Network performance prediction method and device and electronic equipment

Abstract

The invention discloses a network performance prediction method and device and electronic equipment. The method comprises the following steps: obtaining network performance data corresponding to a target area, the network performance data comprising data of a plurality of base station cells collected from a plurality of systems; a plurality of target indexes are determined from the network performance data, the target indexes comprise indexes related to network performance experience, and the network performance experience is used for quantitatively representing the satisfaction degree of users in the target area for network services; the target index is predicted through the prediction model, a prediction result is obtained, and the prediction result is used for quantitatively representing the network performance of the target area. The network performance prediction method and device solve the technical problems that the network performance prediction technology adopted in the related technology only pays attention to the performance state of a single network device, the adopted prediction index is single, the overall network use experience of a user cannot be comprehensively reflected, and the prediction result is inaccurate.

Description

Technical Field

[0001] The present application relates to the field of communication technologies, and in particular, to a method, apparatus, and electronic device for predicting network performance. Background Art

[0002] With the wide deployment of the fifth-generation mobile communication technology (5G), especially 5G networks, the scenarios of the Internet of Everything are becoming increasingly rich, and the demand for network performance is also getting higher and higher. For telecommunications operators, how to predict network performance is not only related to network optimization but also related to the investment in network construction. The network performance prediction technologies adopted by related technologies only focus on the performance status of a single base station or a single network device, without considering the mutual influence between devices in the network coverage area, or only considering a single indicator, which cannot comprehensively reflect the overall network usage experience of users, resulting in inaccurate prediction results.

[0003] In view of the above problems, no effective solution has been proposed yet. Summary of the Invention

[0004] Embodiments of the present application provide a method, apparatus, and electronic device for predicting network performance, so as to at least solve the technical problems that the network performance prediction technologies adopted by related technologies only focus on the performance status of a single network device, and the adopted prediction indicators are single, which cannot comprehensively reflect the overall network usage experience of users, resulting in inaccurate prediction results.

[0005] According to one aspect of the embodiments of the present application, a method for predicting network performance is provided, including: obtaining network performance data corresponding to a target area, where the network performance data includes data of multiple base station cells collected from multiple systems; determining multiple target indicators from the network performance data, where the target indicators include indicators related to network performance experience, and the network performance experience is used to quantitatively represent the satisfaction degree of users in the target area with respect to network services; using a prediction model to predict the target indicators to obtain a prediction result, where the prediction result is used to quantitatively represent the network performance of the target area.

[0006] In some embodiments of the present application, determining multiple target indicators from the network performance data includes: determining a first indicator set corresponding to the network performance data, where the first indicator set includes all network performance indicators corresponding to the network performance data; calculating the mutual information value between the first indicator and the network performance experience, where the first indicator is any one indicator in the first indicator set, and the mutual information value is used to quantitatively represent the correlation between the first indicator and the network performance experience; adding the first indicator corresponding to the highest mutual information value to a second indicator set; determining the target indicators according to the first indicator set and the second indicator set.

[0007] In some embodiments of the present application, determining a target metric based on a first metric set and a second metric set includes: calculating an average mutual information value between the second metric set and a third metric, where the third metric is any one of the first metrics in the first metric set other than the second metric set, and the average mutual information value is used to quantitatively represent the information redundancy degree between the third metric and the second metric set; determining the difference between the mutual information value corresponding to each third metric and the average mutual information value corresponding to each third metric, and adding the third metric corresponding to the maximum difference to the second metric set; repeatedly executing the calculation process of the mutual information value and the average mutual information value until the second metric set meets a preset condition; and determining the second metric set as the target metric.

[0008] In some embodiments of the present application, the prediction model includes an input layer, a first hidden layer, a second hidden layer, and an output layer, where the first hidden layer and the second hidden layer are configured with the same number of neurons; using the prediction model to predict the target metric to obtain a prediction result, including: the input layer converts the target metric into time series data to obtain a feature vector corresponding to each moment, where each feature vector includes the values of all target metrics at the corresponding moment; the first hidden layer processes the feature vector corresponding to each moment in sequence from the start point to the end point of the time series data to obtain a first hidden state corresponding to each moment, where the first hidden state includes the target metric information corresponding to the start point to the current moment of the time series data; the second hidden layer processes the feature vector corresponding to each moment in sequence from the end point to the start point of the time series data to obtain a second hidden state corresponding to each moment, where the second hidden state includes the target metric information corresponding to the current moment to the end point of the time series data; the output layer integrates the first hidden state and the second hidden state corresponding to each moment to obtain a target feature vector, and determines the prediction result according to the target feature vector.

[0009] In some embodiments of the present application, the prediction model further includes a third hidden layer, where the third hidden layer is used to integrate the first hidden state and the second hidden state corresponding to each moment.

[0010] In some embodiments of the present application, the third hidden layer includes a first sub-layer, a second sub-layer, and a third sub-layer. The first sub-layer includes a first number of neurons, and the second sub-layer and the third sub-layer include a second number of neurons, where the first number is less than the second number.

[0011] In some embodiments of the present application, the target area is determined in the following manner: obtaining a first vertex selected by a target object on a map, where the first vertex is the starting vertex of the target area determined from an initial area; detecting a second vertex set selected by the target object on the map, where the second vertex set is used to determine the shape of the target area; when the first vertex is included in the second vertex set, obtaining the first longitude and latitude coordinate information corresponding to each vertex in the second vertex set; and determining the closed area formed by the first longitude and latitude coordinate information as the target area.

[0012] In some embodiments of the present application, after determining the closed area formed by the first longitude and latitude coordinate information as the target area, the method further includes: dividing the initial area into a plurality of grids, where the sizes of the plurality of grids are the same; obtaining the second longitude and latitude coordinate information corresponding to the second center point of each grid, and generating an identifier corresponding to each grid according to the second longitude and latitude coordinate information; determining all the identifiers included in the target area to obtain an identifier set; and determining the grids corresponding to each identifier in the identifier set as sub-areas of the target area.

[0013] In some embodiments of the present application, after obtaining the network performance data corresponding to the target area, the method further includes: mapping the network performance data to the sub-areas of the target area to obtain the sub-network performance data corresponding to each sub-area; determining the sub-target indicators corresponding to each sub-area from the sub-network performance data; and using a prediction model to predict the sub-target indicators to obtain the prediction results corresponding to each sub-area.

[0014] In some embodiments of the present application, after obtaining the prediction results corresponding to each sub-area, the method further includes: determining the result type corresponding to the prediction result of each sub-area, and repainting the sub-areas corresponding to the target area on the map according to the result type, where each result type corresponds to a filling color.

