Link map prediction method and device, computer equipment and storage medium

Abstract

The invention belongs to the field of artificial intelligence and financial science and technology, and relates to a link atlas prediction method, which comprises the steps of obtaining a reference link atlas of a first target time period and a target operation and maintenance data set of a specific time period in a second target time period, and denoising the target operation and maintenance data set to obtain an enhanced operation and maintenance data set; calculating a correlation coefficient of the enhanced operation and maintenance data set to obtain a correlation coefficient map; and inputting the reference link atlas and the correlation coefficient atlas into a trained link atlas prediction model for prediction to obtain a target link atlas in the second target time period. The invention further provides a link map prediction device, computer equipment and a storage medium. In addition, the invention also relates to a block chain technology, and the target link map can be stored in a block chain. According to the invention, the accuracy and reliability of data can be improved, the link map is updated in advance, and the map prediction precision is improved.

Description

Technical Field

[0001] The present application relates to the fields of artificial intelligence and financial technology, and in particular to a link graph prediction method, apparatus, computer equipment and storage medium. Background Art

[0002] With the development of science and technology, especially driven by big data and artificial intelligence technologies, all walks of life are exploring how to effectively use these technologies to improve business efficiency and decision-making quality. In the field of operation and maintenance, the quality and reliability of data are the cornerstones to ensure the stable operation of the system. High-quality operation and maintenance data can help enterprises better understand and predict the behavior of the system, thereby optimizing maintenance strategies and reducing system failures. The intelligent operation and maintenance link map is an effective tool. It is a technology used for intelligent operation and maintenance (AIOps). It analyzes IT operation and maintenance data by integrating big data and machine learning capabilities to help enterprises achieve more efficient operation and maintenance management. At the same time, it helps to fully understand the complexity of the server environment and provides key information for real-time monitoring, troubleshooting and resource optimization.

[0003] However, the current intelligent operation and maintenance link map remains unchanged within a fixed time period and cannot be updated in advance, resulting in a lag in map updates. The old map may not accurately reflect the current system status, and inaccurate maps will also cause great difficulties in completing operation and maintenance tasks; secondly, the current operation and maintenance data often contains a lot of noise, which affects the quality of the data and the accuracy of subsequent analysis, and will have a very adverse impact on subsequent data analysis decisions and the establishment of intelligent operation and maintenance link maps; in addition, since the IT environment is dynamically changing, servers and network equipment may change over time, making the prediction map need to consider timing and dynamics, which increases the difficulty of prediction. Summary of the invention

[0004] The purpose of the embodiments of the present application is to propose a link map prediction method, device, computer equipment and storage medium to solve the technical problems in the related technology that the link map update is delayed and the operation and maintenance data contains a large amount of noise, resulting in inaccurate link map establishment and increased prediction difficulty.

[0005] In order to solve the above technical problems, the present application embodiment provides a link graph prediction method, which adopts the following technical solution:

[0006] Acquire a reference link graph of a first target time period and a target operation and maintenance data set for a specific time period in a second target time period, and denoise the target operation and maintenance data set to obtain an enhanced operation and maintenance data set;

[0007] Calculate the correlation coefficient of the enhanced operation and maintenance data set to obtain a correlation coefficient map;

[0008] The reference link map and the correlation coefficient map are input into a trained link map prediction model for prediction to obtain a target link map for the second target time period.

[0009] Furthermore, the step of denoising the target operation and maintenance data set to obtain an enhanced operation and maintenance data set includes:

[0010] Preprocessing all target operation and maintenance data in the target operation and maintenance data set to obtain a preprocessed operation and maintenance data set;

[0011] Converting the preprocessed operation and maintenance data set into time series operation and maintenance data according to a preset format;

[0012] Analyzing the noise characteristics of the time series operation and maintenance data, and determining Gaussian filter parameters based on the noise characteristics;

[0013] According to the Gaussian filter parameters, a Gaussian filter algorithm is used to denoise the time series operation and maintenance data to obtain an enhanced operation and maintenance data set.

[0014] Furthermore, the step of calculating the correlation coefficient of the enhanced operation and maintenance data set to obtain a correlation coefficient map includes:

[0015] Calculate the correlation coefficient of each pair of nodes in the enhanced operation and maintenance data set, and obtain a correlation heat map according to the correlation coefficient;

[0016] The weakly connected nodes in the enhanced operation and maintenance data set are filtered based on the correlation heat map to obtain a correlation coefficient map.

[0017] Furthermore, the link graph prediction model includes a convolution layer, a pooling layer, a graph fusion layer, a multi-layer perception layer and a prediction layer; the step of inputting the reference link graph and the correlation coefficient graph into the trained link graph prediction model for prediction to obtain the target link graph for the second target time period includes:

[0018] Inputting the reference link map and the correlation coefficient map into the convolution layer for feature extraction to obtain operation and maintenance map features;

[0019] Performing a pooling operation on the operation and maintenance graph features through the pooling layer to obtain pooled graph features;

[0020] The pooled graph features are fused through the graph fusion layer to obtain operation and maintenance node fusion features;

[0021] Inputting the operation and maintenance node aggregation features and the pooled graph features into the multi-layer perception layer for weighted calculation to obtain a weighted feature matrix;

[0022] The weighted feature matrix is ​​input into the prediction layer for prediction to obtain a target link graph.

[0023] Furthermore, the step of fusing the pooled graph features through the graph fusion layer to obtain the operation and maintenance node fusion features includes:

[0024] The graph fusion layer transmits and aggregates the neighborhood node feature information of the pooled graph features to obtain node aggregation features;

[0025] The node aggregation features are calculated using a self-attention mechanism to obtain operation and maintenance node fusion features.

