congestion and resource waste problems, thereby improving operational efficiency and user experience in various fields.

[0004] Existing base station cell load forecasting technologies primarily include time-series modeling based on historical data and determining the maximum load of a base station cell using coefficients set based on statistical rules. However, these technologies struggle to achieve precise, full-time load forecasting in scenarios such as communication support for critical events, and they struggle to address the impact of irregular user flows on forecast accuracy. Furthermore, the load forecasting process is typically only handled by specialized technicians, requiring the initial preparation of feature data to be input into the model based on query requirements, and the model then outputs the forecast results, lacking intelligent interaction. Summary of the Invention

[0005] In view of this, the present application provides a base station load prediction method, device, storage medium and electronic device, which can improve the accuracy of base station load prediction and reduce the complexity of the load prediction process.

[0006] In a first aspect, the present application provides a base station load prediction method, comprising:

[0007] When a load forecast request is received, the load forecast query information is input into the large language model for processing to obtain characteristic data of the target activity; the load forecast query information includes the natural language input by the user to query the load conditions of the base station cells within the communication guarantee area during the period of the target activity;

[0008] The characteristic data of the target activity is input into the prediction model for load prediction to obtain the load prediction result; the load prediction result represents the minute-level load prediction result of the base station cells in the communication guarantee area during the target activity period; the prediction model is obtained based on the relevant information of the activities held, the minute-level load data of the base station cells related to the activities held during the period, and the base station characteristic data.

[0009] In the above embodiment, a large language model is used to analyze the natural language input by the user to obtain the characteristic data of the target activity. Because the large language model can analyze the natural language input by the user and accurately predict the user's intention, there is no need for technical personnel to query the corresponding characteristic data for the target activity, thus simplifying the load forecasting process. Furthermore, because the prediction model is based on relevant information about the held events, minute-level load data of the base station cells related to the held events during the period, and base station characteristic data, the prediction model can achieve refined minute-level load forecasting for all time periods.

[0010] Optionally, before obtaining the load forecast request, the method further includes:

[0011] Obtaining a first activity data set related to a previously held activity; the first activity data set is obtained by crawling relevant information of the previously held activity using a crawler program;

[0012] Obtaining a first base station data set related to base station cells within the communication guarantee area; the first base station data set includes minute-level load data and base station characteristic data of the base station cells related to the held event;

[0013] The model to be trained is trained using the first activity dataset and the first base station dataset.

[0014] In the above embodiment, a crawler program is used to crawl information about events held in the communication support area to obtain a first activity dataset, which can provide training data for the subsequent prediction model. Compared to existing load forecasting methods, using this information and the corresponding base station load data as training data for the prediction model can improve the load forecasting accuracy of the prediction model. In addition, the crawler program can efficiently obtain massive amounts of data, ensuring the comprehensiveness and accuracy of the data.

[0015] Optionally, training the to-be-trained model using the first activity dataset and the first base station dataset includes:

[0016] A first feature data set is obtained based on the first activity data set, the first base station data set, and the first association relationship; the first association relationship represents the correspondence between relevant information of the held activity and minute-level load data and base station feature data of the base station cell during the held activity;

[0017] Training and parameter adjustment of the to-be-trained model according to the first feature data set;

[0018] When the prediction accuracy of the model to be trained meets the requirements, the prediction model is obtained.

[0019] In the above embodiment, since the first feature data set includes relevant information about activities that have been held and the corresponding relationship between the load data of the base station cell and the base station feature data, using the first feature data set to train the model to be trained can enable the model to be trained to better learn the characteristic relationship between the activities and the load data of the base station cell, thereby improving the load prediction accuracy of the prediction model.

[0020] Optionally, the load forecast query information is input into a large language model for processing to obtain characteristic data of the target activity, including:

[0021] Utilize the feature extraction tool included in the large language model to parse the natural language input by the user to obtain the identification information of the target activity;

[0022] The intent understanding tool included in the large language model is used to match the identification information of the target activity with the second feature data set to obtain the feature data of the target activity; the second feature data set includes the identification information of the activities planned to be held in the communication guarantee area, the base station feature data and related information of the planned activities.

[0023] In the above embodiment, a large language model is used to analyze the natural language input of users to obtain characteristic data for the target activity. Because the large language model can analyze the natural language input of users and accurately predict user intent, it eliminates the need for technical personnel to query the corresponding characteristic data for the target activity, thus simplifying the load forecasting process.

[0024] Optionally, the method further includes:

[0025] Obtaining a second activity data set related to the communication support area; the second activity data set is obtained by crawling relevant information of activities planned to be held in the communication support area based on a crawler program;

[0026] Acquire a second base station data set related to a base station cell within the communication guarantee area; the second base station data set includes base station characteristic data of the base station cell;

[0027] A second feature data set is established based on the second activity data set, the second base station data set, and the second association relationship; the second association relationship represents the correspondence between the identification information of the planned activity and the feature data of the planned activity.