[0015] In some embodiments of the present application, after determining a plurality of target indicators from the network performance data, the method further includes: dividing each indicator in the target indicators into different indicator types, where different indicator types are used to reflect different dimensions affecting the network performance experience; and determining the weight corresponding to each indicator in the target indicators according to the indicator type.

[0016] According to another aspect of the embodiments of the present application, there is also provided a device for predicting network performance, including: an acquisition module, configured to acquire network performance data corresponding to a target area, where the network performance data includes data of multiple base station cells collected from multiple systems; a determination module, configured to determine multiple target metrics from the network performance data, where the target metrics include metrics that are correlated with the network performance experience, and the network performance experience is used to quantitatively represent the satisfaction degree of users in the target area with respect to network services; a prediction module, configured to use a prediction model to predict the target metrics to obtain a prediction result, where the prediction result is used to quantitatively represent the network performance of the target area.

[0017] According to still another aspect of the embodiments of the present application, there is also provided an electronic device, including: a memory and a processor, where the memory is used to store program instructions; the processor is connected to the memory and is configured to execute to implement the above-mentioned method for predicting network performance.

[0018] According to still another aspect of the embodiments of the present application, there is also provided a non-volatile storage medium, where the non-volatile storage medium includes a stored computer program, and the device where the non-volatile storage medium is located executes the above-mentioned method for predicting network performance by running the computer program.

[0019] According to still another aspect of the embodiments of the present application, there is also provided a computer program product, including computer instructions, where the computer instructions, when executed by a processor, implement the above-mentioned method for predicting network performance.

[0020] In the embodiments of the present application, by determining metrics that are highly correlated with the network performance experience from the data of multiple base station cells collected from multiple systems and using a prediction model for prediction, the purpose of improving the efficiency and accuracy of network performance prediction is achieved, thereby realizing the technical effects of effectively improving the network service quality and enhancing the user experience, and further solving the technical problem that the network performance prediction technology adopted in the related art only focuses on the performance status of a single network device, and the prediction metrics used are single, and cannot comprehensively reflect the overall network usage experience of users, resulting in inaccurate prediction results. 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 to the present application. In the drawings:

[0022] Figure 1 is a hardware structure block diagram of a computer terminal for a method of predicting network performance according to an embodiment of the present application;

[0023] Figure 2 is a flowchart of a method of predicting network performance according to an embodiment of the present application;

[0024] Figure 3 It is a schematic diagram of the prediction model structure of a network performance prediction method according to an embodiment of the present application;

[0025] Figure 4 It is a schematic diagram of the prediction result of a network performance prediction method according to an embodiment of the present application;

[0026] Figure 5 It is the overall flowchart of a network performance prediction method according to an embodiment of the present application;

[0027] Figure 6 It is a schematic diagram of the structure of a network performance prediction device according to an embodiment of the present application. Detailed implementation manners

[0028] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0029] 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 do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application described here can be implemented in an order different from those illustrated or described here. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0030] In order to better understand the embodiments of the present application, the technical terms involved in the embodiments of the present application are explained as follows:

[0031] Geographic Information System (GIS for short): A geographic information system is a technical system for collecting, storing, managing, and analyzing geospatial data, which can associate various types of data with geographic coordinates, support map rendering, spatial analysis, and decision-making support. In the present application, GIS is used for map selection and visualization of network resources.

[0032] Maximum Relevance Minimum Redundancy algorithm (abbreviated as mRMR): The Maximum Relevance Minimum Redundancy algorithm is a feature selection method that aims to select a feature subset from the original feature set that is most relevant to the target variable and has the least redundancy among each other. The mRMR algorithm can ensure that the selected feature set contains both the most valuable information for predicting the target variable and avoids redundant calculations between features, improving the efficiency and interpretability of the model. In this application, the mRMR algorithm is used to screen out the most valuable performance metrics from a large number of performance metrics, improving the interpretability and efficiency of the model.

[0033] Bidirectional Long Short-Term Memory network (abbreviated as BiLSTM): The Bidirectional Long Short-Term Memory network is a neural network architecture commonly used in time series prediction and natural language processing. It combines the long-term and short-term memory capabilities of LSTM (Long Short-Term Memory) and the ability to process data bidirectionally. BiLSTM can process the input sequence in both the forward and backward directions, enabling each LSTM unit at each time point in the middle to receive the complete information of the sequence. In this application, BiLSTM is used to effectively process time series data (such as performance metric data generated according to time series), capture the changing trend of network performance metrics over time, and improve the adaptability of the model to long-term prediction.

[0034] The network performance experience prediction technologies adopted by related technologies mainly face the following challenges: First, most technologies are limited to the analysis at the single device level and lack the ability to make comprehensive predictions from a broader geographical area perspective, which ignores the importance of the regional dimension in network performance evaluation; Second, the prediction metrics are relatively single and cannot comprehensively reflect the overall network usage experience of customers. This limitation may lead to an incomplete evaluation of the network status and make it difficult to truly reflect the actual experience of users and the quality of network services; Finally, network performance data has a high degree of dynamism and uncertainty, and these characteristics intensify over time, making it more difficult to predict performance changes in the future for a relatively long period. The prediction accuracy of traditional model performance metrics is generally not high. It may perform okay when dealing with short-term predictions, but when the prediction time is extended to several hours, days, or even longer, the prediction errors of the model will accumulate, ultimately resulting in a significant decline in the accuracy of the prediction results.

[0035] To solve the above technical problems, the embodiments of this application provide corresponding solutions, which are described in detail below.

[0036] The embodiment of the network performance prediction method provided by the embodiments of the present application can be executed on a mobile terminal, a computer terminal, or a similar computing device. Figure 1 The following shows a hardware block diagram of a computer terminal for implementing the network performance prediction method. As Figure 1 shown, the computer terminal 10 may include one or more processors (the processors may include, but are not limited to, processing devices such as a microprocessor MCU or a programmable logic device FPGA, shown as 102a, 102b, ……, 102n in the figure), a memory 104 for storing data, and a transmission module 106 for communication functions connected by wired and / or wireless networks. In addition, it may further 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. Those of ordinary skill in the art can understand that Figure 1 the structure shown is only schematic and does not limit the structure of the above electronic device. For example, the computer terminal 10 may further include more or fewer components than those Figure 1 shown, or have a different configuration from that Figure 1 shown.

[0037] It should be noted that the above one or more processors and / or other data processing circuits are generally referred to as "data processing circuits" in this article. The data processing circuit may be embodied in software, hardware, firmware, or any combination thereof, in whole or in part. In addition, the data processing circuit may be a single independent processing module, or be incorporated in whole or in part into any one of the other elements in the computer terminal 10. As involved in the embodiments of the present application, the data processing circuit is used for a processor control (such as the selection of a variable resistor terminal path connected to an interface).