[0026] Furthermore, the prediction layer includes a forward hidden layer, a backward hidden layer and a classification layer; the step of inputting the weighted feature matrix into the prediction layer for prediction to obtain the target link map includes:

[0027] Inputting the weighted feature matrix into the forward hidden layer and the backward hidden layer respectively to transfer and update feature information, and obtaining forward hidden features and backward hidden features respectively;

[0028] The forward hidden feature and the backward hidden feature are concatenated according to corresponding positions to obtain a hidden state;

[0029] The hidden state is classified and predicted by the classification layer to obtain a target link map.

[0030] Furthermore, before the step of inputting the reference link graph and the correlation coefficient graph into the trained link graph prediction model for prediction, the method further includes:

[0031] Acquire a sample data set, the sample data set including a historical link map of a first historical time period and a historical correlation coefficient map of a second historical time period as a historical operation and maintenance data pair, wherein the historical operation and maintenance data pair carries a real link map of the corresponding second historical time period;

[0032] Using the sample data set to train a pre-built initial link graph prediction model to obtain a predicted link graph;

[0033] The loss value between the predicted link graph and the actual link graph is calculated according to the loss function, the model parameters are adjusted based on the loss value, and the training is continued until the training is completed to obtain the final link graph prediction model.

[0034] In order to solve the above technical problems, the embodiment of the present application further provides a link graph prediction device, which adopts the following technical solution:

[0035] An acquisition module, used to acquire a reference link graph of a first target time period and a target operation and maintenance data set for a specific time period in a second target time period, and denoise the target operation and maintenance data set to obtain an enhanced operation and maintenance data set;

[0036] A calculation module, used to calculate the correlation coefficient of the enhanced operation and maintenance data set to obtain a correlation coefficient map;

[0037] The prediction module is used to input the reference link map and the correlation coefficient map into a trained link map prediction model for prediction, so as to obtain a target link map for the second target time period.

[0038] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0039] The computer device includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the link graph prediction method as described above when executing the computer-readable instructions.

[0040] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0041] The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the steps of the link graph prediction method described above are implemented.

[0042] Compared with the prior art, this application has the following beneficial effects:

[0043] The present application provides a link map prediction method, which denoises a target operation and maintenance data set of a specific time period in a second target time period, calculates the correlation coefficient of the denoised target operation and maintenance data set, obtains a corresponding correlation coefficient map, and predicts the target link map of the second time period through a link map prediction model based on the correlation coefficient map and the reference link map; the present application improves the accuracy and reliability of data through denoising, and can update the link map in advance by combining the correlation coefficient map and the reference link map, and more effectively perform operation and maintenance tasks. At the same time, the map prediction accuracy is improved through the link map prediction model, and the future intelligent operation and maintenance link map can be fully and accurately predicted, which is very helpful to quickly locate problems, improve system availability and reliability, reduce fault repair time, and optimize resource utilization efficiency, providing a scientific basis and direction for subsequent operation and maintenance strategy formulation and performance optimization, helping the operation and maintenance team to formulate more effective measures to improve system performance and user experience, and can support more business expansion and technology upgrades in the future. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0045] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0046] Figure 2 is a flowchart of an embodiment of a link graph prediction method according to the present application;

[0047] Figure 3 yes Figure 2 A flowchart of a specific implementation of step S203;

[0048] Figure 4 is a schematic structural diagram of an embodiment of a link graph prediction device according to the present application;

[0049] Figure 5 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0051] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0052] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0053] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0054] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0055] This application provides a link graph prediction method that can be applied to Figure 1 In the system architecture 100 shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0056] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0057] Terminal devices 101, 102, 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, etc.

[0058] The server 105 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal devices 101 , 102 , and 103 .

[0059] It should be noted that the link graph prediction method provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the link graph prediction device is generally arranged in the server / terminal device.

[0060] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0061] Continue to refer Figure 2 , shows a flowchart of an embodiment of a link graph prediction method according to the present application, comprising the following steps:

[0062] Step S201, obtaining a reference link graph of a first target time period and a target operation and maintenance data set of a specific time period in a second target time period, and denoising the target operation and maintenance data set to obtain an enhanced operation and maintenance data set.

[0063] Among them, the reference link map is the intelligent operation and maintenance link map of the first target time period. The intelligent operation and maintenance link map is a technology used in intelligent operation and maintenance (AIOps). It combines big data and machine learning capabilities to analyze and manage IT operation and maintenance data to improve operation and maintenance efficiency and reduce costs. It includes root cause analysis of system abnormal events, abnormal alarm convergence, root cause location methods of business abnormalities, link tracking and topology analysis, etc.

[0064] The intelligent operation and maintenance link map includes but is not limited to monitoring data, log data (Log), indicator data (Metric), link data (Trace), alarm data (Alarm), change data (Change), configuration data, business data, and user and entity behavior data. Among them, monitoring data includes real-time monitoring data of network devices, servers, applications, etc., such as CPU usage, memory usage, network traffic, response time, etc.; log data is the log information generated when the system, application and service are running, which records operations, errors, warnings and other important events; indicator data is key performance indicators used to measure the performance of systems and applications, such as transaction volume, success rate, response time, system load, etc.; link data records the flow path of requests in the system, including Alarm data refers to the alarm information generated when an abnormality occurs in the system or application, which is used to quickly locate the problem; change data is used to record configuration changes, software updates, hardware replacements and other information in the system, which helps to analyze the impact of changes on the system operation status; configuration data is data from the configuration management database (CMDB), which contains configuration information of network devices, servers, applications, etc. and the relationship between them; business data is data related to business processes and transactions, such as transaction volume, transaction success rate, transaction time, etc., which is used to evaluate business impact; user and entity behavior data is user operation behavior data and entity (such as equipment, service) behavior pattern data, which is used to analyze the impact of user behavior on the system and the normal and abnormal behavior patterns of entities.

[0065] The intelligent operation and maintenance link map integrates and analyzes this data to help the operation and maintenance team more effectively monitor, analyze and respond to problems in the IT environment, and achieve automated and intelligent operation and maintenance management.

