Enterprise digital operation real-time data analysis method based on multi-source data fusion

Abstract

The invention discloses an enterprise digital operation real-time data analysis method based on multi-source data fusion, relates to the technical field of data processing, and solves the problems that firstly, an enterprise comprehensive benefit index is difficult to analyze and predict; secondly, it is difficult to embed the knowledge graph into a vector space to construct a dynamic graph neural network; then, the equipment abnormal offset, the supply chain abnormal offset and the customer activeness are difficult to analyze, a risk matrix is constructed, and a comprehensive benefit index decline propagation path is identified; service link resources are difficult to dynamically allocate according to an analysis result of the risk matrix; and finally, a data analysis result is difficult to display visually. According to the method, enterprise multi-source data is collected and preprocessed, a three-dimensional vector space and a dynamic graph neural network are constructed, enterprise benefits and risks are analyzed and predicted in combination with multiple technologies, resources of business links are dynamically allocated, and a data analysis result is visually displayed.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and specifically relates to a real-time data analysis method for enterprise digital operation based on multi-source data fusion. Background Art

[0002] With the rapid development of Internet, Internet of Things and big data technologies, enterprises generate a vast amount of multi-source data during operation. However, traditional data processing and analysis technologies are difficult to effectively integrate and utilize this multi-source data, and cannot meet the needs of enterprise real-time data analysis. By focusing on the data processing, analysis and visualization requirements in enterprise digital operation, and supported by key technologies such as Apache Flink stream processing framework, multi-source data fusion, knowledge graph, deep learning and visualization technologies, comprehensive, real-time and accurate digital operation data analysis results are provided for enterprises.

[0003] The following problems exist in the prior art: First, it is difficult to analyze and predict the comprehensive benefit index of enterprises; second, it is difficult to embed the knowledge graph into the vector space and construct a dynamic graph neural network; then, it is difficult to analyze the abnormal offsets of equipment, the abnormal offsets of the supply chain and customer activity, construct a risk matrix and identify the propagation path of the decline in the comprehensive benefit index; it is difficult to dynamically allocate business link resources based on the analysis results of the risk matrix; finally, it is difficult to visually display the data analysis results. Summary of the Invention

[0004] To solve the problems existing in the above-mentioned prior art, the first aspect of the present invention provides a real-time data analysis method for enterprise digital operation based on multi-source data fusion, including the following modules:

[0005] To solve the above problems, the first aspect of the present invention provides a real-time data analysis method for enterprise digital operation based on multi-source data fusion, including the following steps:

[0006] S1: Real-time collect the operation multi-source data during enterprise operation, construct an operation multi-source data set, and perform preprocessing;

[0007] S2: Extract features according to the preprocessed operation multi-source data set and construct a three-dimensional vector space;

[0008] S3: Use the three-dimensional vector space and Apache Flink, combined with a time series prediction model, to analyze and predict the comprehensive benefit index of the enterprise; analyze and evaluate the future operation trend and potential risks of the enterprise according to the prediction results; embed the knowledge graph into the three-dimensional vector space and construct a dynamic graph neural network;

[0009] S4: Analyze the abnormal offset of the equipment, the abnormal offset of the supply chain, and the customer activity, construct a risk matrix, and identify the propagation path of the decline in the comprehensive benefit index; use the analysis results of the risk conduction matrix to dynamically allocate resources for business processes;

[0010] S5: Use a dashboard to visually display the trends of the enterprise's future operations, the comprehensive benefit index, and the propagation path of the decline in the comprehensive benefit index.

[0011] Preferably, in step S1, the operation multi-source data set includes: equipment operation data, enterprise financial data, customer behavior data, human resources data, and external data.

[0012] Preferably, step S2 includes the following steps:

[0013] Access the preprocessed operation multi-source data set through Apache Flink, and extract the maximum value, minimum value, and average value of the equipment operation data, enterprise financial data, customer behavior data, and human resources data as statistical feature values;

[0014] Extract the difference between the inflow and outflow of funds and the capital turnover rate from the enterprise financial data; extract the customer retention rate and customer activity from the customer behavior data; extract the employee turnover rate from the human resources data;

[0015] The feature values extracted from the external data include: the busyness degree and traffic congestion degree of the enterprise's geographical area;

[0016] According to the results of the above-mentioned feature extraction of the operation multi-source data set, combine them into an equipment status feature vector, an enterprise financial feature vector, a customer behavior feature vector, a human resources feature vector, and an external feature vector;