[0038] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage devices corresponding to the network performance prediction method in the embodiments 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, implements the above network performance 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 memories. In some instances, the memory 104 may further include a memory remotely set relative to the processor, and these remote memories may be connected to the computer terminal 10 through a network. Examples of the above networks include, but are not limited to, the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0039] The transmission module 106 is used to receive or send data via a network. Specific examples of the above-mentioned network may include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission module 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission module 106 can be a Radio Frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0040] The display can be, for example, a touch-screen liquid crystal display (LCD), which enables a user to interact with the user interface of the computer terminal 10.

[0041] It should be noted here that, in some alternative embodiments, the above Figure 1 illustrated computer terminal may include hardware elements (including circuits), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware elements and software elements. It should be pointed out that Figure 1 is only an example of a specific specific instance and is intended to illustrate the types of components that may exist in the above computer terminal.

[0042] Under the above operating environment, an embodiment of a method for predicting network performance is provided in an embodiment of the present application. It should be noted that the steps illustrated in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is illustrated in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0043] Figure 2 is a flowchart of a method for predicting network performance according to an embodiment of the present application. As Figure 2 shown, the method includes the following steps:

[0044] Step S202, obtaining network performance data corresponding to a target area, where the network performance data includes data of a plurality of base station cells collected from a plurality of systems.

[0045] In the above step S202, the target area includes a geographical range selected by a target object on a map interface and is used to define the geographical boundary of the network performance data.

[0046] Network performance data involves various performance indicators of base station cells, including but not limited to uplink PRB utilization rate, UE context establishment success rate, etc., and can be sourced from multiple systems such as 5GR, comprehensive adjustment, mobile command center, big data lake, etc. In some embodiments of the present application, in order to obtain network performance data corresponding to a target area, data from different systems such as 5GR, comprehensive adjustment, mobile command center, big data lake, etc. can be associated and aggregated to form a unified dataset (network performance geographic database) for the target area. This dataset contains multiple network performance indicators such as cell uplink PRB utilization rate, uplink PUSCH PRB utilization rate, UE context establishment success rate, etc. The data in the network performance geographic database integrates geographic information and network performance data.

[0047] Since network performance data may be constantly changing, a data request module can be designed to send data requests to each data source system (such as 5GR, comprehensive adjustment, etc.) according to the definition of the target area to obtain the latest network performance data. The data request module can work in a periodic or event-driven manner. For example, it can automatically request updates every hour or trigger updates when significant changes occur in the network performance data to ensure that the data on which the prediction model is based is the latest and most accurate.

[0048] To improve the flexibility and pertinence of prediction, the target area is determined in the following way: obtain the first vertex selected by the target object on the map, where the first vertex is the starting vertex of the target area determined from the initial area; detect the second vertex set selected by the target object on the map, where the second vertex set is used to determine the shape of the target area; when the first vertex is included in the second vertex set, obtain the first longitude and latitude coordinate information corresponding to each vertex in the second vertex set; determine the closed area formed by the first longitude and latitude coordinate information as the target area.

[0049] The target object can refer to a user or an automated script, which is the initiator of customizing the area selection on the map. The target object interacts on the map interface to select a specific geographical area for comprehensive analysis and prediction of network performance experience.

[0050] The first vertex is the starting vertex of the target area, which is determined when the target object first clicks on the map. The first vertex is the starting point of polygon drawing, and its coordinate information will be recorded by the system and used for the initialization of the subsequent drawing process. The second vertex set is a set of vertices formed by the target object continuously clicking on the map, which is used to depict the complete boundary of the target area. The second vertex set jointly constructs the border of the polygon, and the shape and scope of the target area can be determined through its coordinate information.

[0051] The first longitude and latitude coordinate information refers to the longitude and latitude coordinates of the first vertex and each vertex in the second vertex set obtained.

[0052] In some embodiments of the present application, first, the map is initialized and waits for the interaction of the target object. When the first click of the target object on the map is detected, the longitude and latitude coordinates of the click position are recorded and used as the first vertex of the target area. Subsequently, the subsequent click events of the target object are continuously monitored. Whenever the target object clicks on the map, the longitude and latitude coordinates of the click position are recorded until a closed polygon is formed. When initializing the map, a query instruction of the target object can be received. The query instruction includes the geographical location information of the initial area. The center point of the initial area is determined through the geographical location information of the initial area, and the map is positioned to this center point. The initial area can be displayed to the target object through the interaction interface and wait for the interaction of the target object. During the drawing process, the polygon contour being constructed can be dynamically displayed through the interaction interface to ensure that the path is clearly visible.

[0053] When the target object draws a polygon, the coordinate information of the vertices can be monitored in real time, and the border of the polygon can be constructed. When the target object clicks on the first vertex again, that is, the starting point, this operation will be recognized as a signal to close the polygon. At this time, the longitude and latitude coordinate information of all vertices including the starting point is recorded, and it is calculated through algorithms (such as coordinate sorting, edge detection, or polygon closure verification algorithms) whether these coordinates form a closed polygon. If a closed area is formed, the boundary of the target area can be confirmed according to the recorded coordinate information and determined as the geographical range for subsequent network performance data analysis.

[0054] To improve the prediction accuracy, the target area can be divided into multiple sub-areas, and the prediction model is used for prediction for each sub-area respectively, or after using the prediction model to predict the entire target area, the prediction results are respectively converted into the prediction results corresponding to each sub-area. Specifically, after determining the closed area formed by the first longitude and latitude coordinate information as the target area, the following steps can also be executed: dividing the initial area into multiple grids, where the sizes of the multiple grids are the same; obtaining the second longitude and latitude coordinate information corresponding to the second center point of each grid, and generating identifiers corresponding to each grid respectively according to the second longitude and latitude coordinate information; determining all identifiers included in the target area to obtain an identifier set; and determining the grids corresponding to each identifier in the identifier set as the sub-areas of the target area.

[0055] In some embodiments of the present application, first, the boundary range of the initial area is determined, and it is evenly divided into multiple grids according to the set grid size (for example, 20 meters × 20 meters). For each grid, the coordinate information of its center point (i.e., the second longitude and latitude coordinate information) is calculated, and these coordinates will be used as the reference points for subsequent network performance data association.