[0028] In the above embodiment, using a crawler program to crawl information about events held in the communication support area to obtain the second activity dataset provides key data for the large language model to match the target event's identification information with the second feature dataset. Furthermore, crawlers can efficiently acquire massive amounts of data, ensuring its comprehensiveness and accuracy.

[0029] Optionally, matching the identification information of the target activity with the second feature data set to obtain feature data of the target activity includes:

[0030] Performing similarity matching on the identification information of the target activity and the identification information of the planned activity included in the second feature data set to obtain a similarity matching result; the similarity matching result is used to represent the degree of similarity between the identification information of the target activity and the identification information of the planned activity;

[0031] According to the similarity matching result and the second association relationship, characteristic data of the target activity is obtained.

[0032] In the above embodiment, since the natural language input by the user contains insufficient feature data, by calculating the similarity matching degree between the identification information of the target activity and the identification information of the planned activity included in the second feature data set, obtaining the feature data of the target activity from the pre-constructed second data set can enable the prediction model to obtain accurate and comprehensive input, thereby improving the load prediction accuracy.

[0033] Optionally, after obtaining the load forecast result, the method further includes:

[0034] Determine the load conditions of base station cells within the communication support area during the target event period based on the load forecast results;

[0035] If the base station cell is overloaded during the target event, the communication load capacity in the communication guarantee area will be increased based on the load forecast results.

[0036] In the above embodiment, adjusting the communication load capacity in the communication guarantee area according to the load forecast result can avoid the problem of poor user experience caused by insufficient communication load capacity in the communication guarantee area, thereby improving user experience.

[0037] In a second aspect, the present application provides a base station load prediction device, comprising:

[0038] The acquisition module is configured to, upon receiving a load forecast request, input load forecast query information into a large language model for processing to obtain characteristic data of a target activity; the load forecast query information includes a natural language input by a user to query the load conditions of base station cells within a communication support area during the period of the target activity;

[0039] The prediction module is configured to input characteristic data related to the target activity into the prediction model to perform load prediction and obtain a load prediction result; the load prediction result represents the minute-level load prediction result of the base station cell in the communication guarantee area during the target activity period; the prediction model is obtained based on relevant information of the activities held, minute-level load data of the base station cell related to the activities held during the period, and base station characteristic data.

[0040] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which implements the method described in the first aspect when the computer program is executed by a processor.

[0041] In a fourth aspect, the present application provides an electronic device comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor implements the method described in the first aspect when executing the computer program.

[0042] In a fifth aspect, the present application provides a computer program product having a computer program stored thereon, and when the computer program product is executed by a processor, the method described in the first aspect is implemented.

[0043] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0046] Figure 1 A flow chart of a base station load prediction method provided in an embodiment of the present application is shown;

[0047] Figure 2 A schematic diagram of an example process provided by an embodiment of the present application is shown;

[0048] Figure 3 A schematic diagram of an example process provided by an embodiment of the present application is shown;

[0049] Figure 4A schematic structural diagram of a base station load prediction device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0050] To facilitate the explanation of the embodiments of the present application, some technical terms and technical means related to the embodiments of the present application, as well as application scenarios of the embodiments of the present application are first introduced below.

[0051] A web crawler is a program or script that automatically crawls internet information according to specific rules. Crawler programs can be used to efficiently obtain massive amounts of data. In embodiments of the present application, crawlers can be used to obtain information related to events. These events are those with a high density of people, such as concerts, sporting events, theatrical performances, large-scale lectures, and the like. Information related to these events may include, for example, the start time, venue size, event logo, and the like.

[0052] It should be noted that the information related to the activities crawled by this application using a crawler program is public information on the Internet and does not involve the acquisition of personal privacy information.

[0053] The Large Language Model (LLM) is an artificial intelligence model based on deep learning. The LLM can understand and generate natural language and perform a variety of language-related tasks. In the embodiments of the present application, the LLM can be used as a natural language understanding tool, or as a port for interacting with technicians, etc. The technician can use the question-and-answer port provided by the LLM to input natural language to query the load forecast results of a certain activity in a certain area, and then after processing by the embodiments of the present application, the load forecast results are fed back to the technician.

[0054] The machine learning algorithm based on the gradient boosting tree (eXtreme Gradient Boosting, XGB) is an accurate prediction model. The XGB model is one of the current mainstream prediction models due to its efficient nonlinear fitting, feature interaction mining and computing performance. In an embodiment of the present application, the XGB model (prediction model) is mainly used to predict the load conditions of base station cells in hot spots. These hot spots can be, for example, the venues for hosting events mentioned above and areas with dense crowds around the venues. The load conditions of the base station cell can be defined by the base station cell capacity (CellCapacity), which refers to the maximum amount of business or number of users that the base station cell can handle simultaneously while meeting the quality of service (QoS) requirements.

[0055] During large-scale events, due to the characteristics of dense crowds, large flow of people, and heavy base station load pressure within a short period of time (1-2 hours), it is an important application scenario for communication load prediction.