[0017] Combine the above five constructed feature vectors to form a comprehensive feature vector matrix; adopt the principal component analysis and dimensionality reduction method to retain the principal components with 95% variance to normalize the fused feature vectors, and update the feature vectors in real time in the Apache Flink stream processing framework;

[0018] Project the comprehensive feature vector matrix into a three-dimensional space, adopt a right-handed coordinate system, construct a three-dimensional space vector, where the x-axis is the equipment status feature vector, the y-axis is the enterprise financial feature vector and the customer behavior feature vector, and the z-axis is the human resources feature vector and the external feature vector; select three linearly independent vectors as the basis, marked as i, j, k, and determine the order of the three directions according to the right-hand rule.

[0019] Preferably, in step S3, using the three-dimensional vector space and Apache Flink, combined with the time series prediction model, to analyze and predict the enterprise's comprehensive benefit index, includes the following steps:

[0020] Visualize the vector changes in the three-dimensional vector space through the stream processing ability of Apache Flink; evaluate the real-time cost-benefit status of enterprise operations in real time according to the position and change trend of vectors in the three-dimensional vector space;

[0021] Calculation formula for the comprehensive benefit index:

[0022] Obtain the comprehensive benefit index Q of the enterprise; where, is the modulus length of the current state vector, and the current state vector represents the position in the three-dimensional vector space; E(f) represents the information entropy of the vector distribution of the current state vector; f represents the current state feature vector; S t represents the revenue of the enterprise at time t; C t is the cost of the enterprise at time t; α, β, and γ represent the weight coefficients, which are 0.4, 0.3, and 0.3 respectively;

[0023] Use the long short-term memory network model to predict the future comprehensive benefit index, and input the historical operation multi-source data set into the long short-term memory network LSTM model for training; during the training process, input it into the LSTM model in chronological order, learn the data characteristics and change rules at different time points, and establish a mapping relationship from the past to the present and then to the future;

[0024] Deploy the trained LSTM model to the Apache Flink environment to predict the newly incoming operation multi-source data in real time; when the real-time operation multi-source data enters the Flink stream processing pipeline, according to the current input data and the comprehensive benefit index formula, the prediction result is the future comprehensive benefit index.

[0025] Preferably, in step S3, analyze and evaluate the future operation trend and potential risks of the enterprise according to the prediction result, including the following steps:

[0026] Plot the comprehensive benefit indexes at different predicted time points as a time series graph, where the abscissa represents time and the ordinate represents the comprehensive benefit index;

[0027] Adopt the linear regression method, use time as the independent variable and the comprehensive benefit index as the dependent variable to fit a straight line; judge the rising or falling trend of the comprehensive benefit index with time through the slope of the fitted straight line; when Q shows an upward trend, it indicates that the future operation trend of the enterprise is good; when Q shows a downward trend, it indicates that the future operation trend of the enterprise is poor;

[0028] According to the historical comprehensive benefit index, different thresholds are divided: when Q > 0.8, it indicates that the enterprise operation status is excellent; when 0.4 ≤ Q ≤ 0.8, it indicates that the enterprise operation status is good; when Q < 0.4, it indicates that the enterprise operation status is dangerous;

[0029] Use the LSTM model to predict the future comprehensive benefit index, and judge whether there is a potential risk in the future enterprise operation status according to the set threshold; when the future comprehensive benefit index is less than 0.4, it is identified that there is a potential risk, otherwise, there is no potential risk.

[0030] Preferably, in step S3, embedding the knowledge graph into a three-dimensional vector space and constructing a dynamic graph neural network includes the following steps:

[0031] Embed the knowledge graph into a three-dimensional vector space. In the knowledge graph, the nodes represent entities including: equipment, customers, and employees, and the edges represent the relationships between entities including: the relationship between equipment and customers, the relationship between equipment and employees, and the relationship between customers and employees;

[0032] Use the TransE model to embed entities into a three-dimensional vector space; introduce the time attribute, regard time as a dimension in the graph, convert the time stamp into a sine or cosine periodic vector, and splice it with the three-dimensional space vector, and then reduce the dimension to three through a fully connected layer to form a spatio-temporal joint embedding, and obtain the node representation containing spatio-temporal information;

[0033] Adopt the EvolveGCN model to capture the dynamic changes of the graph structure by evolving the graph convolutional network parameters of the LSTM;

[0034] Combine with the Apache Flink stream processing framework to synchronize the incremental data of the graph structure every 5 minutes; utilize the stream processing ability of Flink to update the new data into the knowledge graph, and update the parameters of the dynamic graph neural network in real time through the EvolveGCN model.