[0056] After obtaining multiple grids, a unique identifier (such as the second center point of the grid) is generated for each grid, which will be used as the index key in the grid database to ensure the traceability and accuracy of each grid. According to the coordinate information of the second center point of each grid, network performance data matching the grid location can also be retrieved from the constructed network performance geographic database. For example, by comparing the grid center point coordinates with the geographic coordinates of the base station cells, the base station cells belonging to the same grid can be found, and then the network performance data of these base station cells can be obtained. In this way, the network performance data can be associated with specific geographic grids, providing a geographically accurate dataset for subsequent feature engineering and model prediction.

[0057] Using the pre-established efficient query index, quickly locate and retrieve the grid identifiers of all grid cells falling within the selected area (target area) based on the geometric object of the circled boundary, and each grid within the target area is respectively determined as a sub-area of the target area. It should be noted that one grid can be determined as a sub-area, or multiple grids can be determined as a sub-area, and the size of each sub-area can be the same or different.

[0058] Step S204, determine multiple target metrics from the network performance data, where the target metrics include metrics that are relevant to the network performance experience, and the network performance experience is used to quantitatively represent the satisfaction degree of users in the target area with the network service.

[0059] In the above step S204, the target metrics are a set of key metrics directly related to the network performance experience selected from the network performance data. For example, the target metrics can be selected by the mRMR algorithm to more accurately reflect the satisfaction degree of users with the network service and provide valuable inputs for the subsequent prediction model.

[0060] The network performance experience is used to quantitatively represent the satisfaction degree of users with the network service, which can be achieved by comprehensively evaluating metrics such as the signal quality, data rate, latency, and connection success rate of the network.

[0061] In some embodiments of the present application, multiple target metrics can be determined from the network performance data through the following steps: determine the first metric set corresponding to the network performance data, where the first metric set includes all network performance metrics corresponding to the network performance data; calculate the mutual information value between the first metric and the network performance experience, where the first metric is any metric in the first metric set, and the mutual information value is used to quantitatively represent the correlation between the first metric and the network performance experience; add the first metric corresponding to the highest mutual information value to the second metric set; determine the target metrics according to the first metric set and the second metric set.

[0062] Determine the target indicator based on the first indicator set and the second indicator set. Specifically: Calculate the average mutual information value between the second indicator set and a third indicator, where the third indicator is any one of the first indicators in the first indicator set other than the second indicator set. The average mutual information value is used to quantitatively represent the information redundancy degree between the third indicator and the second indicator set. Determine the difference between the mutual information value corresponding to each third indicator and the average mutual information value corresponding to each third indicator, and add the third indicator corresponding to the maximum difference to the second indicator set. Repeat the calculation process of the mutual information value and the average mutual information value until the second indicator set meets the preset condition. Determine the second indicator set as the target indicator.

[0063] In a machine learning model, multicollinearity among predictor variables may cause the model to contain redundant information. To simplify model training and improve efficiency, it is necessary to eliminate these redundant independent variables. In some embodiments of the present application, the mRMR feature extraction algorithm can be used to deeply analyze and extract features from the network performance data in the network performance geographic database, and screen out the key indicators that are most closely related to the user network experience as the input parameters of the regional network experience analysis prediction model, which not only improves the prediction accuracy of the model, but also reduces the complexity of the model, making the model more efficient and easy to interpret.

[0064] Determining the first indicator set corresponding to the network performance data is the starting point of the entire prediction process, which ensures that all potential network performance indicators can be considered when constructing the model. Specifically, when determining the first indicator set, the following steps can be performed:

[0065] (1) Comprehensively collect network performance indicators.

[0066] Converge network performance data from multiple systems such as 5GR, comprehensive adjustment, mobile command center, and big data lake to form an initial indicator set of a first preset number (such as 400) of indicators. The indicators in the initial indicator set cover multiple dimensions such as resource load, call access, mobility management, data traffic, service integrity, air interface performance, and base station alarms, ensuring the comprehensiveness and depth of the indicator set.

[0067] (2) Refine the indicator analysis dimension.

[0068] According to the perspective of business impact, further classify the preset number of indicators collected initially to ensure full coverage of each dimension, which helps to understand network performance from a more detailed level. For example, it can be divided into indicators of a second preset number (such as 7) of main dimensions such as resource load, call access, mobility management, data traffic, service integrity, air interface performance, and base station alarms, which are used to understand network performance from different business perspectives and provide a clear framework for subsequent feature selection and model construction.

[0069] (3) Multiple algorithms are integrated for feature selection.

[0070] By using the mRMR algorithm, feature selection can be performed on the initial indicator set, the mutual information value between each indicator and the network performance experience can be calculated, and the correlation and redundancy of the indicators can be quantified to screen out the indicator subset that is most relevant to the network performance experience and has the least redundancy between each other. For example, the top 20 indicators with the highest mutual information values ​​can be used as the target indicator set for subsequent model training and prediction.

[0071] The mRMR algorithm focuses on optimizing the feature selection process, aiming to improve the quality of the feature set. It achieves this goal by maximizing the correlation between the features and the target variable while reducing the redundancy between the features. Specifically, it can prioritize the features with the largest mutual information value with the target variable, while ensuring that the mutual information value between the newly selected features and the existing feature set is minimized, thereby avoiding the introduction of highly redundant information. In this way, not only the prediction accuracy of the model is improved, but also the interpretability of the model is enhanced, making the model more concise and efficient. Specifically:

[0072] (1) Data initialization: Samples are extracted from the network performance geographic database after data preprocessing to form a sample set, all indicator sets establish a full feature set (i.e., the first indicator set), and the target feature subset is established and initialized.

[0073] (2) Calculate the mutual information between the characteristic index and the network performance experience: For each characteristic index x i The mutual information value between (i.e., the first indicator) and the network performance experience y is calculated to evaluate the correlation between the feature and the network performance experience. The calculation formula is:

[0074]

[0075] Among them, I(x i ; y) is the characteristic variable x i The mutual information value (i.e., mutual information value) of the network performance experience y, p(x i ) and p(y) are the characteristic variables x i and the probability density of network performance experience y (i.e., the frequency of occurrence in the sample set or the probability within a certain range of values), p(x i , y) is the characteristic variable x i and the joint probability density of network performance experience y.

[0076] (3) Select the most relevant feature indicator: Based on the calculation results of the mutual information between the feature indicator and the network performance experience, the feature indicator with the highest score (the first indicator added to the second indicator set) is selected and added to the feature selection set S (i.e., the second indicator set).

[0077] (4) Calculate the redundancy between features: For each feature index x in the selected feature set S j (the index in the second index set), calculate the average mutual information between the new candidate feature x i (any first index in the first index set other than the second index set) and each feature in S. The calculation formula is:

[0078]

[0079] where I(x i , y) is the average mutual information value between the new candidate feature x i and the feature selection set S (which can also be expressed as I(x i ; y)), |S| is the total number of features in the feature selection set S, x j is the existing feature in the feature selection set S, and I(x i , x j ) is the mutual information value between the new candidate feature x i and a certain feature x j in the feature selection set S (which can also be expressed as I(x i ; x j )).

