Communication link prediction method, system and device and electronic equipment

Abstract

The invention discloses a communication link prediction method, system and device and electronic equipment. The method comprises the following steps: acquiring communication link data; processing the communication link data through a first model to obtain a first prediction result, the first model being used for capturing trend change and seasonal characteristics in the communication link data, and the first prediction result being an initial utilization rate prediction value of the communication link; determining a residual value between the first prediction result and the communication link data; the residual value is processed through a second model, a second prediction result is obtained, the second model is used for learning a time sequence mode in the residual value, and the second prediction result is used for representing the future change trend of the residual value; and fusing the first prediction result and the second prediction result to obtain a target prediction result of the communication link. According to the method and the device, the technical problem of relatively low real-time monitoring and prediction efficiency of the utilization rate of the communication link caused by frequent traffic congestion phenomena in a large-scale network environment is solved.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a method, system, device and electronic device for predicting a communication link. Background Art

[0002] Modern communications networks, especially IP bearer networks, face increasingly severe traffic management challenges. With the surge in user numbers and the diversification of data service demands, network link utilization is increasing, leading to frequent traffic congestion. Traffic congestion not only degrades network performance, such as increasing latency and reducing throughput, but can also lead to resource waste, network jitter, and service interruptions, severely impacting user network experience and business continuity.

[0003] Current network operations and maintenance practices still rely on manual querying to monitor communication link utilization, a method that is extremely inefficient for large-scale networks. The sheer number of devices and complex network architecture make manually checking the link status of each core node one by one time-consuming and error-prone, failing to meet the demands for rapid response and real-time monitoring. This inefficient monitoring method, especially during peak network times, prevents timely detection and resolution of traffic bottlenecks.

[0004] Furthermore, related technologies typically limit link utilization queries to a single device or link, lacking a comprehensive monitoring solution for the overall link utilization of prefecture-level core networks. This significantly limits the ability to detect and prevent network failures early. Furthermore, due to traditional query methods, network monitoring can often only be performed on a PC, making it unsuitable for modern mobile office environments. Furthermore, data communication equipment has strict login permissions, making them inaccessible to non-maintenance personnel. This complicates cross-disciplinary collaboration and reduces flexibility in problem resolution.

[0005] Furthermore, network traffic exhibits significant seasonal and cyclical characteristics. For example, traffic distribution varies significantly across regions during the Spring Festival and National Day holidays, and traffic patterns differ significantly between weekdays and weekends. However, existing communication link monitoring systems do not fully utilize this cyclical data for intelligent prediction, making it impossible to identify potential network congestion points in advance and take preventative measures.

[0006] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0007] The embodiments of the present application provide a communication link prediction method, system, device and electronic device to at least solve the technical problem of low efficiency in real-time monitoring and prediction of communication link utilization caused by frequent traffic congestion in a large-scale network environment.

[0008] According to one aspect of an embodiment of the present application, a method for predicting a communication link is provided, comprising: acquiring communication link data; processing the communication link data through a first model to obtain a first prediction result, wherein the first model is used to capture trend changes and seasonal characteristics in the communication link data, and the first prediction result is an initial utilization prediction value of the communication link; determining a residual value between the first prediction result and the communication link data; processing the residual value through a second model to obtain a second prediction result, wherein the second model is used to learn a time series pattern in the residual value, and the second prediction result is used to represent a future change trend of the residual value; and fusing the first prediction result and the second prediction result to obtain a target prediction result of the communication link.

[0009] Optionally, after determining the residual value between the first prediction result and the communication link data, the method further includes: optimizing the residual value through a third model to obtain a residual optimization value, wherein the third model is used to enhance the correlation between different eigenvectors in the residual value.

[0010] Optionally, the third model is trained in the following manner: obtaining historical communication link data; determining a first eigenvector and a second eigenvector based on the historical communication link data, wherein the first eigenvector is used to represent a data set of historical communication link data in the same periodic period, and the second eigenvector is used to represent a data set of historical communication link data in a preset time period whose feature similarity is greater than a preset threshold; modulating the first eigenvector and the second eigenvector according to a preset feature scale to obtain a third eigenvector corresponding to the first eigenvector, and obtaining a fourth eigenvector corresponding to the second eigenvector; training the initial optimization model based on the third eigenvector and the fourth eigenvector to obtain the third model.

[0011] Optionally, before training the initial optimization model based on the third eigenvector and the fourth eigenvector, the method also includes: interpolating the third eigenvector and the fourth eigenvector to obtain a fifth eigenvector corresponding to the third eigenvector, and obtaining a sixth eigenvector corresponding to the fourth eigenvector; determining a first compensation factor based on the third eigenvector and the fifth eigenvector, wherein the first compensation factor is used to represent the degree of mapping deviation between the first eigenvector after modulation processing and after interpolation processing; determining a second compensation factor based on the fourth eigenvector and the sixth eigenvector, wherein the second compensation factor is used to represent the degree of mapping deviation between the second eigenvector after modulation processing and after interpolation processing; and correcting the third eigenvector and the fourth eigenvector based on the first compensation factor and the second compensation factor.

[0012] Optionally, the second model is trained in the following manner: predicting historical communication link data through the first model to obtain historical prediction results; determining historical residual values between the historical prediction results and the historical communication link data; optimizing the historical residual values through the third model to obtain historical residual optimization values; training the initial residual prediction model based on the historical residual optimization values to obtain the second model.

[0013] Optionally, the first model is trained in the following manner: obtaining time series characteristics of historical communication link data, wherein the time series characteristics include trend changes, seasonal characteristics and periodic characteristics of the historical communication link data; determining the autocorrelation function and partial autocorrelation function of the historical communication link data based on the time series characteristics, wherein the autocorrelation function is used to represent the data correlation of the historical communication link data at different time points, and the partial autocorrelation function is used to represent the data correlation of the historical communication link data at a specific lag time point; determining the seasonal parameters of the initial prediction model based on the autocorrelation function and the partial autocorrelation function, wherein the seasonal parameters are used to reflect the periodic behavior of the historical communication link data; training the initial prediction model based on the seasonal parameters and the historical communication link data, and optimizing the model parameters of the initial prediction model to obtain the first model.