[0035] Preferably, in step S4, analyze the equipment abnormal offset, the supply chain abnormal offset and the customer activity, construct a risk matrix and identify the propagation path of the decline of the comprehensive benefit index, including the following steps:

[0036] Calculate the average value of each characteristic parameter in the normal equipment operation characteristic vector as the reference value to obtain the reference vector; use the Euclidean distance formula to calculate the offset between the current equipment state vector and the reference vector in real time to obtain the equipment abnormal offset;

[0037] Combine the external characteristic vector, the enterprise financial characteristic vector and the human resource characteristic vector to generate the supply chain state characteristic vector, and obtain the supply chain abnormal offset in the same calculation method as the equipment abnormal offset;

[0038] Obtain the activity data of customer nodes, calculate the difference between the activity in the previous period and the activity in the current period, and then divide it by the activity in the previous period to obtain the activity decline rate;

[0039] Take the device abnormal offset, the supply chain abnormal offset, the customer activity, and the timestamp as input signals; introduce a graph attention network model to construct a dynamic graph neural network, and use the graph attention mechanism to automatically learn the relationships between nodes;

[0040] Input the input signals into the constructed dynamic graph neural network for training, calculate the influence degree of the critical path on the potential risk probability through the graph attention mechanism, and output the potential risk probability;

[0041] According to the output result of the dynamic graph neural network, evaluate the risk influence degree of different nodes and paths, construct a risk conduction matrix, and the elements in the matrix represent the intensity of risk conduction between different nodes or among the three business links; the three business links include: the device operation business link, the customer activity business link, and the supply chain business link;

[0042] When the comprehensive benefit index shows a downward trend, use the risk conduction matrix to identify the risk propagation path and the risk influence degree.

[0043] Preferably, in step S4, using the analysis result of the risk conduction matrix to dynamically allocate resources for business links includes the following steps:

[0044] Adopt the eigenvector weighting method to obtain the contribution weights of the three business links of device operation, customer activity, and supply chain to the comprehensive benefit index respectively;

[0045] According to the intensity of risk conduction among the three business links in the risk conduction matrix, calculate the sensitivity of the three business links during the risk conduction process respectively;

[0046] Through the graph attention mechanism, automatically learn the relationships between nodes, calculate the influence degree of the critical path on the potential risk probability, and determine the influence ranking of the three business links;

[0047] Combine the contribution weights of the three business links to the comprehensive benefit index respectively, the risk conduction sensitivity, and the influence ranking obtained by the graph attention network to construct a resource allocation weight function and determine the weight coefficient of resource allocation;

[0048] After normalizing the obtained resource allocation weight coefficients of the three business links respectively, sort the three business links respectively to determine the resource allocation response priority of the three business links.

[0049] Compared with the prior art, the beneficial effects of the present invention are:

[0050] The present invention collects multi-source data sets during the operation of an enterprise in real time, and performs preprocessing using the Apache Flink stream processing framework and a lightweight geocoding architecture; by utilizing a three-dimensional vector space and real-time data analysis technology, combined with a time series prediction model, it can accurately analyze and predict the comprehensive benefit index of the enterprise, helping the enterprise to understand future operation trends in advance.

[0051] The present invention constructs a risk matrix and identifies the propagation path of the decline in the comprehensive benefit index by analyzing equipment abnormal offsets, supply chain abnormal offsets, and customer activity, enabling the enterprise to promptly discover potential risks. By dynamically allocating resources in business processes, it improves resource utilization efficiency, ensuring that it can respond quickly and effectively when risks occur, and safeguarding the stability and comprehensive benefits of enterprise operations.