[0080] (5) Evaluate and select features: According to the mRMR algorithm, evaluate the score of each candidate feature x i , and select the feature index with the highest score to be added to the feature selection set S. The calculation formula is:

[0081] Score(x i ) = I(x i ; y) - I(x i ; S) (Formula 3)

[0082] where Score(x i ) is the score (i.e., the difference).

[0083] (6) Iteratively calculate until the stop condition is met: The larger the importance score of the feature index, the stronger the correlation between the feature and the network performance experience. In some embodiments of the present application, 20 main features are sufficient to meet the requirements of analysis and modeling. The top 20 features with the importance scores (i.e., the target indicators) can be used for model prediction, reducing data redundancy and improving the prediction efficiency of the model.

[0084] (7) Repeat steps (3) to (5) until 20 feature indicators are selected, and then output the final feature indicator selection set S.

[0085] After determining multiple target metrics from network performance data, the following steps can also be performed: divide each metric in the target metrics into different metric types, where different metric types are used to reflect different dimensions affecting the network performance experience; determine the weight corresponding to each metric in the target metrics according to the metric type.

[0086] In some embodiments of the present application, the target metrics can be classified from the perspective of business impact. For example, the target metrics can be divided into 7 types of dimensions such as resource load, call access, mobility management, data traffic, service integrity, air interface performance, and base station alarm, making the model more interpretable.

[0087] The division of metric types helps to understand network performance from different dimensions and ensures that the prediction model can comprehensively consider various factors in the network.

[0088] Weight assignment is to give appropriate attention to each metric type according to the degree of influence of different dimensions on the user experience in the construction of the prediction model, so that the model prediction result is more accurate and interpretable. For example, based on the features selected by the mRMR algorithm, further analyze the correlation between different metric types and the network performance experience, such as determining its weight by calculating the mutual information value between the resource load type metric and the network experience.

[0089] In some embodiments of the present application, as shown in Table 1, the target metrics can be classified as follows:

[0090] Table 1: Target Metric Classification Table.

[0091]

[0092] Step S206, use the prediction model to predict the target metrics to obtain a prediction result, where the prediction result is used to quantitatively represent the network performance of the target area.

[0093] In the above step S206, the prediction model is trained using the mRMR-BiLSTM algorithm based on historical network performance data and target metrics to capture the time series characteristics of network performance. In the model training stage, the system takes the target metrics as input and the actual data of the network performance experience as output. By optimizing the model parameters, the model can learn the relationship between the features and the network performance experience, so as to realize the prediction of network performance.

[0094] In some embodiments of the present application, the prediction model includes an input layer, a first hidden layer, a second hidden layer, and an output layer, where the first hidden layer and the second hidden layer are configured with the same number of neurons.

[0095] When using a prediction model to predict a target metric, the following steps are included: The input layer converts the target metric into time series data, obtaining a feature vector corresponding to each moment, where each feature vector includes the values of all target metrics at the corresponding moment; The first hidden layer processes the feature vector corresponding to each moment in sequence from the start point to the end point of the time series data, obtaining a first hidden state corresponding to each moment, where the first hidden state includes the target metric information from the start point of the time series data to the current moment; The second hidden layer processes the feature vector corresponding to each moment in sequence from the end point to the start point of the time series data, obtaining a second hidden state corresponding to each moment, where the second hidden state includes the target metric information from the current moment to the end point of the time series data; The output layer integrates the first hidden state and the second hidden state corresponding to each moment, obtains a target feature vector, and determines a prediction result based on the target feature vector.

[0096] Given that performance data has temporal and periodic characteristics, in some embodiments of the present application, a BiLSTM model is selected to capture these characteristics. The BiLSTM (Bidirectional Long Short-Term Memory) network is an extension of the LSTM network, consisting of two LSTM layers, capable of simultaneously processing past and future information in sequence data. The forward LSTM layer processes historical data to capture past information of the time series; the reverse LSTM layer processes the reverse sequence of the same data to obtain future information. The two LSTM layers can be configured with the same number of neurons to ensure balanced information processing. After obtaining the outputs of the forward and reverse LSTM layers, these outputs are linearly fused according to specific weights. This fusion process synthesizes information from both directions and generates the final prediction result.

[0097] In some embodiments of the present application, the prediction model is constructed in the following manner:

[0098] (1) Construct a dataset: Collect the network performance metrics of the region in the past 3 months. The network performance metrics include 20 metrics in 7 dimensions such as resource load, call access, and mobility management. The data granularity is hourly. 70% of the sample data is divided into a training set, and 30% is divided into a test set.

[0099] (2) Train the model: Figure 3 It is a schematic diagram of the prediction model structure of a network performance prediction method according to an embodiment of the present application. As Figure 3 shown, the feature data (such as the target metric) is respectively input into the hidden state h i of the forward LSTM (the first hidden layer) and the hidden layer h j of the backward LSTM (the second hidden layer); The hidden state h i of the forward LSTM and the hidden layer hj They are connected together and fed into the output layer for prediction; the BiLSTM model outputs h; the output layer introduces the output h of the BiLSTM into the sigmod activation function and finally outputs the trained model data (such as the prediction result).

[0100] (3) Evaluate the model: Through continuous optimization in the model training stage, capture the changing trend of network performance metrics. For example, model evaluation metrics such as RMSE, MAPE, and MAE can be used to comprehensively evaluate the prediction accuracy and stability of the model to ensure the reliability and effectiveness of the model in practical applications.

[0101] To determine the optimal parameters of the model, in the BiLSTM model, the optimizer Adam is selected and the maximum number of training times is set to 300. To ensure prediction accuracy, the prediction model also includes a third hidden layer, where the third hidden layer is used to integrate the first hidden state and the second hidden state corresponding to each moment. The third hidden layer includes a first sub-layer, a second sub-layer, and a third sub-layer. The first sub-layer includes a first number of neurons, and the second sub-layer and the third sub-layer include a second number of neurons, where the first number is less than the second number. For example, a 3-layer structure is designed with 10, 20, and 20 hidden unit data for each layer.

[0102] Meanwhile, a dropout layer can be added after the BiLSTM. Through verification of different parameter models, the model has the best effect when the dropout rate is 0.3.