[0066] For example, in the financial industry, intelligent operation and maintenance link maps can be used in scenarios such as abnormal alarms, alarm convergence, fault analysis, and trend prediction. For example, WeBank uses intelligent operation and maintenance link maps to achieve rapid fault identification and root cause location through machine learning algorithms combined with rules, significantly improving operation and maintenance efficiency; another example is a large joint-stock bank, which has achieved multi-node correlation analysis, panoramic perspective of alarm events, and configuration change impact analysis by building intelligent operation and maintenance link maps, thereby improving the accuracy and timeliness of fault location.

[0067] In this embodiment, the first target time period is the t-1th day, the second target time period is the tth day, and the specific time period in the second target time period is the first hour of the tth day. For example, if a link map for July 20, 2024 is to be generated, the first target time period should be from 00:00 to 23:59 on July 19, 2024, the second target time period is from 00:00 to 23:59 on July 20, 2024, and the time range of the specific time period of the second target time period is from 00:00 to 01:00 on October 2, 2023.

[0068] In this embodiment, the intelligent operation and maintenance link map relies on the support of operation and maintenance data. The operation and maintenance data provides the original data and information required to build the link map. Accordingly, the target operation and maintenance data set includes but is not limited to monitoring data, log data, indicator data, link data, alarm data, change data, configuration data, business data, and user and entity behavior data. Each type of data contains a corresponding timestamp.

[0069] In this embodiment, the reference link map and the target operation and maintenance data set can be obtained through a wired connection or a wireless connection. It should be noted that the above wireless connection method may include but is not limited to 3G / 4G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future.

[0070] The acquired target operation and maintenance data set is denoised to obtain an enhanced operation and maintenance data set after smoothing. The denoising methods include mean filtering, median filtering, wavelet denoising, non-local mean denoising, bilateral filtering, local mean decomposition, iterative filtering decomposition method, synchronous compression transformation, etc.

[0071] In some optional implementations, the step of denoising the target operation and maintenance data set to obtain an enhanced operation and maintenance data set includes:

[0072] All target operation and maintenance data in the target operation and maintenance data set are processed to obtain a preprocessed operation and maintenance data set;

[0073] Convert the preprocessed operation and maintenance data set into time series operation and maintenance data according to the preset format;

[0074] Analyze the noise characteristics of time series operation and maintenance data, and determine Gaussian filter parameters based on the noise characteristics;

[0075] According to the Gaussian filter parameters, the Gaussian filter algorithm is used to denoise the time series operation and maintenance data to obtain an enhanced operation and maintenance data set.

[0076] Among them, all target operation and maintenance data in the target operation and maintenance data set are preprocessed, and the preprocessing includes data cleaning (removing invalid or erroneous data points) and data standardization (scaling the data to a specific range, such as 0 to 1).

[0077] Get each operation and maintenance data record from the preprocessed operation and maintenance data set, and ensure that each operation and maintenance data record has a corresponding timestamp. If there is no timestamp in the original operation and maintenance data, it may be necessary to assign a timestamp based on the time when the data was generated or other logic; format each operation and maintenance data record into a unified structure, specifically a list or data frame containing a timestamp and data value, sorted by timestamp to form a continuous time series, that is, time-series operation and maintenance data.

[0078] In this embodiment, Gaussian filtering is used to remove noise from the time series operation and maintenance data, wherein Gaussian filtering is a linear filtering method for smoothing operation and maintenance data and removing Gaussian noise. The noise characteristics of the time series operation and maintenance data are analyzed, and the noise characteristics include noise intensity, frequency characteristics, spatial characteristics, time characteristics, and amplitude. The Gaussian filter parameters are determined based on the noise characteristics. The parameters of the Gaussian filter mainly include the filter kernel size (ksize) and the standard deviation (sigma), wherein the filter kernel size determines the smoothness of the filter. Generally, a larger filter kernel can remove noise more effectively, but may cause loss of details of the image or data; the standard deviation determines the shape of the Gaussian filter kernel. The larger the standard deviation, the wider the filter kernel and the stronger the smoothing effect.

[0079] In a specific implementation, the Gaussian filtering process is as follows:

[0080] 1) Data preparation

[0081] Obtain the operation and maintenance data of the first hour of the tth day and convert the operation and maintenance data into time series operation and maintenance data.

[0082] 2) Gaussian kernel (i.e. filter kernel) function generation

[0083] The Gaussian kernel is a two-dimensional matrix whose center is usually located at the center of the matrix, and the standard deviation determines the shape of the Gaussian distribution. The specific formula of the Gaussian function is as follows:

[0084]

[0085] Where x is the deviation of each data point from the center of the time series operation and maintenance data, and σ is the standard deviation of the Gaussian kernel.

[0086] 3) Apply Gaussian filtering:

[0087] For each data point x in the time series operation and maintenance data i Do the following:

[0088] Use a window size of 5*5 (Gaussian kernel) to calculate the Gaussian weight w for each data point in the window j , using the formula:

[0089] Calculate the smoothed data points The value of , using the formula:

[0090] 4) Repeat step 3: Repeat step 3 for each data point in the time series operation and maintenance data to generate a smoothed version of the entire data sequence, thus completing denoising.

[0091] Through Gaussian filtering and denoising, random fluctuations and outliers in the data can be effectively smoothed and removed, resulting in clearer and more stable data representation, improving the quality and reliability of the data, which is very helpful for subsequent analysis and decision-making, as well as the establishment of the Pearson graph.

[0092] Step S202, calculating the correlation coefficient of the enhanced operation and maintenance data set to obtain a correlation coefficient map.