[0014] Optionally, after obtaining the time series characteristics of the historical communication link data, the method further includes: performing a unit root test on the time series characteristics to obtain a test result, wherein the test result is used to indicate the degree of stationarity of the time series characteristics; when the test result meets a preset condition, determining that the time series characteristics are a stationary sequence; when the test result does not meet the preset condition, determining that the time series characteristics are an unstationary sequence, and performing differential processing on the unstationary sequence until the unstationary sequence reaches stationarity.

[0015] According to another aspect of an embodiment of the present application, a communication link prediction system is also provided, including: a mobile terminal, a data processing platform, an acquisition platform and a performance database, wherein the mobile terminal is used to initiate a query request for the communication link; the acquisition platform is used to collect communication link data of the communication equipment according to the query request, and send the communication link data to the performance database for storage, and send the communication link data to the data processing platform for processing; the performance database is used to store the communication link data; the data processing platform is used to process the communication link data and return the data processing result to the mobile terminal; the data processing platform is also used to detect the communication equipment according to a preset inspection frequency; the data processing platform is also used to obtain the communication link data from the performance database, and execute the communication link prediction method according to any one of claims 1 to 7.

[0016] According to another aspect of the embodiments of the present application, a communication link prediction device is also provided, including: an acquisition module for acquiring communication link data; a first prediction module for processing the communication link data through a first model to obtain a first prediction result, wherein the first model is used to capture trend changes and seasonal characteristics in the communication link data, and the first prediction result is an initial utilization prediction value of the communication link; a determination module for determining a residual value between the first prediction result and the communication link data; a second prediction module for processing the residual value through a second model to obtain a second prediction result, wherein the second model is used to learn a time series pattern in the residual value, and the second prediction result is used to represent a future change trend of the residual value; and a fusion module for fusing the first prediction result and the second prediction result to obtain a target prediction result of the communication link.

[0017] According to another aspect of the embodiments of the present application, an electronic device is provided, including: a memory and a processor, wherein the memory is used to store program instructions; the processor is connected to the memory and is used to execute the prediction method for implementing the above-mentioned communication link.

[0018] According to another aspect of the embodiments of the present application, a non-volatile storage medium is provided. The non-volatile storage medium includes a stored computer program, wherein the device where the non-volatile storage medium is located executes the above-mentioned communication link prediction method by running the computer program.

[0019] According to another aspect of the embodiments of the present application, a computer program product is provided, including computer instructions, which implement the above-mentioned communication link prediction method when executed by a processor.

[0020] In an embodiment of the present application, communication link data is acquired; the communication link data is processed by a first model to obtain a first prediction result, wherein the first model is used to capture trend changes and seasonal characteristics in the communication link data, and the first prediction result is an initial utilization prediction value of the communication link; a residual value between the first prediction result and the communication link data is determined; the residual value is processed by a second model to obtain a second prediction result, wherein the second model is used to learn a time series pattern in the residual value, and the second prediction result is used to represent a future change trend of the residual value; the first prediction result and the second prediction result are integrated to obtain a target prediction result of the communication link, thereby achieving the purpose of accurately predicting the future utilization of the communication link, thereby realizing the technical effect of optimizing network resources, preventing network congestion and ensuring network stability, and further solving the technical problem of low efficiency in real-time monitoring and prediction of communication link utilization caused by frequent traffic congestion in a large-scale network environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0022] Figure 1 is a hardware structure diagram of a computer terminal for implementing a communication link prediction method according to an embodiment of the present application;

[0023] Figure 2 is a flowchart of a method for predicting a communication link according to an embodiment of the present application;

[0024] Figure 3 is a flowchart of another communication link prediction method according to an embodiment of the present application;

[0025] Figure 4 is a structural diagram of a communication link prediction system according to an embodiment of the present application;

[0026] Figure 5 It is a structural diagram of a communication link prediction device according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0028] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in a sequence other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0029] First, some nouns or terms that appear in the process of explaining the embodiments of this application are subject to the following explanations:

[0030] SARIMA (Seasonal Autoregressive Integrated Moving Average): Used to forecast data with seasonal patterns. The SARIMA model combines autoregressive (AR), differencing (I), moving average (MA), and seasonal components to effectively handle trends and seasonal variations in the data.

[0031] xLSTM network: An extended LSTM network (Long Short-Term Memory) used to process long-distance dependencies in sequence data, especially for residual prediction based on SARIMA forecasts to improve forecast accuracy.

[0032] IP bearer network: A network that uses IP technology as its primary transmission method and can carry a variety of services such as voice, video, and data. It is the core part of modern communication networks.

[0033] Internet Protocol Radio Access Network (IPRAN): An IP-based network architecture that connects cellular base stations and core networks, providing efficient data transmission and flexible network management. A key component of modern mobile communications networks, IPRAN supports multiple generations of wireless technologies, including 3G, 4G, and 5G. It can carry a variety of services, including voice, video, and data. It offers high bandwidth, low latency, and high reliability, making it a crucial technology for the convergence of wireless and wired networks.

[0034] IDC (Internet Data Center): A place that provides services such as hosting, resource leasing, system maintenance, information consulting, and content provision. It is an important facility for enterprise data storage and processing.

[0035] ACF (Autocorrelation Function): A function that measures the correlation between a time series and its past values. It calculates the correlation coefficient of the series at different time lags.

[0036] PACF (Partial Autocorrelation Function): Similar to ACF, but it removes the influence of intermediate variables in the series and focuses only on the direct correlation between the current value and the past value of a specific lag period.

[0037] In order to solve the problem of low communication link detection efficiency in the related art, the embodiment of the present application provides a communication link prediction method, which can be run on Figure 1 Among the computer terminals shown, the computer terminal will be described below.

[0038] The communication link prediction method embodiment provided in the embodiment of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 FIG1 shows a hardware structure block diagram of a computer terminal for implementing a method for predicting a communication link. Figure 1 As shown, the computer terminal 10 may include one or more (illustrated by 102a, 102b, ..., 102n in the figure) processors (the processor may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission module 106 for communication functions connected via a wired and / or wireless network. In addition, it may also include: a display, a keyboard, a cursor control device, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, and a BUS bus. It will be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components than shown, or with Figure 1 Different configurations shown.