[0103] (4) Model iteration: Integrate the model into the production environment, and implement a continuous performance monitoring and evaluation mechanism for the regional network performance prediction model to regularly check the prediction accuracy of the model. Based on the latest collected network performance data, fine-tune and optimize the model parameters to ensure that the prediction model can accurately predict the network status and provide the most timely and accurate data support for network planning and resource allocation, thereby continuously improving the network service quality.

[0104] In the case of dividing the target area into multiple sub-areas, the following method can be used to predict the network performance of the sub-areas: map the network performance data to the sub-areas of the target area to obtain the sub-network performance data corresponding to each sub-area; determine the sub-target indicators corresponding to each sub-area from the sub-network performance data; use the prediction model to predict the sub-target indicators to obtain the prediction results corresponding to each sub-area.

[0105] (1) Map network performance data to sub-regions of the target area: The GIS (Geographic Information System) technology can be used to locate and divide network performance data in the geographical space to ensure that the network performance data of each sub-region corresponds to its geographical coordinates. For example, the target area can be divided into multiple sub-regions according to a preset grid size (such as 20 meters × 20 meters), and the data in the national network performance geographical database can be mapped to the corresponding sub-regions based on the geographical coordinates of each data point to form a sub-network performance data set for each sub-region.

[0106] (2) Determine the sub-target indicators corresponding to each sub-region from the sub-network performance data: Refer to the feature selection process in step S204, but at the sub-region level, that is, calculate the mutual information between each indicator in the sub-network performance data and the network performance experience, and select the top 20 indicators with the highest mutual information values as the sub-target indicators, taking into account the characteristics of the network performance data within the sub-region and local influencing factors.

[0107] (3) Predict the sub-target indicators using the prediction model: Apply the mRMR-BiLSTM prediction model to the sub-target indicators of each sub-region to predict future network performance according to historical data. Considering the spatio-temporal characteristics of network performance, ensure the accuracy and practicality of the prediction results. For example, input the sub-target indicators of each sub-region into the trained mRMR-BiLSTM model for network performance prediction; the model outputs the specific predicted values of the network performance within the next 7 days for each sub-region, as well as possible network performance experience levels, such as excellent, good, medium, poor, etc., providing zoned performance prediction results for network operation and maintenance and optimization.

[0108] After obtaining the prediction results corresponding to each sub-region, the following steps can also be executed: Determine the result type corresponding to the prediction result of each sub-region, and repopulate the sub-regions corresponding to the target area on the map according to the result type, where each result type corresponds to a filling color.

[0109] Display the comprehensive analysis and prediction results of the regional network performance experience in an intuitive form on the GIS map. This can not only help decision-makers better understand the spatial distribution characteristics and change trends of network performance, but also significantly improve their decision-making efficiency and accuracy in key stages such as network planning, deployment, and operation and maintenance. In this way, the configuration of network resources can be effectively optimized, unnecessary cost expenditures can be reduced, and at the same time, users can be ensured to obtain a more stable and high-speed network service experience.

[0110] For example, Figure 4 is a schematic diagram of the prediction results of a network performance prediction method according to an embodiment of the present application, as Figure 4As shown, the model can be used to evaluate the performance experience of the regional network, and the evaluation results can be processed by GIS rasterization to visually display the situation of the regional network experience score. The predicted results of the regional network experience are divided into four types: excellent (dark green), good (light green), medium (yellow), and poor (red), and are divided by color.

[0111] Through the above steps S202 to S206, by determining the indicators highly related to the network performance experience from the data of multiple base station cells collected from multiple systems and using the prediction model for prediction, the purpose of improving the efficiency and accuracy of network performance prediction is achieved, thus realizing the technical effect of effectively improving the network service quality and enhancing the user experience, and further solving the technical problem that the network performance prediction technology adopted in the related technology only focuses on the performance status of a single network device, and the prediction indicators adopted are single, and cannot comprehensively reflect the overall network usage experience of users, resulting in inaccurate prediction results.

[0112] Figure 5 is the overall flowchart of a method for predicting network performance according to an embodiment of the present application. As Figure 5 shown, in some embodiments of the present application, the network performance can be predicted through the following steps:

[0113] Step S502: Construct a data set.

[0114] (1) Collect original data such as performance and high alarms: Collect performance index data and alarm information of base station cells from multiple systems such as 5GR, integrated dispatching, mobile command center, and big data lake, covering data in multiple dimensions such as resource load, call access, mobility management, data traffic, service integrity, air interface performance, and base station alarms, to ensure the comprehensiveness of the data.

[0115] (2) Data preprocessing: Clean, check, and convert the format of the collected original data. This includes removing invalid data, handling missing values, unifying the data format, etc., to ensure the data quality and consistency, and establish a good data foundation for subsequent analysis.

[0116] (3) Construct a network performance geographic database: Associate the preprocessed data with GIS geographic information to form a database containing geographic coordinates, network performance indicators, and alarm information. This step maps the network performance data to the geographic space for subsequent regional analysis and prediction.

[0117] (4) Collect data and establish a sample set: Extract data from the past 3 months from the network performance geographic database to form a sample set for model training and testing. The data granularity is hourly to ensure that the sample set covers the periodic changes of network performance.

[0118] Step S504: Feature engineering.

[0119] (1) Screening network experience features using the mRMR algorithm: From over 200 network performance metrics, calculate the mutual information between each metric and network performance experience based on the mRMR algorithm, and screen out 20 feature metrics most relevant to user experience.

[0120] (2) Ranking the feature importance and selecting TOP20: Sort all metrics according to the calculated mutual information values, and select the top 20 metrics as the feature set S. These metrics have the highest correlation and the lowest redundancy, and are the preferred inputs for constructing the prediction model.

[0121] (3) Obtaining the feature set S: After the feature engineering steps, finally obtain the set S containing 20 key features as the input variables for subsequent model training. These features will be used to predict the regional network performance experience.

[0122] Step S506: Data modeling.

[0123] (1) Constructing a BiLSTM regional network performance experience prediction model: According to the feature set S, design and construct a BiLSTM model, and use the bidirectional LSTM structure to process time series data to better capture the forward and backward dependencies of network performance metrics.

[0124] (2) Training the model based on the sample set: Divide the sample set into a 70% training set and a 30% test set, and use the training set data to train the BiLSTM model, and adjust the model parameters to optimize the prediction effect.

[0125] (3) Evaluating the model: Use the test set data to evaluate the prediction accuracy of the model, and measure the model performance through metrics such as RMSE (Root Mean Square Error), MAPE (Mean Absolute Percentage Error), and MAE (Mean Absolute Error).