[0056] The embodiment of the application provides a base station load prediction method, which can be applied to a load prediction scene of a large-scale event. The embodiment of the application obtains relevant information of events held by using a crawler program, and obtains a prediction model (for example, an XGB model) according to the information of the events and operating parameter data, load data and the like of a communication base station. Then, a large language model is used as an interactive port and an intention understanding tool to understand the intention of a user to query a load prediction result of a relevant base station cell during an event. According to the natural language input by the user, feature data that can be understood by the prediction model is matched. The prediction model reasons the feature data and outputs a load prediction result. Finally, the load prediction result is fed back to the user through the interactive port of the LLM model. The technical scheme provided by the embodiment of the application simplifies a load prediction process and improves load prediction accuracy.

[0057] It should be noted that the operating parameter data of the communication base station includes key information such as a physical position, a hardware configuration and network parameters of the base station, and the running data includes data amounts of uplink and downlink at each time node of the base station and a number of users accessing the base station and the like. The operating parameter data and the running data of the communication base station are obtained from an operation and maintenance database after authorization.

[0058] The embodiments of the application will be described in more detail below with reference to the accompanying drawings. It should be noted that the embodiments in the application and the features in the embodiments can be combined with each other without conflict.

[0059] The embodiment provides a base station load prediction method, as shown in Figure 1 The method comprises the following steps.

[0060] S101, in the case of obtaining a load prediction request, inputting load prediction query information to a large language model for processing to obtain feature data of a target event.

[0061] The load prediction request may, for example, be a control instruction generated based on a user asking for a base station cell load prediction result, and the control instruction is used to call an interactive port of a large language model (LLM). The load prediction query information includes natural language input by the user to query load conditions of a base station cell in a communication guarantee area during holding of a target event.

[0062] In some examples, the load prediction query information may, for example, be "Please help me query the load prediction condition of the xx concert held on xx day in xx month." The large language model processes the load prediction query information to obtain feature data of the target event (the xx concert).

[0063] The characteristic data of the target activity is structured data related to the target activity. Such structured data may include, for example, the start and end time of the target activity, as well as base station characteristic data. Specifically, the base station characteristic data may include base station point of interest (POI) data and / or base station engineering parameter data. Base station POI data may include, for example, the geographic coordinates of the base station and base station attribute information. The base station attribute information may include, for example, the base station's communication demand level and the number of users.

[0064] In some examples, feature data of a target activity is used as input to a prediction model.

[0065] S102: Input the characteristic data of the target activity into the prediction model to perform load prediction and obtain the load prediction result.

[0066] The load forecast result is a minute-level load forecast result representing the base station cells within the communication guarantee area during the target activity period.

[0067] In some examples, the communication guarantee area includes, for example, the venue where the target event is held and crowded places around the venue, such as subway stations and bus stops around the venue.

[0068] The prediction model (eg, XGB model) is obtained based on relevant information of past events and minute-level load data of base station cells related to the past events during the period of the events.

[0069] The embodiment of the present application uses a large language model to analyze the natural language input by the user to obtain the characteristic data of the target activity. Since the large language model can analyze the natural language input by the user and can accurately predict the user's intention, there is no need for technical personnel to query the corresponding characteristic data for the target activity, thereby simplifying the load forecasting process. Furthermore, since the prediction model is based on relevant information about the activities held and the minute-level load data of the base station cells related to the activities held during the period, the prediction model can be used to achieve refined minute-level load forecasting for all time periods.

[0070] The base station load prediction method provided in the embodiment of the present application is described in detail below.

[0071] Before performing load forecasting, the model needs to be trained so that it can accurately output load forecasting results.

[0072] The specific implementation method of training the model is given in the embodiment of this application, such as Figure 2 Shown, including:

[0073] S201: Acquire a first activity data set related to a previously held activity.

[0074] The first activity data set is obtained by crawling relevant information of past activities using a crawler program.

[0075] In some examples, the first activity data set may include relevant information about activities held in different communication guarantee areas. It may also include relevant information about activities held in the same communication guarantee area. For example, taking a concert as an example, different communication guarantee areas include venues in multiple cities or regions such as venue A, venue B, and venue C. The crawler program can be used to obtain information about concerts held in different venues from web pages. The information about these concerts includes the identification information of the concert (for example, the name of the concert, the name of the singer of the concert, etc.), the start time and end time of the concert. In addition, it can also include concert attendance, the number of seats in the venue, and other multiple dimensions of concert-related data, which will not be described in detail in this application.

[0076] It can be seen that obtaining concert data in multiple dimensions can increase the richness of the data, thereby increasing the accuracy of the load forecast results output by the prediction model to a certain extent.

[0077] The same communication support area includes a specific venue, for example, Venue A. A crawler program is used to retrieve information about concerts held at Venue A from web pages. This information is the same as the above method for obtaining concert information for multiple venues.

[0078] In some examples, crawler programs can also be used to obtain information about concerts in a certain venue or multiple venues within a period of time from web pages.

[0079] For example, by setting crawling rules of a crawler program, information about concerts held in a certain venue or multiple venues in the past year can be obtained from the web page.