[0093] In this embodiment, after obtaining the enhanced operation and maintenance data set after Gaussian filtering, the linear correlation between the attributes of each node of each operation and maintenance data is analyzed using the correlation coefficient calculation method. Specifically, by calculating the correlation coefficient between each pair of node attributes in the enhanced operation and maintenance data, a correlation coefficient matrix is ​​obtained, and the value of each element in the matrix represents the correlation coefficient between the corresponding two node attributes. Through the obtained correlation coefficient matrix, a correlation coefficient map can be drawn, and the correlation coefficient map can intuitively show the size and direction of the correlation between the attributes of each node. The positive correlation coefficient is expressed as a positive value, and the negative correlation coefficient is expressed as a negative value. The larger the absolute value of the correlation coefficient, the stronger the correlation.

[0094] In some optional implementations of this embodiment, the step of calculating the correlation coefficient of the enhanced operation and maintenance data set to obtain the correlation coefficient map includes:

[0095] Calculate the correlation coefficient of each pair of nodes in the enhanced operation and maintenance data set, and obtain the correlation heat map based on the correlation coefficient;

[0096] Based on the correlation heat map, weakly connected nodes in the operation and maintenance data set are filtered and enhanced to obtain the correlation coefficient map.

[0097] Among them, the correlation coefficient algorithms include Pearson correlation coefficient, Spearman rank correlation coefficient, Kendall-Taub correlation coefficient, cosine similarity, etc.

[0098] In some optional implementations of this embodiment, the Pearson correlation coefficient is used to establish a Pearson correlation map.

[0099] The Pearson correlation coefficient is a statistic used to measure the degree of linear correlation between two variables. It measures the current relationship between the two variables. The Pearson correlation coefficient is used to create a Pearson correlation map. The steps to calculate the node Pearson correlation map using the Pearson correlation coefficient are as follows:

[0100] ① Data normalization

[0101] In order to eliminate the unit differences between different nodes, the Z-score is used to scale the data to a similar range. The formula is as follows:

[0102]

[0103] Among them, X represents the data point to be standardized, μ is the mean of the enhanced operation and maintenance dataset, and σ is the standard deviation of the enhanced operation and maintenance dataset.

[0104] ② Calculate the Pearson correlation coefficient between nodes and obtain the Pearson correlation graph

[0105] For each pair of nodes, the Pearson correlation coefficient between them is calculated using the following formula:

[0106]

[0107] Among them, p ab Indicates the edge relationship between two nodes a and b in the Pearson correlation graph, with a value range of 0-1, x ai and x bi represents the filtered enhanced operation and maintenance data of nodes a and b at the i-th moment in the first hour of the t-th day. and Respectively represent x ai and x bi , n represents the number of data nodes in the enhanced operation and maintenance dataset.

[0108] ③Remove weak connections in the graph

[0109] By setting a threshold, the weak connections in the Pearson correlation graph are filtered out. The formula is as follows:

[0110] p=Relu(p-ε)

[0111] Among them, p is the Pearson correlation map, Relu is the rectified linear unit function, which is used to remove negative connections in the Pearson correlation map; ε is used as a threshold to filter weak connections in the graph, and ε can be adjusted by business personnel based on experience.

[0112] By calculating the Pearson correlation map, we can intuitively display the size and direction of the correlation between the attributes of each node, so as to analyze the key related attributes, which will play an important role in subsequent analysis. In addition, after determining the key related attributes, we can further obtain the business logic and potential operation and maintenance optimization direction behind these attributes, which provides data support and analysis basis for formulating more effective operation and maintenance strategies and improvement measures.

[0113] Step S203: input the reference link graph and the correlation coefficient graph into the trained link graph prediction model for prediction to obtain a target link graph for the second target time period.

[0114] In this embodiment, the Pearson correlation graph is combined with the reference intelligent operation and maintenance link graph provided by the business party through the trained link graph prediction model to predict the intelligent operation and maintenance link graph of the tth day (the second target time period), that is, the target link graph.

[0115] It should be emphasized that in order to further ensure the privacy and security of the target link map, the above-mentioned target link map can also be stored in a node of a blockchain.

[0116] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0117] In some optional implementations, the link graph prediction model includes a convolution layer, a pooling layer, a graph fusion layer, a multi-layer perception layer and a prediction layer. The convolution layer, the pooling layer, the graph fusion layer, the multi-layer perception layer and the prediction layer are connected in sequence, and the input reference operation and maintenance link graph and correlation coefficient graph are processed.

[0118] Among them, see Figure 3 As shown, the step of inputting the reference link graph and the correlation coefficient graph into the trained link graph prediction model for prediction to obtain the target link graph corresponding to the second target time period includes:

[0119] Step S301: Input the reference operation and maintenance link map and the correlation coefficient map into the convolution layer for feature extraction to obtain the operation and maintenance map features.

[0120] The convolution layer can use a common convolution layer or a Ghost convolution module. When the convolution layer uses a common convolution layer, the convolution layer includes at least two convolution sublayers with different numbers of channels. Taking two convolution sublayers as an example, the convolution kernel of the first convolution sublayer is 3×3 and the number of channels is 64, and the convolution kernel of the second convolution sublayer is 3×3 and the number of channels is 128.

[0121] When the Ghost convolution module is adopted in the convolution layer, the Ghost convolution module includes an intrinsic convolution layer, a first ghost convolution layer, a BN layer and a ReLU activation function, a depth-separable convolution layer, a second ghost convolution layer, a BN layer and a splicing layer. Specifically, the reference operation and maintenance link map and the correlation coefficient map are input into the intrinsic convolution layer for convolution to obtain an intrinsic feature map; the intrinsic feature map is input into the first ghost convolution layer for processing to obtain a first convolution feature; the first convolution feature is normalized by the BN layer and the ReLU activation function to obtain a normalized convolution feature; the normalized convolution feature is convolved channel by channel by the depth-separable convolution layer to obtain a depth-convolution feature; the depth-convolution feature is input into the second ghost convolution layer to obtain an optimized depth-convolution feature; the optimized depth-convolution feature is normalized by the BN layer to obtain a depth-separable feature map; the intrinsic feature map and the depth-separable feature map are spliced ​​and fused by the splicing layer to output the first Ghost convolution feature.