[0126] (4) Judging whether the prediction result meets the expectation. If not, continue training; if so, obtain the regional network performance experience prediction model: According to the performance of the model on the test set, if the prediction result deviates greatly from the actual network experience, return to the training stage to adjust the model parameters; if the prediction result is satisfactory, output the trained regional network performance experience prediction model.

[0127] Step S508: Model iteration.

[0128] After the model is deployed to the production environment, continuously collect actual network performance data for performance monitoring and effectiveness evaluation of the model. According to the latest data feedback and model evaluation results, regularly perform fine-tuning and optimization of the model's parameters to ensure that the model can adapt to changes in the network environment and maintain prediction accuracy. For example, through model iteration, the changing patterns of network performance can be continuously learned, improving the generalization ability of the model and ensuring that the model can stably and accurately predict the regional network performance experience in the long term, providing continuous data support for network operation and maintenance, planning, and resource allocation.

[0129] This application first uses the mRMR algorithm for feature selection in the prediction model, overcoming the problem of feature redundancy commonly existing in related technologies. The mRMR algorithm can not only screen out the features that have the greatest impact on the network performance experience but also ensure that these features have the minimum redundancy with each other, thus constructing a more refined and efficient feature set, significantly improving the prediction accuracy and interpretability of the model. Moreover, different from related technologies that often focus on the performance prediction of a single base station or device, this application innovatively proposes to conduct comprehensive analysis and prediction of network performance in units of geographical regions, making the prediction results more comprehensive and able to consider the interactions and coverage overlaps among multiple base station cells within the region.

[0130] At the same time, this application uses BiLSTM to be able to process past and future information simultaneously, capture the periodicity and trend of network performance indicators, enhance the model's adaptability to long-term prediction, and be able to more accurately predict future changes in network performance. In addition, this application displays the prediction results in an intuitive graphical form on the GIS map, supporting custom circle selection on the map, providing a clear and easy-to-understand network performance view for decision-makers. Most related technologies focus on the performance prediction at the network device level, while this application starts from the perspective of user experience, comprehensively analyzes network performance, and predicts the network experience of users in a specific geographical region. This user-oriented prediction method can more directly reflect the network service quality and is conducive to improving user satisfaction.

[0131] This application can provide in-depth insights and accurate guidance for scenarios such as network planning and optimization, fault detection and repair, user experience improvement, resource management and cost control, and network guarantee.

[0132] (1) First, when performing network planning and optimization, this application can help operators deeply understand the trends and fluctuations of network performance in a specific geographical region, providing data-driven decision-making basis for base station site selection, frequency planning, capacity upgrade, etc. Through the prediction model, managers can foresee possible network congestion and performance degradation in specific regions in the future, such as large event venues and densely populated areas, and thus plan network expansion and optimization measures in advance to ensure the continuous stability of network services.

[0133] (2) Secondly, for the detection and repair of network failures, this application can warn of potential problems before failures occur by real-time monitoring and predicting network performance metrics, allowing the maintenance team to intervene promptly and prevent large-scale service interruptions. For example, when the prediction model indicates that the alarm metrics of a base station in a certain area are abnormally rising, which may signal an impending hardware failure, the operation and maintenance personnel can arrange preventive inspections and repairs accordingly, reducing network downtime and the decline in user experience caused by failures.

[0134] (3) Furthermore, from the perspective of improving user experience, this application can accurately identify poor-quality grids and areas with poor network experience by comprehensively analyzing and predicting network performance, and then prescribe the right remedy, specifically optimizing network configuration or resource scheduling to enhance the network speed and stability perceived by users. For example, if the prediction model finds that the network experience in certain areas deteriorates during the evening peak hours, the operator can adjust the resource allocation strategy, increase the uplink bandwidth or optimize the scheduling algorithm to improve the user experience during this period.

[0135] (4) Also, in terms of resource management and cost control, this application can assist the operator in more efficiently allocating network resources on the premise of ensuring service quality, avoiding over-investment and resource waste. For example, if the predictive analysis shows that the network resources in a certain area are underutilized, the operator can consider adjusting the base station configuration in that area or reallocating the resources to areas with higher demand, thereby achieving optimal utilization of resources and reducing the costs of network construction and operation and maintenance.

[0136] (5) Finally, in high-load and unpredictable situations such as major events and natural disasters, by predicting potential challenges to network performance in advance, the operator can formulate emergency communication plans, enhance the network bearing capacity in key areas, and ensure unimpeded critical communications, such as measures like urgently dispatching temporary base stations and activating backup routes, to improve the elasticity and disaster resistance of the network.

[0137] Figure 6 is a structural diagram of a network performance prediction device according to an embodiment of the present application, as Figure 6 shown, the device includes:

[0138] An acquisition module 602, configured to acquire network performance data corresponding to a target area;

[0139] A determination module 604, configured to determine a plurality of target metrics from the network performance data;

[0140] A prediction module 606, configured to predict the target metrics using a prediction model to obtain a prediction result.

[0141] It should be noted that Figure 6 the network performance prediction device shown is used to execute Figure 2The prediction method of network performance shown, thus Figure 2 The relevant explanations in the prediction method of network performance in Figure 6 also apply to the prediction device of network performance shown, which will not be elaborated here.

[0142] The embodiments of the present application also provide an electronic device, which includes a memory and a processor. Among them, the memory is used to store program instructions; the processor is connected to the memory and is used to execute the steps of implementing the prediction method of network performance in various embodiments of the present application.

[0143] For example, the processor executes the following functions by executing the program instructions stored in the memory:

[0144] Obtain network performance data corresponding to the target area, where the network performance data includes data of multiple base station cells collected from multiple systems; determine multiple target metrics from the network performance data, where the target metrics include metrics that are relevant to the network performance experience, and the network performance experience is used to quantitatively represent the satisfaction of users in the target area with the network service; use the prediction model to predict the target metrics to obtain a prediction result, where the prediction result is used to quantitatively represent the network performance of the target area.

[0145] The embodiments of the present application also provide a non-volatile storage medium, which includes a stored computer program. Among them, the device where the non-volatile storage medium is located executes the steps of the prediction method of network performance in various embodiments of the present application by running the computer program.

[0146] The embodiments of the present application also provide a computer program product, including computer instructions, which implement the steps of the prediction method of network performance in various embodiments of the present application when executed by a processor.

[0147] The embodiments of the present application also provide a computer program, which implements the steps of the prediction method of network performance in various embodiments of the present application when executed by a processor.

[0148] The serial numbers of the above embodiments of the present application are only for description and do not represent the advantages and disadvantages of the embodiments.