[0080] It's clear that setting a time range can prevent outdated and distorted data from impacting prediction model training, improving data quality and, to a certain extent, increasing the accuracy of the prediction model. For example, venue renovations (adding seats) or changes in a singer's fame (rising or falling popularity) can be used. For this purpose, you can also set targeted crawling rules based on the popularity of certain singers and venue renovation information. This prevents poor-quality data from affecting the prediction model's accuracy.

[0081] S202: Acquire a first base station data set in a communication guarantee area where the event was held.

[0082] The first base station data set includes minute-level load data and base station characteristic data of base station cells related to the activities held during the period of the activities;

[0083] In some examples, the operation and maintenance database is accessed through query instructions to obtain minute-level load data and base station characteristic data of the base station cell within a preset time range.

[0084] For example, the start and end time of a concert can be determined based on the first activity data set obtained above, and the query range of the query instruction can be constrained based on the start and end time of the concert. Furthermore, minute-level load data and base station characteristic data of each base station cell during the concert period can be obtained.

[0085] It should be noted that the minute-level load data for a base station cell and the base station characteristic data may be stored in the same operation and maintenance database or in different databases. When querying base station characteristic data using a query command, the query range can be directly constrained based on the geographical characteristics of the concert to obtain base station characteristic data within the specified range. Base station characteristic data can include base station POI data and / or base station operating parameter data.

[0086] S203: Train the to-be-trained model using the first activity data set and the first base station data set to obtain a prediction model.

[0087] In the above embodiment, a crawler program is used to crawl information about events held in the communication support area to obtain a first activity dataset, which can provide training data for the subsequent prediction model. Compared to existing load forecasting methods, using this information and the corresponding base station load data as training data for the prediction model can improve the load forecasting accuracy of the prediction model. In addition, the crawler program can efficiently obtain massive amounts of data, ensuring the comprehensiveness and accuracy of the data.

[0088] Furthermore, the embodiment of the present application also provides a specific implementation method for using the first activity data set and the first base station data set to train the to-be-trained model to obtain a prediction model. The training process of the model includes the following steps:

[0089] Step 1: Obtain a first feature data set according to the first activity data set, the first base station data set, and the first association relationship.

[0090] The first association relationship represents the corresponding relationship between the relevant information of the held event and the minute-level load data and base station characteristic data of the base station cell during the held event.

[0091] In some examples, the first association relationship represents the correspondence between feature data and labels in the model training data, which constitutes sample data in the model training.

[0092] In some examples, the first feature dataset can be divided into a training set, a validation set and a test set. Among them, the training set accounts for 60 to 80 percent of the total amount of the first feature dataset, and the validation set accounts for 10 to 20 percent of the total amount of the first feature dataset. The test set accounts for the remaining part of the first feature dataset.

[0093] Step two, training and parameter adjustment of the to-be-trained model according to the first feature dataset.

[0094] In some examples, by dividing the first feature dataset through the above-mentioned embodiments, the training set can be used to train the to-be-trained model, the validation set can be used to adjust the parameters of the trained model, and finally the test set can be used to test the prediction accuracy of the prediction model.

[0095] Specifically, training the to-be-trained model using the training set mainly includes the following stages: initialization of the model, iterative training, model saving, etc.

[0096] Among them, the model training can include the following specific steps:

[0097] First step: initialize the prediction value; usually the mean of the target variable (regression) or the logit (classification). In the embodiments of the present application, it can be set as the mean of the target variable (load data).

[0098] Second step: calculate the gradient and Hessian matrix; for each sample data (feature data and label), calculate the first-order derivative (gradient) and second-order derivative (Hessian) of the current prediction value.

[0099] Third step: build a decision tree; based on the gradient and Hessian matrix, find the optimal split point (by maximizing the objective function gain).

[0100] Apply regularization (such as max_depth limit tree depth, gamma control split threshold).

[0101] Fourth step: update the model; add the prediction result of the new tree to the total prediction, and multiply it by the learning rate (learning_rate) to control the step size.

[0102] Fifth step: repeat iteration: until the maximum number of trees is reached or the validation set performance no longer improves (early stopping).

[0103] In the embodiments of the present application, the iteration stage can be performed in the form of minimizing the loss function. For example, the loss function of the model can be expressed as the following formula (I):

[0104]

[0105] Among them, is the overall loss of the iteration, is the loss base value predicted by the model after the previous round of iteration, g i Is the loss function for the previous round of prediction results The first derivative (gradient), h i is the second-order derivative (Hessian matrix element), x i are the input variables of the model.

[0106] The above formula (1) is an approximate form based on the second-order Taylor expansion, which can more accurately approximate the loss function and assist in optimizing the new tree f t (x i ) The correction of the prediction results makes the model optimization more efficient and accurate, and improves the generalization ability of the model.

[0107] After the maximum number of trees is reached or the performance of the validation set no longer improves, the trained model is obtained.

[0108] Then, the parameters of the trained model are adjusted. Parameter adjustment can be performed using grid search strategy, random search strategy, Bayesian optimization strategy, etc. I will not go into details here.

[0109] Step 3: When the prediction accuracy of the model to be trained meets the requirements, a prediction model is obtained.