[0052] The present invention visually displays important information such as the future operation trends, comprehensive benefit index, and propagation path of the decline in the comprehensive benefit index of the enterprise through a dashboard, enabling enterprise managers to intuitively and clearly understand the enterprise operation status. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0054] Figure 1 It is the system method flowchart of the present invention;

[0055] Figure 2 It is the schematic diagram of the prediction model process in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0056] The following will clearly and completely describe the technical solutions of the present invention in combination with the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0057] Please refer to Figure 1-2 , the first aspect embodiment of the present invention provides a real-time data analysis method for enterprise digital operation based on multi-source data fusion, including the following steps:

[0058] S1: Collect the operation multi-source data during the operation of the enterprise in real time, construct an operation multi-source data set, and perform preprocessing;

[0059] S2: Extract features from the preprocessed multi-source operation dataset and construct a three-dimensional vector space;

[0060] S3: Use the three-dimensional vector space and Apache Flink, combined with a time series prediction model, to analyze and predict the enterprise's comprehensive benefit index; Analyze and evaluate the future operation trends and potential risks of the enterprise based on the prediction results; Embed the knowledge graph into the three-dimensional vector space to construct a dynamic graph neural network;

[0061] S4: Analyze the abnormal offsets of equipment, the abnormal offsets of the supply chain, and customer activity, construct a risk matrix and identify the propagation path of the decline in the comprehensive benefit index; Dynamically allocate business link resources using the analysis results of the risk conduction matrix;

[0062] S5: Use a dashboard to visually display the future operation trends, comprehensive benefit index, and propagation path of the decline in the comprehensive benefit index of the enterprise.

[0063] Specifically, through various sensors, information system interfaces, etc., a multi-source operation dataset during enterprise operation is collected in real time, including: equipment operation data, enterprise financial data, customer behavior data, human resources data, and external data, etc. The connectors of Apache Flink can be used to integrate with each data source to achieve real-time data access. The Apache Flink stream processing framework is adopted to clean the collected data, removing duplicate, incorrect or missing data. Using its functions such as window operations, the data is transformed, such as unit conversion and unification of equipment operation data. At the same time, a lightweight geocoding architecture is used to encode the data with geographical location information, unifying the data format and timestamp for subsequent analysis. Feature extraction is performed on the preprocessed multi-source operation dataset, and features that can reflect the key information of enterprise operation are selected, and then a three-dimensional vector space is constructed, and the extracted features are mapped into this space, and each dimension represents different features or feature combinations respectively, so as to achieve multi-dimensional representation of enterprise operation data. Using the constructed three-dimensional vector space and real-time data analysis technology, combined with the time series prediction model, the comprehensive benefit index of the enterprise is analyzed and predicted to understand the overall operation benefit trend of the enterprise in the future period in advance. According to the prediction results, the future operation trend and potential risks of the enterprise are analyzed and evaluated. The knowledge graph is embedded into the vector space, and a dynamic graph neural network is constructed. Through this network and the graph attention mechanism, the abnormal offset of the equipment and the customer activity are analyzed, a risk matrix is constructed, the propagation path of the decline of the comprehensive benefit index is identified, and the risk propagation path is determined. Using the analysis results of the risk conduction matrix, the resource allocation priorities of the three business links of equipment, customers and the supply chain are analyzed. Real-time monitoring and resource optimization allocation of the three key business links of equipment operation, customer activity and supply chain in enterprise digital operation are realized, and the operation efficiency of the enterprise is improved. Using visualization tools such as dashboards, the analysis results such as the future operation trend of the enterprise, the comprehensive benefit index, and the propagation path of the decline of the comprehensive benefit index are displayed in intuitive charts, graphs, etc., which is convenient for enterprise managers, etc. to quickly understand the enterprise operation status and make decisions in a timely manner.

[0064] In one embodiment, step S1, wherein the multi-source operation dataset includes: equipment operation data, enterprise financial data, customer behavior data, human resources data, and external data.

[0065] Specifically, real-time operational multi-source data within the current cycle time is collected. The cycle is used as an operational cycle for multi-source data collection according to specific time spans such as hours, days, weeks, months, quarters, or years, and is specifically determined according to the enterprise's own business operation model. Equipment operation data collection includes installing temperature sensors, vibration sensors, production counting sensors, timers, and energy monitors at key parts of the enterprise's production equipment. The collected equipment operation data includes equipment temperature, equipment vibration, equipment production quantity, production time, and equipment energy consumption. Enterprise financial data collection includes interfacing with the enterprise's existing financial software and obtaining financial data such as total sales, sales quantity, cash inflows and outflows in real time through APIs or middleware technology. A data synchronization mechanism is established to ensure the timeliness and accuracy of financial data. During data synchronization, the data is verified and compared to avoid data duplication or errors. By primary key matching and timestamp comparison, the consistency of sales data between different systems is checked, and data synchronization problems are promptly discovered and resolved. At the same time, historical financial data is regularly backed up for data recovery and analysis when needed. Customer behavior data collection includes embedding the code of an analysis tool in the enterprise's official website and mobile application to monitor customers' access behaviors, and real-time tracking of information such as the source of customer access, accessed pages, stay time, and click counts, so as to calculate key indicators such as the number of customer accesses, access duration, and interaction times. When a customer opens a product page on the enterprise website, information such as the customer's access time and the content of the browsed page is recorded to obtain the customer's stay time and interaction behavior on this page. Human resources data collection includes integrating with the enterprise's human resources management system and obtaining data such as the number of employees and employees' working hours through data interfaces. External data collection includes collecting geographical location information of the enterprise's location through geographic information system technology, including the enterprise's longitude and latitude coordinates, the busyness degree of the enterprise's geographical area, and the traffic congestion degree.