[0039] It should be noted that the one or more processors and / or other data processing circuits described above may generally be referred to herein as "data processing circuitry." The data processing circuitry may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuitry may be a single, independent processing module, or may be incorporated in whole or in part into any of the other components of the computer terminal 10. As described in the embodiments of the present application, the data processing circuitry serves as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0040] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the communication link prediction method in the embodiment of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, implementing the above-mentioned communication link prediction method. The memory 104 may include a high-speed random access memory and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely located relative to the processor, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0041] The transmission module 106 is configured to receive or transmit data via a network. A specific example of the aforementioned network may include a wireless network provided by the communications provider of the computer terminal 10. In one embodiment, the transmission module 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In another embodiment, the transmission module 106 may be a radio frequency (RF) module, which is configured to communicate with the Internet wirelessly.

[0042] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 .

[0043] It should be noted that, in some optional embodiments, the above Figure 1 The computer terminal shown may include hardware elements (including circuits), software elements (including computer code stored on a computer-readable medium), or a combination of hardware elements and software elements. Figure 1 This is merely one example of a particular embodiment and is intended to illustrate the types of components that may be present in the computer terminal described above.

[0044] In the above-mentioned operating environment, an embodiment of the present application provides an embodiment of a method for predicting a communication link. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0045] Figure 2 is a flow chart of a method for predicting a communication link according to an embodiment of the present application, such as Figure 2As shown, the method includes the following steps:

[0046] Step S202: Acquire communication link data.

[0047] In the above step S202, communication link data can be obtained from a performance database (a data warehouse for storing network performance data, alarm and obstacle data, service quality data and other data), wherein the communication link data may include the real-time communication link utilization and historical communication link utilization of network core equipment (such as CR, ASBR, IPRAN CE, VoLTE CE, 5G CoreCE, IDC export equipment and other communication equipment) in each region and city.

[0048] Step S204: Process the communication link data through the first model to obtain a first prediction result, wherein the first model is used to capture trend changes and seasonal characteristics in the communication link data, and the first prediction result is an initial utilization prediction value of the communication link.

[0049] In step S204, the SARIMA model is used as the first model for data processing. Using the SARIMA model, a preliminary prediction of the link utilization of each communication device is obtained, yielding a first prediction result, namely, the initial link utilization prediction value. The SARIMA model primarily utilizes real-time link utilization data, historical link utilization data, and its seasonal variation patterns to predict link utilization over a period of time, providing a foundation for subsequent predictions.

[0050] Step S206: Determine a residual value between the first prediction result and the communication link data.

[0051] In step S206, after obtaining the first prediction result, it is necessary to compare it with the actual communication link data and calculate the difference between the two, namely the residual value. This step is key to evaluating the prediction accuracy of the SARIMA model and is also a prerequisite for introducing the second model for optimization. The residual value reveals the parts that the SARIMA model failed to accurately predict. These errors often include complex fluctuations and randomness that the model fails to capture.

[0052] Optionally, after determining the residual value between the first prediction result and the communication link data, the above method also includes: optimizing the residual value through a third model to obtain a residual optimization value, wherein the third model is used to enhance the correlation between different eigenvectors in the residual value.

[0053] In an embodiment of the present application, a periodic characteristic model can also be introduced as a third model to optimize the residual values obtained by SARIMA prediction to enhance the correlation between different eigenvectors in the residual data and provide a more optimized input for the xLSTM network, thereby improving the accuracy and stability of the entire prediction system in predicting the communication link utilization.

[0054] Step S208, processing the residual value through the second model to obtain a second prediction result, wherein the second model is used to learn the time series pattern in the residual value, and the second prediction result is used to represent the future change trend of the residual value.

[0055] In step S208, an xLSTM network may be used as the second model to perform deep learning processing on the residual optimization value obtained in step S206. The xLSTM network is used to learn the time series pattern in the residual value to predict the future trend of the residual value, thereby obtaining a second prediction result.

[0056] Step S210: The first prediction result and the second prediction result are integrated to obtain a target prediction result of the communication link.

[0057] In step S210, the two sets of prediction results obtained above are comprehensively considered and combined through an appropriate fusion strategy (such as weighted averaging or other machine learning methods) to obtain the final communication link target prediction result. This target prediction result takes into account both long-term trends and seasonal factors, as well as the possibility of complex short-term fluctuations, thus achieving higher prediction accuracy. In this way, the prediction system can more accurately predict the future utilization of communication links, helping to identify potential network bottlenecks in advance and providing strong data support for network management.

[0058] Through the above steps S202 to S210, the purpose of accurately predicting the future utilization of the communication link is achieved, thereby realizing the technical effects of optimizing network resources, preventing network congestion and ensuring network stability, and further solving the technical problem of low efficiency in real-time monitoring and prediction of communication link utilization caused by frequent traffic congestion in large-scale network environments.

[0059] Figure 3 is a flow chart of another communication link prediction method according to an embodiment of the present application, such as Figure 3The figure shows a clear depiction of the communication link prediction process, intuitively illustrating the entire process from data acquisition to prediction result generation. Specifically, starting with manually triggered prediction, real-time and historical communication link data are first obtained from the performance database. A preliminary analysis is then performed using the SARIMA model to obtain a first prediction result that accounts for long-term trends and seasonal fluctuations. Next, the residual value between the first prediction result and the actual link data is calculated. This data is input into a periodic feature model. Through feature vector modulation and mapping deviation compensation, the periodic characteristics in the data and the interactions between different features are captured and optimized. Finally, the processed residual data is fed into the xLSTM network to predict the future trend of the residual value, resulting in a second prediction result. By weighted fusion of the first and second prediction results, the target prediction value for the communication link is ultimately derived, achieving accurate time-series prediction of link utilization. Among them, the SARIMA model can be trained through steps such as data preprocessing, stationarity test, seasonal parameter determination and self-parameter selection; the periodic characteristic model can be trained through steps such as data acquisition, eigenvector division, eigenvector modulation and mapping deviation compensation; the xLSTM network can be trained through data acquisition and using the residual data after preliminary prediction of the SARIMA model and optimization of the periodic characteristic model.

[0060] In short, Figure 3 It presents a structured time series prediction process, from data acquisition to prediction result generation, covering the collaborative work of SARIMA model, periodic characteristic model and xLSTM network. Through the combined optimization of multiple models, it effectively improves the accuracy and reliability of prediction, providing strong technical support for network resource management and fault prevention.

[0061] The following describes in detail the model training process of the SARIMA model (first model), the periodic feature model (third model), and the xLSTM network (second model).