[0110] In some examples, a test set can be used to test the training model, and the mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE), etc. can be used as criteria to determine whether the prediction accuracy meets the requirements.

[0111] In the above embodiment, since the first feature data set includes relevant information about activities that have been held and the corresponding relationship between the load data of the base station cell and the base station feature data, using the first feature data set to train the model to be trained can enable the model to be trained to better learn the characteristic relationship between the activities and the load data of the base station cell, thereby improving the load prediction accuracy of the prediction model.

[0112] After obtaining the trained prediction model, the load prediction process begins in response to receiving the load prediction request.

[0113] To further illustrate the process of inputting load forecast query information into a large language model for processing and obtaining characteristic data of a target activity, a specific implementation method is provided, including the following steps:

[0114] Step 1: Use the feature extraction tool included in the large language model to parse the natural language input by the user to obtain the identification information of the target activity.

[0115] Among them, the feature extraction tool of the large language model is mainly used to structure the natural language input by the user. For example, the user inputs "Please help me check the load forecast of the A concert on xx / xx / xx." Then the feature extraction tool can extract feature data such as: [xx / xx / xx], [A concert], [load forecast]. Among them, the identification information of the target activity can include the name of the concert, that is, in the above example, the identification information of the target activity is: [A concert] in the feature data. For another example, the identification information of the target activity can also include the date plus the name of the concert. That is, in the above example, the identification information of the target activity is: [xx / xx / xx] + [A concert] in the feature data.

[0116] In some examples, if the user inputs "Please help me check the load forecast of concert A in area xx.", the identification information of the target activity extracted by the feature extraction tool may be, for example: [concert A] or [area xx] + [concert A].

[0117] It is known that, in the process of extracting identification information of target activities, feature extraction tools need to go through at least text preprocessing, keyword candidate generation, keyword weight calculation and ranking, entity recognition and keyword filtering, keyword optimization and calibration, etc. The embodiments of this application will not be described in detail one by one.

[0118] In some examples, the user enters the activity information (recorded as activity_msg) containing the name of the security activity in the form of natural language through the interaction port of the large language model. After the large language model receives the natural language input by the user, it calls the feature extraction tool of the large language model to extract the identification information of the target activity (recorded as S a ). As shown in the following formula (II):

[0119]

[0120] in, Represents the keyword extraction capability of a large language model.

[0121] Step 2: Use the intent understanding tool included in the large language model to match the identification information of the target activity with the second feature data set to obtain feature data of the target activity.

[0122] The second feature data set includes identification information of an activity planned to be held in the communication guarantee area, base station feature data, and relevant information of the activity planned to be held.

[0123] Among them, the intention understanding tool of the large language model is mainly used to accurately capture the purpose and needs behind the user input text. In the embodiment of the present application, the intention understanding tool can be used to match the characteristic data of the target activity according to the identification information of the target activity.

[0124] In some examples, when obtaining the identification information S of the target activity, a Then, the intention understanding tool is called to perform similarity matching between the identification information of the target activity and the identification information of the planned activities in the second data set to obtain feature data of the target activity.

[0125] In the above embodiment, a large language model is used to analyze the natural language input of users to obtain characteristic data for the target activity. Because the large language model can analyze the natural language input of users and accurately predict user intent, it eliminates the need for technical personnel to query the corresponding characteristic data for the target activity, thus simplifying the load forecasting process.

[0126] Specifically, the embodiment of the present application provides a specific implementation method for the process of obtaining the second feature data set, including the following steps:

[0127] Step 1: Acquire a second activity data set related to the communication guarantee area.

[0128] The second activity data set is obtained by crawling relevant information of activities planned to be held in the communication guarantee area based on a crawler program.

[0129] In some practical application scenarios, communication operation and maintenance personnel in a specific area may only want to know the load status of the communication base stations in the area. Therefore, the crawling rules of the crawler program can be used to obtain relevant information about planned activities in a targeted manner.

[0130] In some examples, since the planned dates of some events may not be published on the Internet long in advance, the crawling rules of the crawler program can be used to obtain relevant information about events planned to be held in the communication guarantee area in the future.

[0131] Step 2: Acquire a second base station data set related to base station cells within the communication guarantee area;

[0132] The second base station data set includes base station characteristic data of the base station cell.

[0133] In some examples, the base station feature data can be obtained by accessing the operation and maintenance database through query instructions. The specific method is similar to the above embodiment and will not be described in detail here.

[0134] Step 3: Create a second feature data set based on the second activity data set, the second base station data set, and the second association relationship.

[0135] The second association relationship represents the correspondence between the identification information of the planned activity and the characteristic data of the planned activity. The characteristic data of the planned activity includes: relevant information of the planned activity and base station POI information within the communication support area where the planned activity is located.

[0136] In the above embodiment, a crawler program is used to analyze and obtain information about events held in the communication support area to obtain a second activity dataset, which provides key data for the large language model to match the identification information of the target activity with the second feature dataset. In addition, the crawler program can efficiently obtain massive amounts of data, ensuring the comprehensiveness and accuracy of the data. Furthermore, based on the pre-established second feature dataset and the large language model, the load forecasting process can be simplified. This eliminates the need for technicians to prepare the feature data of the target activity before each load forecast result query. Instead, they can obtain the load forecast result for the communication support area by simply inputting natural language into the interactive port provided by the large language model, reducing the workload of technicians.