[0149] In the above embodiments of the present application, the descriptions of each embodiment have their own focuses. For the parts not elaborated in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0150] In several embodiments provided by 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 merely illustrative. For example, the division of the units can be a logical function division. In actual implementation, there can be other division methods. For example, 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 couplings, direct couplings, or communication connections shown or discussed with each other can be through some interfaces. The indirect couplings or communication connections of units or modules can be in electrical or other forms.

[0151] The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they can be located in one place or distributed to multiple units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0152] In addition, in each embodiment of this application, the functional units can be integrated into a processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0153] If the above-mentioned integrated units are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable 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 methods described in each embodiment of this application. The foregoing storage medium includes: USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical discs and other various media that can store program codes.

[0154] The above is only the preferred embodiment of this application. It should be noted that for those of ordinary skill in the art, without departing from the principle of this application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of this application.

Claims

1. A method for predicting network performance, characterized in that Including: Obtain network performance data corresponding to a target area, where the network performance data includes data of multiple base station cells collected from multiple systems; Determine multiple target metrics from the network performance data, where the target metrics include metrics that are relevant to the network performance experience, and the network performance experience is used to quantitatively represent the satisfaction degree of users in the target area with respect to network services; Use a prediction model to predict the target metrics to obtain a prediction result, where the prediction result is used to quantitatively represent the network performance of the target area.

2. The method according to claim 1, wherein Determine multiple target metrics from the network performance data, including: Determine a first metric set corresponding to the network performance data, where the first metric set includes all network performance metrics corresponding to the network performance data; Calculate the mutual information value between a first metric and the network performance experience, where the first metric is any metric in the first metric set, and the mutual information value is used to quantitatively represent the correlation between the first metric and the network performance experience; Add the first metric corresponding to the highest mutual information value to a second metric set; Determine the target metrics according to the first metric set and the second metric set.

3. The method according to claim 2, wherein Determine the target metrics according to the first metric set and the second metric set, including: Calculate the average mutual information value between the second metric set and a third metric, where the third metric is any first metric in the first metric set other than the second metric set, and the average mutual information value is used to quantitatively represent the information redundancy degree between the third metric and the second metric set; Determine the difference between the mutual information value corresponding to each third metric and the average mutual information value corresponding to each third metric, and add the third metric corresponding to the largest difference to the second metric set; Repeat the calculation process of the mutual information value and the average mutual information value until the second metric set meets a preset condition; Determine the second metric set as the target metrics.

4. The method according to claim 1, characterized in that, The prediction model includes an input layer, a first hidden layer, a second hidden layer, and an output layer, where the first hidden layer and the second hidden layer are configured with the same number of neurons; using the prediction model to predict the target metrics to obtain a prediction result, including: The input layer converts the target metrics into time series data to obtain a feature vector corresponding to each moment, where each feature vector includes the values of all the target metrics at the corresponding moment; The first hidden layer processes the feature vector corresponding to each moment in sequence along the direction from the start point to the end point of the time series data to obtain a first hidden state corresponding to each moment, where the first hidden state includes the target metric information corresponding to the time series data from the start point to the current moment; The second hidden layer processes the feature vectors corresponding to each moment in sequence from the end point to the start point of the time series data, and obtains a second hidden state corresponding to each moment, where the second hidden state includes target index information corresponding to the current moment to the end point of the time series data; The output layer integrates the first hidden state and the second hidden state corresponding to each moment to obtain a target feature vector, and determines the prediction result according to the target feature vector.

5. The method according to claim 4, wherein The prediction model further includes a third hidden layer, where the third hidden layer is used to integrate the first hidden state and the second hidden state corresponding to each moment.

6. The method according to claim 5, wherein The third hidden layer includes a first sub-layer, a second sub-layer, and a third sub-layer. The first sub-layer includes a first number of neurons, and the second sub-layer and the third sub-layer include a second number of neurons, where the first number is less than the second number.

7. The method according to claim 1, wherein The target area is determined by the following method: Obtain a first vertex selected by the target object on the map, where the first vertex is the starting vertex of the target area determined from the initial area; Detect a second vertex set selected by the target object on the map, where the second vertex set is used to determine the shape of the target area; When the first vertex is included in the second vertex set, obtain the first longitude and latitude coordinate information corresponding to each vertex in the second vertex set; Determine the closed area formed by the first longitude and latitude coordinate information as the target area.

8. The method according to claim 7, characterized in that, After determining the closed area formed by the first longitude and latitude coordinate information as the target area, the method further includes: Divide the initial area into multiple grids, where the sizes of the multiple grids are the same; Obtain the second longitude and latitude coordinate information corresponding to the second center point of each grid, and generate an identifier corresponding to each grid according to the second longitude and latitude coordinate information; Determine all the identifiers included in the target area to obtain an identifier set; Determine the grid corresponding to each identifier in the identifier set as a sub-area of the target area.

9. The method according to claim 8, wherein After obtaining the network performance data corresponding to the target area, the method further includes: Map the network performance data to the sub-areas of the target area to obtain sub-network performance data corresponding to each sub-area; Determine sub-target indicators corresponding to each sub-area from the sub-network performance data; Use the prediction model to predict the sub-target indicators to obtain prediction results corresponding to each sub-area.

10. The method according to claim 9, wherein After obtaining the prediction results corresponding to each sub-area, the method further includes: determining the result type corresponding to the prediction result of each sub-area, and repainting the sub-areas corresponding to the target area on the map according to the result type, where each result type corresponds to a filling color.

11. The method according to claim 1, wherein After determining multiple target indicators from the network performance data, the method further includes: Divide each of the target indicators into different indicator types, where different indicator types are used to reflect different dimensions affecting the network performance experience; Determine the weight corresponding to each index in the target index according to the index type.

12. A prediction device for network performance, characterized in that, Including: An acquisition module, configured to acquire network performance data corresponding to a target area, where the network performance data includes data of a plurality of base station cells collected from a plurality of systems; A determination module, configured to determine a plurality of target indexes from the network performance data, where the target indexes include indexes related to network performance experience, and the network performance experience is used to quantitatively represent the satisfaction degree of users in the target area with network services; A prediction module, configured to predict the target index by using a prediction model to obtain a prediction result, where the prediction result is used to quantitatively represent the network performance of the target area.

13. An electronic device, characterized in that, Including: A memory and a processor, where the memory is configured to store program instructions; the processor is connected to the memory and is configured to execute the prediction method for network performance according to any one of claims 1 to 11.

14. A non-volatile storage medium, characterized in that, The non-volatile storage medium includes a stored computer program, where the device where the non-volatile storage medium is located executes the prediction method for network performance according to any one of claims 1 to 11 by running the computer program.

15. A computer program product, comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, the prediction method for network performance according to any one of claims 1 to 11 is implemented.

Images