[0062] Optionally, the first model is trained in the following manner: obtaining time series characteristics of historical communication link data, wherein the time series characteristics include trend changes, seasonal characteristics and periodic characteristics of the historical communication link data; determining the autocorrelation function and partial autocorrelation function of the historical communication link data based on the time series characteristics, wherein the autocorrelation function is used to represent the data correlation of the historical communication link data at different time points, and the partial autocorrelation function is used to represent the data correlation of the historical communication link data at a specific lag time point; determining the seasonal parameters of the initial prediction model based on the autocorrelation function and the partial autocorrelation function, wherein the seasonal parameters are used to reflect the periodic behavior of the historical communication link data; training the initial prediction model based on the seasonal parameters and the historical communication link data, and optimizing the model parameters of the initial prediction model to obtain the first model.

[0063] After obtaining the time series characteristics of the historical communication link data, the method further includes: performing a unit root test on the time series characteristics to obtain a test result, wherein the test result is used to indicate the degree of stationarity of the time series characteristics; when the test result meets a preset condition, determining that the time series characteristics are a stationary sequence; when the test result does not meet the preset condition, determining that the time series characteristics are an unstationary sequence, and performing differential processing on the unstationary sequence until the unstationary sequence reaches stationarity.

[0064] In the embodiment of the present application, the training process of the SARIMA model reflects the statistical principles of time series data prediction. Through careful parameter adjustment and data preprocessing, the model can learn and simulate complex trends and periodic patterns in the data. This process ensures the accuracy and reliability of the model prediction, provides a high-quality prediction basis and residual data for the subsequent periodic feature model and xLSTM network, and thus achieves more accurate communication link utilization prediction. The specific steps can be as follows:

[0065] S1: Obtain the time series characteristics of historical communication link data.

[0066] First, historical communication link utilization data is obtained from the performance database. This data is loaded into a DataFrame to ensure that the data is arranged in chronological order, facilitating subsequent analysis and processing. Next, a time series graph is plotted to visually observe long-term trends, seasonal changes, and cyclical patterns in the data, and to extract time series features.

[0067] S2: Unit root test and stationarity treatment.

[0068] After obtaining time series characteristics, a unit root test (ADF test) is performed to determine whether the data is stationary. The ADF test focuses on the unit root problem in the series, that is, whether the series exhibits random walk characteristics, which can lead to non-stationarity. The basic idea of the test is to use a regression model to test whether the autoregressive coefficient of the series is equal to 1, that is, whether the series has a unit root. If the series has a unit root, it is likely to exhibit long-term non-stationary characteristics, such as trend or seasonal fluctuation patterns.

[0069] Specifically, the results of the ADF test include the ADF statistic and the corresponding p-value. If the p-value is less than the preset significance level (usually 0.05), the null hypothesis of the presence of a unit root can be rejected, which means that the data is stationary or can be considered stationary within a certain error range. If the p-value is greater than this significance level, the null hypothesis cannot be rejected and the data is considered non-stationary, possibly containing trend or seasonal factors.

[0070] Furthermore, if the ADF test indicates that the series is stationary, then the data can be used directly for parameter selection and model training of the SARIMA model without further processing. If the series is determined to be non-stationary, then differencing is required. The purpose of differencing is to remove trend or seasonality from the series and make it stationary. The number of differencing steps depends on the stationarity of the series after the ADF test. Typically, differencing and ADF testing are repeated until the series becomes stationary.

[0071] S3: Determine seasonal parameters through the autocorrelation function and partial autocorrelation function of historical communication link data.

[0072] The autocorrelation function (ACF) and partial autocorrelation function (PACF) are used to analyze the correlation between data points. The ACF measures the correlation between time series data at different points in time. By calculating the correlation coefficient for each lag, it reveals the linear correlation between the current data point and past data points. Unlike the ACF, the PACF removes the influence of intermediate lags and focuses on the direct correlation between the current data point and past data points at a specific lag.

[0073] Through the ACF and PACF graphs, we can identify the lag order of non-zero correlation coefficients, and thus estimate the non-seasonal parameters (p, d, q) and seasonal parameters (P, D, Q, S) in the model. Specifically:

[0074] p (autoregressive order): The order of an autoregressive term refers to the influence of the values at the previous p time points on the value at the current time point. For example, if p = 1, the model will consider the influence of the value at the time point before the current time point; if p = 2, the model will consider the influence of the values at the previous two time points, and so on.

[0075] d (order of differencing): The order of differencing refers to how many times the data is differencing to eliminate trends and make the time series stationary. d = 1 means performing a single differencing operation, i.e., calculating the item-by-item difference of the time series; d = 2 means performing a double differencing operation, i.e., calculating the difference of item-by-item differences of the time series, and so on.

[0076] q (moving average order): The order of the moving average term refers to the influence of the first q random error terms in the model on the current time point. For example, if q = 1, the model will consider the influence of the forecast error at the previous time point on the current forecast; if q = 2, the model will consider the influence of the forecast errors at the previous time point, and so on.

[0077] P (order of seasonal autoregressive term): Similar to the non-seasonal autoregressive parameter p, P refers to the order of seasonal autoregressive term, but it considers the influence of the first P time points relative to the length of the seasonal cycle S.

[0078] D (order of seasonal differencing): D represents the number of times seasonal differencing is performed to remove seasonal trends from the data. Seasonal differencing is the difference between a value in the time series and its value S time units ago.

[0079] Q (order of the seasonal moving average term): Q is similar to the non-seasonal q parameter, but it considers the effect of the first Q seasonal errors associated with the length of the seasonal cycle S.

[0080] S (Seasonal Cycle Length): The seasonal cycle length refers to the number of time units that the seasonal pattern repeats in the time series. For example, for annual seasonality, S = 12 (months); for quarterly seasonality, S = 4 (quarters).

[0081] S4: Automated parameter selection.

[0082] Use the auto_arima function (from the pmdarima library) to automatically select the optimal parameter combination. Auto_arima determines a reasonable model configuration by comparing the model performance (such as AIC or BIC information criteria) under different parameter combinations.

[0083] S5: Model training.

[0084] Use the SARIMAX (Seasonal Autoregressive Integrated Moving Average Extended Model) function in the statsmodels library to train the SARIMA model. Pass the determined parameters (p, d, q, P, D, Q, S) to the model to begin model training. Model training involves minimizing the forecast error. This can be achieved by optimizing the coefficients of the autoregressive, moving average, and seasonal autoregressive and moving average terms.