[0122] That is, the Ghost convolution module performs ordinary convolution on the image to be segmented through the intrinsic convolution layer to obtain the intrinsic feature map, and then performs the intrinsic feature map one by one. Linear transformation and attention calculation are performed to obtain another part of the depth-separable feature map, which is then fused with the intrinsic feature map. In this way, the total number of parameters of the model is reduced to simplify the model.

[0123] Step S302, performing a pooling operation on the operation and maintenance graph features through a pooling layer to obtain pooled graph features.

[0124] The pooling operation compresses the input operation and maintenance graph features, reduces the features, leads to a reduction in parameters, and simplifies the complexity of the calculation; secondly, it increases the receptive field layer by layer, retains important information in the operation and maintenance graph features, and improves the overfitting problem. Commonly used pooling methods include mean-pooling, max-pooling, and stochastic-pooling.

[0125] Step S303, performing feature fusion on the pooled graph features through the graph fusion layer to obtain the operation and maintenance node fusion features.

[0126] In this embodiment, a fusion algorithm is used through the graph fusion layer to integrate the pooled graph features extracted from the reference operation and maintenance link graph and the correlation coefficient graph. This fusion process not only considers local features, but also integrates global information to form a high-level feature representation that describes the overall system architecture and node associations.

[0127] In some optional implementations, the step of fusing the pooled graph features through the graph fusion layer to obtain the fusion features of the operation and maintenance node includes:

[0128] The graph fusion layer transfers and aggregates the neighborhood node feature information of the pooled graph features to obtain node aggregation features;

[0129] The self-attention mechanism is used to calculate the node aggregation features to obtain the operation and maintenance node fusion features.

[0130] Specifically, the key path is extracted from the pooled graph features, a subgraph of the timing path graph is constructed, the adjacency matrix of the subgraph is established according to the connection relationship of the nodes in the subgraph, and the feature vector matrix of the subgraph is established according to the feature information of the nodes in the subgraph; the established adjacency matrix and feature vector matrix are used to transfer and aggregate the feature information of the neighborhood nodes through the graph fusion layer, and the feature vector matrix of the subgraph is updated according to the feature vectors of each node after transfer and aggregation to obtain the node aggregation feature. For the feature vector of each node in the node aggregation feature, according to the importance of other nodes on the key path relative to the current node, the weighted information is transferred and aggregated through the self-attention mechanism, and the feature vector of each node is updated to obtain the operation and maintenance node fusion feature.

[0131] The graph fusion layer uses graph neural network (GNN), which is a technology for deep learning on graph data structures. GNN can use the structural information of graph data to effectively process irregular and non-grid data structures.

[0132] The graph fusion layer is used to further learn meaningful node representations for the pooled graph features, and the self-attention mechanism is used to transfer global feature information to improve data accuracy and further improve the prediction accuracy of the model.

[0133] Step S304: Input the aggregated features of the operation and maintenance nodes and the pooled graph features into the multi-layer perception layer for weighted calculation to obtain a weighted feature matrix.

[0134] Specifically, the operation and maintenance node aggregation features and the operation and maintenance pooling features are input into the multi-layer perception layer for weighted representation to obtain a weighted feature matrix, wherein the multi-layer perception layer is an algorithm model based on a neural network, which consists of an input layer, a hidden layer, and an output layer. The multi-layer perception layer is used to construct a weighted representation.

[0135] Step S305, input the weighted feature matrix into the prediction layer for prediction to obtain the target link graph.

[0136] In this embodiment, the prediction layer includes a Bi-LSTM layer and a classification layer. The Bi-LSTM sublayer can automatically obtain data features, and use sequential and reverse recurrent neural networks to obtain two independent hidden layer representations for each input data, and then perform certain calculations (splicing or adding) on ​​the two hidden layer representations to obtain a final hidden layer representation, which is sent to the output layer for subsequent calculations. This hidden layer representation contains semantic information from both the previous and the following contexts for each data in the time series data. Among them, the Bi-LSTM layer includes a forward hidden layer and a backward hidden layer.

[0137] In some optional implementations of this embodiment, the step of inputting the weighted feature matrix into the prediction layer for prediction to obtain the target link graph includes:

[0138] The weighted feature matrix is ​​input into the forward hidden layer and the backward hidden layer respectively to transfer and update feature information, and the forward hidden features and the backward hidden features are obtained respectively;

[0139] Concatenate the forward hidden features and the backward hidden features according to the corresponding positions to obtain the hidden state;

[0140] The hidden state is classified and predicted through the classification layer to obtain the target link map.

[0141] Among them, the forward hidden feature is The backward hidden feature is Splicing to get h t for Then we can obtain the complete hidden state (h 1 ,h 2 ,…,h n )∈R n×m .

[0142] In this embodiment, the classification layer may be a CRF layer, or a fully connected layer + a softmax layer.

[0143] The prediction layer can make full use of past and future context information to predict the target link map of the second target time period and improve the accuracy of the prediction.

[0144] Operation and maintenance data generally do not change drastically often, so the map information on the t-1 day has a certain reference value. However, it cannot provide real-time feedback on some new situations. The operation and maintenance data one hour before the t day can roughly reflect the data changes on that day. Therefore, this application jointly uses the map information on the t-1 day and the operation and maintenance data one hour before the t day to accurately predict the intelligent operation and maintenance link map on the t day, so that the operation and maintenance team no longer needs to use the intelligent operation and maintenance link map on the t-1 day when performing the operation and maintenance tasks on the t day, thereby solving the lag problem in establishing the intelligent operation and maintenance link map.