[0137] Specifically, the embodiment of the present application provides a specific implementation method for the process of obtaining identification information of a target activity, including the following steps:

[0138] Step 1: Perform similarity matching based on the identification information of the target activity and the identification information of the planned activities included in the second feature data set to obtain a similarity matching result.

[0139] The similarity matching result is used to represent the similarity between the identification information of the target activity and the identification information of the planned activity.

[0140] In some examples, the identification information of the target activity and the identification information of the planned activity may be mapped into dense vectors respectively, and then matched by calculating the distance or approximation between the two vectors.

[0141] Step 2: Obtain feature data of the target activity based on the similarity matching result and the second association relationship.

[0142] In some examples, the similarity matching result may include the identification information of the planned event with the highest speed match value to the identification information of the target event. The characteristic data of the target activity is then obtained based on the correspondence between the identification information of the planned event, the characteristic data of the planned event, and the base station characteristic data. In other words, the characteristic data of the planned event is identical to the characteristic data of the target activity.

[0143] In the above embodiment, since the natural language input by the user contains insufficient feature data, by calculating the similarity matching degree between the identification information of the target activity and the identification information of the planned activity included in the second feature data set, obtaining the feature data of the target activity from the pre-constructed second data set can enable the prediction model to obtain accurate and comprehensive input, thereby improving the load prediction accuracy.

[0144] After obtaining the load forecast results, it is also necessary to formulate a strategy for adjusting the communication load capacity within the communication guarantee area based on the load forecast results.

[0145] Specifically, the embodiment of the present application provides a specific implementation method for a strategy of adjusting the communication load capacity within a communication guarantee area based on load forecast results, including:

[0146] Determine the load conditions of the base station cells within the communication guarantee area during the target event based on the load forecast results; if the base station cells are overloaded during the target event, increase the communication load capacity within the communication guarantee area based on the load forecast results.

[0147] In some examples, improving the communication load capacity within the communication guarantee area may include, for example, adding base stations, expanding the capacity of existing base stations, etc.

[0148] In the above embodiment, adjusting the communication load capacity in the communication guarantee area according to the load forecast result can avoid the problem of poor user experience caused by insufficient communication load capacity in the communication guarantee area, thereby improving user experience.

[0149] The following describes the embodiment of the present application in detail with reference to an application scenario of querying the load forecast of a concert.

[0150] like Figure 3 As shown, the overall flow chart of the method provided in the embodiment of the present application is introduced, and the process includes:

[0151] S301. Use a crawler program to crawl from web pages information related to activities held in the communication support area in the past year.

[0152] Optionally, you can use the crawler program's crawling conditions described in the above example to obtain information about events held in the communications assurance area in the past year.

[0153] S302: Use query instructions to obtain minute-level base station load data and base station characteristic data of activities held in the communication support area from the operation and maintenance database.

[0154] Optionally, the method of setting the query range of the query instruction described in the above example can be used to obtain minute-level load data and base station characteristic data of activities held in the communication guarantee area during the period of the activities.

[0155] S303: Construct a first feature data set based on the relevant information of the held activities, the minute-level base station load data and base station feature data during the holding period, and the first association relationship.

[0156] In some examples, relevant information of an event including a concert, minute-level base station load data and base station feature data during the event, and the first association relationship can be used to construct a first feature data set to enrich the training data of the model.

[0157] S304: Utilize the first feature data set to train and obtain a prediction model.

[0158] In some examples, the method described in the above example of dividing the first feature dataset into a training set, a validation set, and a test set can be used to train the model and adjust its parameters.

[0159] S310. Use a crawler program to crawl relevant information about activities planned to be held in the communication guarantee area within the next year.

[0160] Optionally, you can use the crawler program's crawling conditions described in the above example to obtain information about events planned to be held in the communication assurance area within the next year.

[0161] S311. Obtain base station characteristic data in the communication guarantee area.

[0162] S312: Construct a second feature data set based on the relevant information of the planned activity, the base station feature data, and the second association relationship.

[0163] Optionally, the method of setting the query scope of the query instruction described in the above example can be used to obtain minute-level load data and base station characteristic data during the period of the activities planned to be held in the communication guarantee area.

[0164] S320: Obtain user input.

[0165] In some examples, obtaining user input includes obtaining natural language input by the user to query the load status of base station cells in the communication guarantee area during the target activity.

[0166] S321. Utilize the feature extraction tool of the large language model to obtain identification information of the concert that the user wants to query.

[0167] Optionally, the identification information of the concert can be extracted by using the feature extraction tool of the large language model as described in the above example to extract the identification information of the target activity.

[0168] S322: Using the intention understanding tool of the large language model, the identification information of the concert is matched with the second feature data set to obtain feature information of the concert.