[0085] During training, the model automatically adjusts its parameters to achieve the best fit. Once the optimal parameter combination is found, the model parameters are fixed, providing a stable and reliable model for subsequent prediction tasks.

[0086] Optionally, the second model is trained in the following manner: predicting historical communication link data through the first model to obtain historical prediction results; determining historical residual values between the historical prediction results and the historical communication link data; optimizing the historical residual values through the third model to obtain historical residual optimization values; training the initial residual prediction model based on the historical residual optimization values to obtain the second model.

[0087] In the embodiment of the present application, the training of the xLSTM network relies on the synergy of the periodic feature model of the SARIMA model. The SARIMA model provides a preliminary forecast under seasonal influences, and the periodic feature model optimizes the residual of this forecast, especially focusing on periodic characteristics. By learning the pattern of historical residual optimization values, an xLSTM network is finally formed that can accurately predict the link utilization time series. The specific steps can be as follows:

[0088] S1: Preliminary forecasts from the SARIMA model.

[0089] First, we used the SARIMA model to predict historical link utilization data and generate preliminary historical forecast results. The SARIMA model captures the seasonality and long-term trends of the data, providing a basic forecasting starting point for the xLSTM network and reducing the burden of network training.

[0090] S2: Calculate historical residual values.

[0091] The SARIMA model's predictions are compared with actual historical link utilization data to calculate historical residuals. These residuals reflect data variations that the SARIMA model fails to fully capture, including complex fluctuations and subtle patterns in the data—details that the SARIMA model may overlook.

[0092] S3: Periodic characteristic model optimization residuals.

[0093] The calculated historical residual values are further optimized using a periodic feature model. This model accounts for the cyclical nature of the data, such as specific patterns within an annual or weekly period. This is crucial when processing link utilization data in telecommunications networks, which is subject to periodic influences such as holidays and weekdays. The optimization process involves modulating the feature vectors, compensating for mapping deviations, and performing multidimensional significant interactions between features. Ultimately, the optimized historical residual values are generated, significantly improving the accuracy and stability of the forecast.

[0094] S4: xLSTM network training and fixing.

[0095] Finally, the historical residual values optimized by the periodic feature model are used as training data to train the initial residual prediction model, resulting in an xLSTM network. By learning these optimized residual values, the xLSTM network can more accurately predict the residual of future link utilization—the difference between the model prediction and actual data. The trained xLSTM network fully considers the long-term dependencies between link cycle data and network nodes, enabling precise time-series prediction of link utilization in complex real-world networks, providing strong technical support for network operations and maintenance.

[0096] Optionally, the third model is trained in the following manner: obtaining historical communication link data; determining a first eigenvector and a second eigenvector based on the historical communication link data, wherein the first eigenvector is used to represent a data set of historical communication link data in the same periodic period, and the second eigenvector is used to represent a data set of historical communication link data in a preset time period whose feature similarity is greater than a preset threshold; modulating the first eigenvector and the second eigenvector according to a preset feature scale to obtain a third eigenvector corresponding to the first eigenvector, and obtaining a fourth eigenvector corresponding to the second eigenvector; training the initial optimization model based on the third eigenvector and the fourth eigenvector to obtain the third model.

[0097] Among them, before the initial optimization model is trained based on the third eigenvector and the fourth eigenvector, it also includes: interpolating the third eigenvector and the fourth eigenvector to obtain a fifth eigenvector corresponding to the third eigenvector, and obtaining a sixth eigenvector corresponding to the fourth eigenvector; determining a first compensation factor based on the third eigenvector and the fifth eigenvector, wherein the first compensation factor is used to represent the degree of mapping deviation between the first eigenvector after modulation processing and after interpolation processing; determining a second compensation factor based on the fourth eigenvector and the sixth eigenvector, wherein the second compensation factor is used to represent the degree of mapping deviation between the second eigenvector after modulation processing and after interpolation processing; and correcting the third eigenvector and the fourth eigenvector based on the first compensation factor and the second compensation factor.

[0098] In the embodiment of the present application, the periodic feature model plays a core role in the entire training process. It not only focuses on the processing of periodic data, but also deeply analyzes the information distortion in the data processing process. Through a series of compensation mechanisms, it enhances the efficiency and relevance of information transmission between features. Especially in the calculation of modulation, interpolation and compensation factors, the periodic feature model ensures that even in the data preprocessing stage, the integrity and subtle differences of the original data can be preserved as much as possible, providing high-quality input for the training of the xLSTM network. The specific steps can be as follows:

[0099] S1: Data acquisition and feature vector determination.

[0100] First, two different sets of feature vectors are extracted from the historical communication link data:

[0101] The first eigenvector focuses on data with the same historical period (the same annual span), such as link utilization data at the same time point (such as Spring Festival and National Day), which is used to capture the periodic characteristics within the year or week.

[0102] The second eigenvector is a data set whose link utilization feature similarity is greater than a preset threshold within a preset time period (such as the past few months). It is used to capture the commonalities and trends of the data within a specific time window and help the model understand short-term dynamic changes.

[0103] S2: Eigenvector modulation.

[0104] The first and second eigenvectors are modulated and converted into vectors with the same preset feature scale, namely the third and fourth eigenvectors, through the co-spatial mapper of the fully connected layer. This ensures that feature vectors from different sources have a unified representation framework before entering the model training, facilitating effective feature interaction and learning in the model.

[0105] The fully connected layer learns nonlinear mapping relationships, allowing vectors that might have existed in different feature spaces to be converted into the same space, facilitating subsequent feature interaction operations. Furthermore, this step not only unifies feature scales but also reduces noise in the raw data and enhances useful signals, providing a solid foundation for subsequent processing.

[0106] S3: Mapping deviation compensation.

[0107] Considering that some of the original feature information may be lost during feature space mapping, this information loss may reduce the effectiveness of feature interaction analysis because some subtle differences that are important to the task may not be captured. At the same time, if there is a certain degree of correlation or covariation between different features, such intrinsic connections may be destroyed after undergoing independent spatial transformations, making subsequent attempts to explore the interaction patterns between them more difficult. Therefore, after spatial mapping, feature mapping deviation compensation can also be performed on the modulated third and fourth eigenvectors.

[0108] Specifically, after modulation processing, interpolation is performed to obtain the fifth and sixth eigenvectors, which have the same characteristic scale as the third and fourth eigenvectors. Interpolation helps fill gaps in the time series, improves the continuity and integrity of the data, and further reveals the relationship between features.