[0145] This application improves the accuracy and reliability of data through denoising operations. Combining the correlation coefficient map and the reference link map can update the link map in advance and perform operation and maintenance tasks more effectively. At the same time, the link map prediction model is used to improve the map prediction accuracy, which can fully and accurately predict future intelligent operation and maintenance link maps. It is very helpful to quickly locate problems, improve system availability and reliability, reduce fault repair time, and optimize resource utilization efficiency, providing a scientific basis and direction for subsequent operation and maintenance strategy formulation and performance optimization.

[0146] In some optional implementations, before the step of inputting the reference link map and the correlation coefficient map into the trained link map prediction model for prediction, the following steps are further included:

[0147] Acquire a sample data set, the sample data set including a historical link map of a first historical time period and a historical correlation coefficient map of a second historical time period as a historical operation and maintenance data pair, wherein the historical operation and maintenance data pair carries a corresponding real link map of the second historical time period;

[0148] The pre-built initial link graph prediction model is trained using the sample data set to obtain a predicted link graph;

[0149] The loss value between the predicted link graph and the actual link graph is calculated according to the loss function, the model parameters are adjusted based on the loss value, and training is continued until the training is completed to obtain the final link graph prediction model.

[0150] Among them, the sample data set may include a historical link map of a first historical time period and a real link map of a second historical time period, extract historical operation and maintenance data of a specific time period of the second historical time period from the real link map, denoise the historical operation and maintenance data, and calculate the Pearson correlation map of the historical operation and maintenance data. The historical link map and the Pearson correlation map are convolutionally fused through a pre-built initial link map prediction model to obtain a predicted link map. The real link map is used as a label and the root mean square error loss is used as a loss function. The loss values ​​of the real link map and the predicted link map are calculated, and the model parameters are continuously optimized through the back propagation algorithm to obtain the optimal parameter set.

[0151] Among them, the link graph prediction model is mainly predicted by the convolution fusion method, which is an effective pattern matching operation. Since the connection relationship between two nodes depends on the connection relationship of all nodes connected to them, the convolution method is used here to predict the historical link graph A. t-1 And the Pearson correlation graph P t Fusion is performed to predict the intelligent operation and maintenance link map A for the day t .

[0152] A t-1 and P t The fusion is performed by convolution, and the specific demonstration is as follows:

[0153]

[0154] Among them, A ab is the node relationship between two nodes a and b in the fused graph; P ab Represents the Pearson correlation graph P t The node relationship between two nodes a and b in A t-1-ab Represents the historical link graph A t-1 The node relationship between the two nodes a and b; m and n are the number of rows and columns of the graph structure representation matrix respectively; w p-ab and w t-1-ij are the weights in the convolution operations in the two graphs, respectively, and are learnable parameters.

[0155] The trained prediction model can then make accurate predictions on the day’s intelligent operation and maintenance link map based on the previous day’s intelligent operation and maintenance link map and the filtered operation and maintenance data of the first hour of the day, thereby improving the efficiency and accuracy of the prediction.

[0156] The application of intelligent operation and maintenance link maps in the financial field is gradually deepening. Through technological innovation and scenario-based applications, it has brought significant benefits and value to the financial industry. In the financial field, the intelligent operation and maintenance link map can realize technologies such as second-level resource expansion, dynamic flow control and fault self-healing, allocate resources in advance, and enhance application performance and emergency response capabilities. For example, the intelligent operation and maintenance link map of the target time period can be predicted in advance through the link map prediction method of this application, which can be applied during major promotions such as "Double Eleven". The shopping platform can ensure the stable operation of the system through the intelligent operation and maintenance link map. The financial field has extremely high requirements for risk management. The intelligent operation and maintenance link map can timely discover and rectify potential risks and loopholes through real-time monitoring and analysis. For example, the bank predicts the target intelligent operation and maintenance link map through the link map prediction method of this application, and realizes minute-level fault demarcation and positioning through the target intelligent operation and maintenance link map, which improves business stability and continuity.

[0157] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0158] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0159] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0160] Further references Figure 4 , as a response to the above Figure 2 The present application provides an embodiment of a link graph prediction device, and the device embodiment is similar to Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0161] like Figure 4 As shown, the link graph prediction device 400 described in this embodiment includes: an acquisition module 401, a calculation module 402 and a prediction module 403. Among them:

[0162] The acquisition module 401 is used to acquire a reference link graph of a first target time period and a target operation and maintenance data set of a specific time period in a second target time period, and denoise the target operation and maintenance data set to obtain an enhanced operation and maintenance data set;

[0163] The calculation module 402 is used to calculate the correlation coefficient of the enhanced operation and maintenance data set to obtain a correlation coefficient map;

[0164] The prediction module 403 is used to input the reference link map and the correlation coefficient map into a trained link map prediction model for prediction, so as to obtain a target link map for the second target time period.

[0165] It should be emphasized that in order to further ensure the privacy and security of the target link map, the above-mentioned target link map can also be stored in a node of a blockchain.

[0166] Based on the above-mentioned link map prediction device 400, the accuracy and reliability of the data are improved through denoising. Combining the correlation coefficient map and the reference link map can update the link map in advance and perform operation and maintenance tasks more effectively. At the same time, the link map prediction model is used to improve the map prediction accuracy, and the future intelligent operation and maintenance link map can be fully and accurately predicted, which is very helpful to quickly locate problems, improve system availability and reliability, reduce fault repair time, and optimize resource utilization efficiency, providing a scientific basis and direction for subsequent operation and maintenance strategy formulation and performance optimization.

[0167] In some optional implementations, the acquisition module 401 includes a preprocessing submodule, a conversion submodule, a determination submodule, and a filtering submodule, wherein:

[0168] The preprocessing submodule is used to preprocess all target operation and maintenance data in the target operation and maintenance data set to obtain a preprocessed operation and maintenance data set;

[0169] The conversion submodule is used to convert the pre-processed operation and maintenance data set into time series operation and maintenance data according to a preset format;

[0170] The determination submodule is used to analyze the noise characteristics of the time series operation and maintenance data, and determine Gaussian filter parameters based on the noise characteristics;

[0171] The filtering submodule is used to denoise the time series operation and maintenance data using a Gaussian filtering algorithm according to the Gaussian filtering parameters to obtain an enhanced operation and maintenance data set.