[0169] Optionally, the feature data of the concert can be obtained by matching the feature data of the target activity using the intent understanding tool of the large language model as described in the above example.

[0170] S323: Input the characteristic information of the concert into the prediction model.

[0171] S324. The prediction model outputs minute-level load prediction results for base station cells within the communication guarantee area during the concert.

[0172] Furthermore, this embodiment provides a base station load prediction device, such as Figure 4 As shown, the device includes: an acquisition module 401 and a prediction module 402.

[0173] Acquisition module 401 is configured to, upon receiving a load forecast request, input load forecast query information into a large language model for processing to obtain characteristic data of a target activity; the load forecast query information includes natural language input by a user to query the load conditions of base station cells within a communication support area during the target activity period;

[0174] The prediction module 402 is configured to input the characteristic data of the target activity into the prediction model to perform load prediction and obtain a load prediction result; the load prediction result represents the minute-level load prediction result of the base station cell in the communication guarantee area during the target activity period; the prediction model is obtained based on relevant information of the activities held, minute-level load data of the base station cell related to the held activities during the period, and base station characteristic data.

[0175] In some examples of this embodiment, the apparatus further includes a training module;

[0176] The training module is configured to obtain a first activity data set related to the communication guarantee area; the first activity data set is obtained by crawling relevant information of previously held activities using a crawler program;

[0177] Obtaining a first base station data set related to a base station cell within the communication guarantee area; the first base station data set includes minute-level load data and base station characteristic data of the base station cell during an event held in the communication guarantee area;

[0178] The first activity dataset and the first base station dataset are used to train the model to be trained.

[0179] In some examples of this embodiment, the training module is further configured to obtain a first feature dataset based on the first activity dataset, the first base station dataset, and the first association relationship; the first association relationship represents the correspondence between relevant information of the held activity and minute-level load data and base station feature data of the base station cell during the held activity;

[0180] Training and parameter adjustment of the to-be-trained model according to the first feature data set;

[0181] When the prediction accuracy of the model to be trained meets the requirements, the prediction model is obtained.

[0182] In some examples of this embodiment, the acquisition module 401 is specifically configured to parse the natural language input by the user using a feature extraction tool included in the large language model to obtain identification information of the target activity;

[0183] The intent understanding tool included in the large language model is used to match the identification information of the target activity with the second feature data set to obtain the feature data of the target activity; the second feature data set includes the identification information of the activities planned to be held in the communication guarantee area, the base station feature data and related information of the planned activities.

[0184] In some examples of this embodiment, the acquisition module 401 is further configured to acquire a second activity data set related to the communication assurance area; the second activity data set is obtained by crawling relevant information of activities planned to be held in the communication assurance area based on a crawler program;

[0185] Acquire a second base station data set related to a base station cell within the communication guarantee area; the second base station data set includes base station characteristic data of the base station cell;

[0186] A second feature data set is established based on the second activity data set, the second base station data set, and the second association relationship; the second association relationship represents the correspondence between the identification information of the planned activity and the feature data of the planned activity.

[0187] In some examples of this embodiment, the acquisition module 401 is further configured to perform similarity matching based on the identification information of the target activity and the identification information of the planned activity included in the second feature data set to obtain a similarity matching result; the similarity matching result is used to represent the degree of similarity between the identification information of the target activity and the identification information of the planned activity;

[0188] According to the similarity matching result and the second association relationship, characteristic data of the target activity is obtained.

[0189] In some examples of this embodiment, the apparatus further includes an execution module;

[0190] An execution module is configured to determine the load conditions of base station cells within the communication guarantee area during the target activity period based on the load forecast result;

[0191] If the base station cell is overloaded during the target event, the communication load capacity in the communication guarantee area will be increased based on the load forecast results.

[0192] It should be noted that, for other corresponding descriptions of the functional units involved in the base station load prediction device provided in this embodiment, reference can be made to the description of the base station load prediction method in the above embodiment, which will not be repeated here.

[0193] Based on the base station load prediction method shown in the above embodiment, this embodiment also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the method shown in the above embodiment.

[0194] Based on the method shown in the above embodiment, this embodiment also provides a computer program product, which stores a computer program. When the computer program product is executed by a processor, it implements the method shown in the above embodiment.

[0195] Based on this understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, USB flash drive, mobile hard disk, etc.), and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of the present application.

[0196] Based on the method shown in the above embodiment, and Figure 4 In the virtual device embodiment shown, in order to achieve the above-mentioned purpose, the embodiment of the present application also provides an electronic device, such as a terminal device, etc., which includes a storage medium and a processor; the storage medium is used to store a computer program; the processor is used to execute the computer program to implement the method shown in the above embodiment.

[0197] Optionally, the physical device may further include a user interface, a network interface, a camera, a radio frequency (RF) circuit, a sensor, an audio circuit, a Wi-Fi module, and the like. The user interface may include a display, an input unit such as a keyboard, and the like. The optional user interface may also include a USB interface, a card reader interface, and the like. The network interface may optionally include a standard wired interface, a wireless interface (such as a Wi-Fi interface), and the like.