[0109] Next, the feature mapping deviation semantic metric network is used to calculate the mapping deviation semantic compensation factors between the third and fifth eigenvectors, and between the fourth and sixth eigenvectors, namely the first compensation factor and the second compensation factor. These compensation factors quantify the information deviation that may occur during the modulation and interpolation process and are used for subsequent feature vector correction.

[0110] S4: Feature vector correction and model training.

[0111] The first and second compensation factors are used to modify the third and fourth eigenvectors, ensuring that the processed eigenvectors are more accurate and consistent in expressing the semantic information of the data. The modified eigenvectors better reflect the complete information of the original data while maintaining a uniform scale after processing.

[0112] Finally, the initial optimization model is trained using the modified third and fourth eigenvectors. By optimizing the model's parameters, it can effectively process and understand the information in these eigenvectors, thereby improving the model's predictive ability for historical communication link data. The specific training process can be as follows:

[0113] First, the third eigenvector and the fourth eigenvector are fed into the initial optimization model as input.

[0114] Secondly, within the model, the deviations of the eigenvectors after modulation are calculated, and then these deviations are compensated according to the first compensation factor and the second compensation factor.

[0115] Furthermore, the model parameters are adjusted through a backpropagation algorithm, enabling the model to optimize the residual prediction based on the input feature vector. The training process involves multiple iterations, and in each iteration, the model updates its parameters based on the prediction error, hoping to reduce the error and improve the prediction performance.

[0116] During the training process, the model's predictive performance is continuously evaluated to ensure that it can accurately predict the historical residual optimization value. After sufficient training, when the model's performance reaches a satisfactory level, the model parameters are fixed to form the final third model for future communication link utilization prediction.

[0117] According to an embodiment of the present application, a communication link prediction system is provided. It should be noted that the communication link prediction system of the embodiment of the present application can be used to execute the communication link prediction method provided in the embodiment of the present application. The following introduces the communication link prediction system provided in the embodiment of the present application.

[0118] Figure 4 This is a structural diagram of a communication link prediction system provided according to an embodiment of the present application. Figure 4 As shown, the system includes: a mobile terminal 40, a data processing platform 42, a collection platform 44 and a performance database 46, wherein:

[0119] The mobile terminal is used to initiate a query request for the communication link.

[0120] In this embodiment of the application, a trigger query mechanism is integrated into the mobile terminal. Users only need to enter a specific command (such as "one-click link census" plus the name of a city) on the mobile terminal to quickly trigger a system query request. This design reflects the consideration of mobile office needs in this application and simplifies the link utilization query process. Regardless of the user's location, link status can be obtained instantly, significantly improving operation and maintenance efficiency.

[0121] The collection platform is used to collect the communication link data of the communication device 48 according to the query request, and send the communication link data to the performance database for storage, and send the communication link data to the data processing platform for processing.

[0122] In an embodiment of the present application, when a query request is received from a mobile terminal, the data processing platform first accesses the core equipment in the metropolitan area network, IPRAN, IDC and new metropolitan area network of each city, and initiates a query instruction to the collection platform. The collection platform automatically starts the data collection process to collect the communication link data of each device.

[0123] After the collection is completed, the communication link data is sent to the performance database for storage, providing a data source for historical data analysis; on the other hand, it is transmitted to the data processing platform in real time for processing to ensure the real-time and accuracy of the information and meet the needs of rapid response.

[0124] Performance database, used to store communication link data.

[0125] In this embodiment, the performance database stores link utilization data collected by the acquisition platform. This data forms the cornerstone of the periodic feature model and xLSTM network training, providing the historical dataset required for time series analysis. The database's efficient storage and retrieval capabilities support the subsequent execution of the prediction algorithm, ensuring data integrity and accessibility.

[0126] The data processing platform is used to process the communication link data and return the data processing results to the mobile terminal; the data processing platform is also used to detect the communication equipment according to the preset inspection frequency; the data processing platform is also used to obtain the communication link data from the performance database and execute the prediction method for realizing the above-mentioned communication link.

[0127] In the embodiment of this application, the data processing platform undertakes the core tasks of data processing and prediction, including real-time analysis of link utilization, automated inspections, and the execution of prediction algorithms. By integrating the SARIMA model, the periodic characteristic model, and the xLSTM network, the data processing platform realizes the automated monitoring and prediction of link utilization. In particular, the automatic inspection during the evening peak period can promptly detect link anomalies, while the prediction algorithm proactively analyzes link utilization trends and provides accurate early warning information for network operation and maintenance. In addition, the platform is also responsible for returning the processed data results to the mobile terminal, ensuring that users can quickly obtain the analysis results.

[0128] Among them, the daily inspection mechanism of the communication link is to use the data processing platform to automatically perform standardized network monitoring tasks, which focuses on conducting thorough link utilization checks on the core network equipment in various cities during the evening traffic peak. Through automated processes, the system can efficiently collect and analyze data and automatically identify abnormal situations where the link utilization exceeds 80%. The selection of this threshold is intended to provide early warning of potential network congestion risks and ensure the rational use of network resources and the stability of user experience. The inspection results are compiled into detailed reports and pushed to network maintenance personnel in real time in the form of emails, allowing them to obtain key information at the first time and take timely measures to avoid or mitigate the impact of network failures. This mechanism not only greatly reduces the complexity and burden of manual operations, but also improves the accuracy and real-time nature of inspections. It is an important practice of network operation and maintenance automation and intelligence, and reflects the significant innovation of this application scheme in improving network maintenance efficiency and preventing network anomalies.

[0129] In an embodiment of the present application, in order to provide a user-friendly interface, Python's Qt library (PyQt) can be used to develop a front-end interface. PyQt is a library for creating GUI applications in Python. It can be used to build an intuitive and easy-to-use interface, making it convenient for users to trigger a one-click survey of links and view prediction and inspection results. Front-end development includes the creation and layout of interface components, the connection of signals and slots, and the processing of event loops to ensure the responsiveness and interactivity of the application, providing network operation and maintenance personnel with more efficient work tools.