[0172] Through Gaussian filtering and denoising, random fluctuations and outliers in the data can be effectively smoothed and removed, resulting in clearer and more stable data representation and improved data quality and reliability.

[0173] In some optional implementations, the calculation module 402 includes a calculation submodule and a filtering submodule, wherein:

[0174] The calculation submodule is used to calculate the correlation coefficient of each pair of nodes in the enhanced operation and maintenance data set, and obtain a correlation heat map according to the correlation coefficient;

[0175] The filtering submodule is used to filter the weakly connected nodes in the enhanced operation and maintenance data set based on the correlation heat map to obtain a correlation coefficient map.

[0176] The correlation coefficient map can intuitively display the size and direction of the correlation between the attributes of each node, so as to analyze the key related attributes, which will play an important role in subsequent analysis.

[0177] In some optional implementations, the link graph prediction model includes a convolution layer, a pooling layer, a graph fusion layer, a multi-layer perception layer and a prediction layer; the prediction module 403 includes a convolution submodule, a pooling submodule, a fusion submodule, a weighted submodule and a prediction submodule, wherein:

[0178] The convolution submodule is used to input the reference link map and the correlation coefficient map into the convolution layer for feature extraction to obtain operation and maintenance map features;

[0179] The pooling submodule is used to perform a pooling operation on the operation and maintenance graph features through the pooling layer to obtain pooled graph features;

[0180] The fusion submodule is used to perform feature fusion on the pooled graph features through the graph fusion layer to obtain the operation and maintenance node fusion features;

[0181] The weighted submodule is used to input the operation and maintenance node aggregation features and the pooled graph features into the multi-layer perception layer for weighted calculation to obtain a weighted feature matrix;

[0182] The prediction submodule is used to input the weighted feature matrix into the prediction layer for prediction to obtain a target link map.

[0183] Combining the correlation coefficient map and the reference link map can update the link map in advance and perform operation and maintenance tasks more effectively. At the same time, the link map prediction model can be used to improve the map prediction accuracy, and the future intelligent operation and maintenance link map can be fully and accurately predicted, which is very helpful to quickly locate problems, improve system availability and reliability, and reduce fault repair time.

[0184] In some optional implementations of this embodiment, the fusion submodule is further used to:

[0185] The graph fusion layer transmits and aggregates the neighborhood node feature information of the pooled graph features to obtain node aggregation features;

[0186] The node aggregation features are calculated using a self-attention mechanism to obtain operation and maintenance node fusion features.

[0187] The graph fusion layer is used to further learn meaningful node representations for the pooled graph features, and the self-attention mechanism is used to transfer global feature information to improve data accuracy and further improve the prediction accuracy of the model.

[0188] In some optional implementations of this embodiment, the prediction layer includes a forward hidden layer, a backward hidden layer and a classification layer; the prediction submodule is further used for:

[0189] Inputting the weighted feature matrix into the forward hidden layer and the backward hidden layer respectively to transfer and update feature information, and obtaining forward hidden features and backward hidden features respectively;

[0190] The forward hidden feature and the backward hidden feature are concatenated according to corresponding positions to obtain a hidden state;

[0191] The hidden state is classified and predicted by the classification layer to obtain a target link map.

[0192] The prediction layer can make full use of past and future context information to predict the target link map of the second target time period and improve the accuracy of the prediction.

[0193] In some optional implementations, the link graph prediction device 400 further includes a training module, which is used to:

[0194] Acquire a sample data set, the sample data set including a historical link map of a first historical time period and a historical correlation coefficient map of a second historical time period as a historical operation and maintenance data pair, wherein the historical operation and maintenance data pair carries a real link map of the corresponding second historical time period;

[0195] Using the sample data set to train a pre-built initial link graph prediction model to obtain a predicted link graph;

[0196] The loss value between the predicted link graph and the actual link graph is calculated according to the loss function, the model parameters are adjusted based on the loss value, and the training is continued until the training is completed to obtain the final link graph prediction model.

[0197] The training prediction model can make an accurate prediction of the intelligent operation and maintenance link map of the day based on the intelligent operation and maintenance link map of the previous day and the filtered operation and maintenance data of the first hour of the day, thereby improving the efficiency and accuracy of the prediction.

[0198] To solve the above technical problems, the present application also provides a computer device. Figure 5 , Figure 5 This is a basic structural block diagram of the computer device in this embodiment.

[0199] The computer device 5 includes a memory 51, a processor 52, and a network interface 53 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 5 with components 51-53, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0200] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0201] The memory 51 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 51 can be an internal storage unit of the computer device 5, such as a hard disk or memory of the computer device 5. In other embodiments, the memory 51 can also be an external storage device of the computer device 5, such as a plug-in hard disk equipped on the computer device 5, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. Of course, the memory 51 can also include both the internal storage unit of the computer device 5 and its external storage device. In this embodiment, the memory 51 is generally used to store the operating system and various application software installed on the computer device 5, such as computer-readable instructions of the link map prediction method, etc. In addition, the memory 51 can also be used to temporarily store various types of data that have been output or are to be output.

[0202] The processor 52 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip in some embodiments. The processor 52 is generally used to control the overall operation of the computer device 5. In this embodiment, the processor 52 is used to run computer-readable instructions stored in the memory 51 or process data, such as computer-readable instructions for running the link map prediction method.

[0203] The network interface 53 may include a wireless network interface or a wired network interface. The network interface 53 is generally used to establish a communication connection between the computer device 5 and other electronic devices.