[0198] Those skilled in the art will understand that the above-mentioned physical device structure provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or a combination of certain components, or different component arrangements.

[0199] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the physical device, supporting the execution of information processing programs and other software and / or programs. The network communication module is used to enable communication between components within the storage medium, as well as with other hardware and software within the physical information processing device.

[0200] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present application can be implemented by means of software plus the necessary general hardware platform, or by hardware. Compared with the current related technologies, by applying the technical solution of this embodiment, it is possible to analyze the natural language input by the user and accurately predict the user's intention, without the need for technical personnel to query the corresponding feature data for the target activity, thereby simplifying the load forecasting process. Furthermore, since the prediction model is based on relevant information of the activities that have been held, minute-level load data of the base station cells related to the activities that have been held during the period, and base station feature data, the prediction model can be used to achieve refined minute-level load forecasting for all time periods.

[0201] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0202] The foregoing is merely a list of specific embodiments of the present application, intended to enable those skilled in the art to understand and implement the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments described herein, but is intended to conform to the broadest scope consistent with the principles and novel features of the present application.

Claims

1. A base station load prediction method, characterized in that: include: When a load forecast request is received, the load forecast query information is input into the large language model for processing to obtain characteristic data of the target activity; The load forecast query information includes natural language input by the user to query the load conditions of base station cells within the communication guarantee area during the period of the target activity; Inputting characteristic data related to the target activity into a prediction model to perform load prediction and obtain a load prediction result; the load prediction result represents a minute-level load prediction result of a base station cell within the communication guarantee area during the target activity period; The prediction model is obtained based on relevant information of the activities held, minute-level load data of the base station cells related to the activities held during the period of holding, and base station characteristic data.

2. The method according to claim 1, characterized in that Before obtaining the load forecast request, the method further includes: Acquire a first activity data set related to the previously held activity; the first activity data set is obtained by crawling relevant information of the previously held activity using a crawler program; Acquire a first base station data set related to a base station cell within the communication guarantee area; the first base station data set includes minute-level load data and base station characteristic data of the base station cell related to the held event; The to-be-trained model is trained using the first activity data set and the first base station data set.

3. The method according to claim 2, characterized in that The training of the to-be-trained model using the first activity data set and the first base station data set includes: A first feature data set is obtained based on the first activity data set, the first base station data set, and the first association relationship; the first association relationship represents the correspondence between the relevant information of the held activity and the minute-level load data and base station feature data of the base station cell during the held activity; Training and adjusting parameters of the to-be-trained model according to the first feature data set; When the prediction accuracy of the model to be trained meets the requirements, the prediction model is obtained.

4. The method according to any one of claims 1 to 3, characterized in that Inputting the load forecast query information into a large language model for processing to obtain characteristic data of a target activity includes: Utilize the feature extraction tool included in the large language model to parse the natural language input by the user to obtain the identification information of the target activity; The intention understanding tool included in the large language model is used to match the identification information of the target activity with the second feature data set to obtain the feature data of the target activity; the second feature data set includes the identification information of the activity planned to be held in the communication guarantee area, the base station feature data and the relevant information of the planned activity.

5. The method according to claim 4, characterized in that The method further comprises: Acquire a second activity data set related to the communication assurance area; the second activity data set is obtained by crawling relevant information of activities planned to be held in the communication assurance area using a crawler program; Acquire a second base station data set related to a base station cell within the communication guarantee area; the second base station data set includes base station characteristic data of the base station cell; A second feature data set is established based on the second activity data set, the second base station data set, and a second association relationship; the second association relationship represents a correspondence between the identification information of the planned activity and the feature data of the planned activity.

6. The method according to claim 5, characterized in that The matching of the identification information of the target activity with the second feature data set to obtain feature data of the target activity includes: performing similarity matching on the identification information of the target activity and the identification information of the planned activity included in the second feature data set to obtain a similarity matching result; the similarity matching result is used to represent the degree of similarity between the identification information of the target activity and the identification information of the planned activity; The characteristic data of the target activity is obtained according to the similarity matching result and the second association relationship.

7. The method according to claim 1, characterized in that After obtaining the load forecast result, the method further includes: Determine the load conditions of the base station cells within the communication guarantee area during the target activity period according to the load forecast result; If the base station cell is overloaded during the target activity, the communication load capacity within the communication guarantee area is increased according to the load prediction result.

8. A base station load prediction device, characterized in that: include: an acquisition module configured to, upon receiving a load forecast request, input the load forecast query information into a large language model for processing to obtain characteristic data of a target activity; the load forecast query information comprising natural language input by a user inquiring about load conditions of base station cells within a communication support area during the period of the target activity; A prediction module is configured to input characteristic data related to the target activity into a prediction model to perform load prediction and obtain a load prediction result; the load prediction result represents a minute-level load prediction result of a base station cell within the communication guarantee area during the target activity period; The prediction model is obtained based on relevant information of the activities held, minute-level load data of the base station cells related to the activities held during the period of the activities, and base station characteristic data.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

10. An electronic device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

11. A computer program product having a computer program stored thereon, characterized in that: When the computer program product is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images