[0130] In the embodiment of the present application, a deep integration of the terminal and the mobile terminal is introduced through the prediction system of the communication link (a one-click link census system), which realizes cross-professional and cross-regional rapid link utilization query, greatly improving the network operation and maintenance efficiency and response speed. At the same time, the automation and standardization of the daily inspection mechanism of the link, combined with the automatic inspection during the evening peak and the push of email reports, effectively monitors the health of the network, reduces the manual burden, and ensures the stable operation of the network. Finally, the link timing prediction algorithm, especially the integration of the SARIMA model, the periodic feature model and the xLSTM network, pioneered the processing of the periodicity and long-term dependence of the link utilization, significantly improved the prediction accuracy and stability, and provided forward-looking technical support for preventing network congestion and optimizing resource allocation.

[0131] Overall, this application creates a comprehensive link operation and maintenance solution covering monitoring, early warning and prediction through the combination of mobile terminal integration, automated inspection and deep learning prediction. It reflects the practicality and foresight of the technology and is a major innovation in the traditional network operation and maintenance model.

[0132] According to an embodiment of the present application, a communication link prediction device is also provided. It should be noted that the communication link prediction device of the embodiment of the present application can be used to execute the communication link prediction method provided in the embodiment of the present application. The following introduces the communication link prediction device provided in the embodiment of the present application.

[0133] Figure 5 FIG. 1 is a structural diagram of a communication link prediction device provided according to an embodiment of the present application. Figure 5 As shown, the device includes:

[0134] An acquisition module 50 is used to acquire communication link data;

[0135] a first prediction module 52 for processing the communication link data using a first model to obtain a first prediction result, wherein the first model is used to capture trend changes and seasonal characteristics in the communication link data, and the first prediction result is an initial utilization prediction value of the communication link;

[0136] a determination module 54, configured to determine a residual value between the first prediction result and the communication link data;

[0137] a second prediction module 56 for processing the residual values using a second model to obtain a second prediction result, wherein the second model is used to learn a time series pattern in the residual values, and the second prediction result is used to represent a future trend of the residual values;

[0138] The fusion module 58 is configured to fuse the first prediction result and the second prediction result to obtain a target prediction result of the communication link.

[0139] Through the acquisition module, first prediction module, determination module, second prediction module and fusion module in the above-mentioned communication link prediction device, the purpose of accurately predicting the future utilization of the communication link is achieved, thereby realizing the technical effects of optimizing network resources, preventing network congestion and ensuring network stability, and further solving the technical problem of low efficiency in real-time monitoring and prediction of communication link utilization caused by frequent traffic congestion in large-scale network environments.

[0140] In the communication link prediction device provided in the embodiment of the present application, the second prediction module is also used to optimize the residual value through the third model to obtain the residual optimization value, wherein the third model is used to enhance the correlation between different feature vectors in the residual value.

[0141] The communication link prediction device provided in the embodiment of the present application also includes a training module 510, which is used to obtain historical communication link data; determine a first eigenvector and a second eigenvector based on the historical communication link data, wherein the first eigenvector is used to represent a data set of historical communication link data in the same periodic period, and the second eigenvector is used to represent a data set of historical communication link data in a preset time period whose feature similarity is greater than a preset threshold; modulate the first eigenvector and the second eigenvector according to a preset feature scale to obtain a third eigenvector corresponding to the first eigenvector, and obtain a fourth eigenvector corresponding to the second eigenvector; train the initial optimization model based on the third eigenvector and the fourth eigenvector to obtain a third model.

[0142] In the prediction device of the communication link provided in the embodiment of the present application, the training module is also used to perform interpolation processing on the third eigenvector and the fourth eigenvector to obtain a fifth eigenvector corresponding to the third eigenvector, and to obtain a sixth eigenvector corresponding to the fourth eigenvector; determine a first compensation factor based on the third eigenvector and the fifth eigenvector, wherein the first compensation factor is used to represent the degree of mapping deviation between the first eigenvector after modulation processing and after interpolation processing; determine a second compensation factor based on the fourth eigenvector and the sixth eigenvector, wherein the second compensation factor is used to represent the degree of mapping deviation between the second eigenvector after modulation processing and after interpolation processing; and correct the third eigenvector and the fourth eigenvector based on the first compensation factor and the second compensation factor.

[0143] In the communication link prediction device provided in an embodiment of the present application, the training module is also used to obtain time series characteristics of historical communication link data, wherein the time series characteristics include trend changes, seasonal characteristics and periodic characteristics of the historical communication link data; determine the autocorrelation function and partial autocorrelation function of the historical communication link data based on the time series characteristics, wherein the autocorrelation function is used to represent the data correlation of the historical communication link data at different time points, and the partial autocorrelation function is used to represent the data correlation of the historical communication link data at a specific lag time point; determine the seasonal parameters of the initial prediction model based on the autocorrelation function and the partial autocorrelation function, wherein the seasonal parameters are used to reflect the periodic behavior of the historical communication link data; train the initial prediction model based on the seasonal parameters and the historical communication link data, and optimize the model parameters of the initial prediction model to obtain a first model.

[0144] In the prediction device for the communication link provided in the embodiment of the present application, the training module is also used to perform a unit root test on the time series characteristics to obtain a test result, wherein the test result is used to indicate the degree of stationarity of the time series characteristics; when the test result meets the preset conditions, the time series characteristics are determined to be a stationary sequence; when the test result does not meet the preset conditions, the time series characteristics are determined to be an unstationary sequence, and the unstationary sequence is subjected to differential processing until the unstationary sequence reaches stationarity.

[0145] In the communication link prediction device provided in the embodiment of the present application, the training module is also used to predict historical communication link data through a first model to obtain historical prediction results; determine the historical residual value between the historical prediction results and the historical communication link data; optimize the historical residual value through a third model to obtain a historical residual optimization value; and train the initial residual prediction model based on the historical residual optimization value to obtain a second model.

[0146] An embodiment of the present application further provides an electronic device, comprising: a memory and a processor, wherein the memory is used to store program instructions; the processor is connected to the memory and is used to execute the prediction method for implementing the above-mentioned communication link.

[0147] It should be noted that the above electronic equipment is used to perform Figure 2 The communication link prediction method shown in the figure, therefore the relevant explanations in the above communication link prediction method are also applicable to the electronic device and will not be repeated here.

[0148] An embodiment of the present application further provides a non-volatile storage medium, which includes a stored computer program, wherein the device where the non-volatile storage medium is located executes the above-mentioned communication link prediction method by running the computer program.

[0149] It should be noted that the above non-volatile storage medium is used to execute Figure 2 The communication link prediction method shown in the figure, therefore, the relevant explanations in the above communication link prediction method are also applicable to the non-volatile storage medium, and will not be repeated here.