[0204] This embodiment implements the steps of the link map prediction method of the above embodiment when the processor executes computer-readable instructions stored in the memory. The accuracy and reliability of the data are improved by denoising. The link map can be updated in advance by combining the correlation coefficient map and the reference link map, and the operation and maintenance tasks can be performed more effectively. At the same time, the map prediction accuracy is improved by the link map prediction model, and the future intelligent operation and maintenance link map can be fully and accurately predicted, which is very helpful to quickly locate problems, improve system availability and reliability, reduce fault repair time, and optimize resource utilization efficiency, providing a scientific basis and direction for subsequent operation and maintenance strategy formulation and performance optimization.

[0205] The present application also provides another implementation mode, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor so that the at least one processor executes the steps of the link graph prediction method as described above, and improves the accuracy and reliability of the data through denoising, and can update the link graph in advance by combining the correlation coefficient graph and the reference link graph, and perform operation and maintenance tasks more effectively. At the same time, the graph prediction accuracy is improved through the link graph prediction model, and the future intelligent operation and maintenance link graph can be fully and accurately predicted, which is very helpful to quickly locate problems, improve system availability and reliability, reduce fault repair time, and optimize resource utilization efficiency, providing a scientific basis and direction for subsequent operation and maintenance strategy formulation and performance optimization.

[0206] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0207] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A link graph prediction method, It is characterized in that The steps include: Acquire a reference link graph of a first target time period and a target operation and maintenance data set for a specific time period in a second target time period, and denoise the target operation and maintenance data set to obtain an enhanced operation and maintenance data set; Calculate the correlation coefficient of the enhanced operation and maintenance data set to obtain a correlation coefficient map; The reference link map and the correlation coefficient map are input into a trained link map prediction model for prediction to obtain a target link map for the second target time period.

2. The link graph prediction method according to claim 1, It is characterized in that The step of denoising the target operation and maintenance data set to obtain an enhanced operation and maintenance data set comprises: Preprocessing all target operation and maintenance data in the target operation and maintenance data set to obtain a preprocessed operation and maintenance data set; Converting the preprocessed operation and maintenance data set into time series operation and maintenance data according to a preset format; Analyzing the noise characteristics of the time series operation and maintenance data, and determining Gaussian filter parameters based on the noise characteristics; According to the Gaussian filter parameters, a Gaussian filter algorithm is used to denoise the time series operation and maintenance data to obtain an enhanced operation and maintenance data set.

3. The link graph prediction method according to claim 1, It is characterized in that The step of calculating the correlation coefficient of the enhanced operation and maintenance data set to obtain a correlation coefficient map comprises: Calculate the correlation coefficient of each pair of nodes in the enhanced operation and maintenance data set, and obtain a correlation heat map according to the correlation coefficient; The weakly connected nodes in the enhanced operation and maintenance data set are filtered based on the correlation heat map to obtain a correlation coefficient map.

4. The link graph prediction method according to claim 1, It is characterized in that The link graph prediction model includes a convolution layer, a pooling layer, a graph fusion layer, a multi-layer perception layer and a prediction layer; the step of inputting the reference link graph and the correlation coefficient graph into the trained link graph prediction model for prediction to obtain the target link graph for the second target time period includes: Inputting the reference link map and the correlation coefficient map into the convolution layer for feature extraction to obtain operation and maintenance map features; Performing a pooling operation on the operation and maintenance graph features through the pooling layer to obtain pooled graph features; The pooled graph features are fused through the graph fusion layer to obtain operation and maintenance node fusion features; Inputting the operation and maintenance node aggregation features and the pooled graph features into the multi-layer perception layer for weighted calculation to obtain a weighted feature matrix; The weighted feature matrix is ​​input into the prediction layer for prediction to obtain a target link graph.

5. The link graph prediction method according to claim 4, It is characterized in that The step of fusing the pooled graph features through the graph fusion layer to obtain the operation and maintenance node fusion features includes: The graph fusion layer transmits and aggregates the neighborhood node feature information of the pooled graph features to obtain node aggregation features; The node aggregation features are calculated using a self-attention mechanism to obtain operation and maintenance node fusion features.

6. The link graph prediction method according to claim 4, It is characterized in that The prediction layer includes a forward hidden layer, a backward hidden layer and a classification layer; the step of inputting the weighted feature matrix into the prediction layer for prediction to obtain a target link map includes: Inputting the weighted feature matrix into the forward hidden layer and the backward hidden layer respectively to transfer and update feature information, and obtaining forward hidden features and backward hidden features respectively; The forward hidden feature and the backward hidden feature are concatenated according to corresponding positions to obtain a hidden state; The hidden state is classified and predicted by the classification layer to obtain a target link map.

7. The link graph prediction method according to claim 1, It is characterized in that Before the step of inputting the reference link graph and the correlation coefficient graph into the trained link graph prediction model for prediction, the step further includes: Acquire a sample data set, the sample data set including a historical link map of a first historical time period and a historical correlation coefficient map of a second historical time period as a historical operation and maintenance data pair, wherein the historical operation and maintenance data pair carries a real link map of the corresponding second historical time period; Using the sample data set to train a pre-built initial link graph prediction model to obtain a predicted link graph; The loss value between the predicted link graph and the actual link graph is calculated according to the loss function, the model parameters are adjusted based on the loss value, and the training is continued until the training is completed to obtain the final link graph prediction model.

8. A link graph prediction device, It is characterized in that include: An acquisition module, used to acquire a reference link graph of a first target time period and a target operation and maintenance data set for a specific time period in a second target time period, and denoise the target operation and maintenance data set to obtain an enhanced operation and maintenance data set; A calculation module, used to calculate the correlation coefficient of the enhanced operation and maintenance data set to obtain a correlation coefficient map; The prediction module is used to input the reference link map and the correlation coefficient map into a trained link map prediction model for prediction, so as to obtain a target link map for the second target time period.

9. A computer device, It is characterized in that It includes a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the link graph prediction method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the steps of the link graph prediction method according to any one of claims 1 to 7 are implemented.

Images