[0150] An embodiment of the present application also provides a computer program product, including computer instructions, which implement the above-mentioned communication link prediction method when executed by a processor.

[0151] It should be noted that the above-mentioned computer program product is used to execute Figure 2 The communication link prediction method shown in the figure, therefore, the relevant explanations in the above communication link prediction method are also applicable to the computer program product and will not be repeated here.

[0152] The serial numbers of the above embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0153] In the above embodiments of the present application, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0154] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of the units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0155] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0156] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0157] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0158] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for predicting a communication link, characterized in that: include: Obtain communication link data; Processing the communication link data using a first model to obtain a first prediction result, wherein the first model is used to capture trend changes and seasonal characteristics in the communication link data, and the first prediction result is an initial utilization prediction value of the communication link; determining a residual value between the first prediction result and the communication link data; Processing the residual value through a second model to obtain a second prediction result, wherein the second model is used to learn a time series pattern in the residual value, and the second prediction result is used to represent a future change trend of the residual value; The first prediction result and the second prediction result are integrated to obtain a target prediction result of the communication link.

2. The method according to claim 1, characterized in that After determining the residual value between the first prediction result and the communication link data, the method further includes: The residual value is optimized by a third model to obtain a residual optimization value, wherein the third model is used to enhance the correlation relationship between different eigenvectors in the residual value.

3. The method according to claim 2, characterized in that The third model is trained in the following way: Obtain historical communication link data; Determining a first eigenvector and a second eigenvector based on the historical communication link data, wherein the first eigenvector is used to represent a data set of the historical communication link data in the same periodic period, and the second eigenvector is used to represent a data set of the historical communication link data having a feature similarity greater than a preset threshold within a preset time period; Modulating the first eigenvector and the second eigenvector according to a preset characteristic scale to obtain a third eigenvector corresponding to the first eigenvector and a fourth eigenvector corresponding to the second eigenvector; The initial optimization model is trained according to the third eigenvector and the fourth eigenvector to obtain the third model.

4. The method according to claim 3, characterized in that Before training the initial optimization model according to the third eigenvector and the fourth eigenvector, the method further includes: performing interpolation processing on the third eigenvector and the fourth eigenvector to obtain a fifth eigenvector corresponding to the third eigenvector, and obtaining a sixth eigenvector corresponding to the fourth eigenvector; determining a first compensation factor based on the third eigenvector and the fifth eigenvector, wherein the first compensation factor is used to represent a degree of mapping deviation between the first eigenvector after modulation processing and after interpolation processing; determining a second compensation factor based on the fourth eigenvector and the sixth eigenvector, wherein the second compensation factor is used to represent a degree of mapping deviation between the second eigenvector after modulation processing and after interpolation processing; The third eigenvector and the fourth eigenvector are modified according to the first compensation factor and the second compensation factor.

5. The method according to claim 2, characterized in that The second model is trained in the following way: Predicting historical communication link data using the first model to obtain historical prediction results; Determining a historical residual value between the historical prediction result and the historical communication link data; Optimizing the historical residual value by the third model to obtain a historical residual optimization value; The initial residual prediction model is trained according to the historical residual optimization value to obtain the second model.

6. The method according to claim 3, characterized in that The first model is trained in the following way: Acquire time series characteristics of the historical communication link data, wherein the time series characteristics include trend changes, seasonal characteristics, and periodic characteristics of the historical communication link data; Determining an autocorrelation function and a partial autocorrelation function of the historical communication link data based on the time series characteristics, wherein the autocorrelation function is used to represent the data correlation of the historical communication link data at different time points, and the partial autocorrelation function is used to represent the data correlation of the historical communication link data at a specific lag time point; Determining a seasonal parameter of an initial prediction model based on the autocorrelation function and the partial autocorrelation function, wherein the seasonal parameter is used to reflect the periodic behavior of the historical communication link data; The initial prediction model is trained according to the seasonal parameters and the historical communication link data, and the model parameters of the initial prediction model are optimized to obtain the first model.

7. The method according to claim 6, characterized in that After obtaining the time series characteristics of the historical communication link data, the method further includes: Performing a unit root test on the time series characteristics to obtain a test result, wherein the test result is used to indicate the degree of stationarity of the time series characteristics; When the test result meets the preset conditions, determining that the time series feature is a stationary sequence; When the test result does not meet the preset condition, the time series feature is determined to be an unsteady sequence, and a differential process is performed on the unsteady sequence until the unsteady sequence becomes stable.

8. A communication link prediction system, characterized in that: include: Mobile terminal, data processing platform, collection platform and performance database, among which, The mobile terminal is used to initiate a query request for the communication link; The collection platform is used to collect communication link data of the communication device according to the query request, and send the communication link data to the performance database for storage, and send the communication link data to the data processing platform for processing; The performance database is used to store the communication link data; The data processing platform is used to process the communication link data and return the data processing results to the mobile terminal; the data processing platform is also used to detect the communication equipment according to a preset inspection frequency; the data processing platform is also used to obtain communication link data from the performance database and execute the communication link prediction method described in any one of claims 1 to 7.

9. A communication link prediction device, characterized in that: include: An acquisition module, used for acquiring communication link data; a first prediction module, configured to process the communication link data using a first model to obtain a first prediction result, wherein the first model is configured to capture trend changes and seasonal characteristics in the communication link data, and the first prediction result is an initial utilization prediction value of the communication link; a determination module, configured to determine a residual value between the first prediction result and the communication link data; a second prediction module, configured to process the residual value using a second model to obtain a second prediction result, wherein the second model is configured to learn a time series pattern in the residual value, and the second prediction result is configured to represent a future change trend of the residual value; A fusion module is used to fuse the first prediction result and the second prediction result to obtain a target prediction result of the communication link.

10. An electronic device, characterized in that: include: A memory and a processor, wherein the memory is used to store program instructions; The processor is connected to the memory and is used to execute the communication link prediction method according to any one of claims 1 to 7.

11. A non-volatile storage medium, characterized in that: The non-volatile storage medium includes a stored computer program, wherein the device where the non-volatile storage medium is located executes the communication link prediction method according to any one of claims 1 to 7 by running the computer program.

12. A computer program product comprising computer instructions, characterized in that When the computer instructions are executed by a processor, the communication link prediction method according to any one of claims 1 to 7 is implemented.

Images