[0066] In one of the embodiments, step S2 includes the following steps:

[0067] Access the preprocessed operational multi-source data set through Apache Flink, and extract the maximum, minimum, and average values of the equipment operation data, enterprise financial data, customer behavior data, and human resources data as statistical feature values;

[0068] Extract the difference between cash inflows and outflows and the cash turnover rate from the enterprise financial data; extract the customer retention rate and customer activity from the customer behavior data; extract the employee turnover rate from the human resources data;

[0069] The extracted feature values of the external data include: the busyness degree and traffic congestion degree of the enterprise's geographical area;

[0070] According to the results of the above-mentioned feature extraction of the operation multi-source data set, combine them into device status feature vectors, enterprise financial feature vectors, customer behavior feature vectors, human resource feature vectors, and external feature vectors;

[0071] Combine the five feature vectors constructed above to form a comprehensive feature vector matrix; adopt the principal component analysis dimensionality reduction method to retain the principal components with 95% variance to normalize the fused feature vectors, and update the feature vectors in real time in the Apache Flink stream processing framework;

[0072] Project the comprehensive feature vector matrix into a three-dimensional space, adopt a right-handed coordinate system, and construct three-dimensional space vectors, where the x-axis is the device status feature vector, the y-axis is the enterprise financial feature vector and the customer behavior feature vector, and the z-axis is the human resource feature vector and the external feature vector; select three linearly independent vectors as the basis, marked as i, j, k, and determine the order of the three directions according to the right-hand rule.

[0073] Specifically, through the data source interfaces of Apache Flink, such as KafkaConsumer, SocketStream, etc., access the preprocessed real-time data stream, including: device operation data, enterprise financial data, customer behavior data, human resources data, and external data, etc. Use the built-in functions or custom functions of Flink to calculate the maximum, minimum, and average values of data such as device temperature, vibration, production quantity, production time, and energy consumption in real time, and combine these statistical feature values into a device status feature vector. Similarly, use the functions of Flink to calculate the maximum, minimum, and average values of total sales, sales quantity, capital inflow and outflow. At the same time, calculate the capital turnover rate by dividing the total sales by the difference between the capital inflow and outflow divided by 2, and combine the difference between the capital inflow and outflow, the capital turnover rate, and the statistical feature values into an enterprise financial feature vector. Extract the maximum, minimum, and average values of the number of customer visits, visit duration, and interaction times. In addition, calculate the customer retention rate by dividing the difference between the total number of customers at the end of the current period and the total number of new customers in the current period by the total number of customers at the end of the previous period; use the RFE model to calculate the customer activity based on the sum of the visit time R, visit frequency F, and interaction times E of customers at the end of the current period; combine the statistical feature values, customer retention rate, and customer activity into a customer behavior feature vector. Calculate the maximum, minimum, and average values of the number of employees and the working hours of employees. At the same time, calculate the employee turnover rate by dividing the number of employees leaving by the total number of employees, and combine it with the statistical feature values into a human resources feature vector. Extract the feature values such as longitude and latitude coordinates, area code, busyness, and traffic congestion degree in the geographical location information, and combine them into an external feature vector. Combine the constructed device status feature vector, enterprise financial feature vector, customer behavior feature vector, human resources feature vector, and external feature vector to form a comprehensive feature vector matrix. In Flink, the PCA algorithm in the FlinkML library can be used to achieve dimensionality reduction. In order to retain the principal components with 95% variance, the number of principal components to be retained can be determined according to the size of the eigenvalue and the cumulative contribution rate. Normalize the fused feature vector so that its value is within the range of [0, 1], and methods such as MinMaxScaler can be used. In the Apache Flink stream processing framework, real-time updating of the feature vector can be achieved by periodically recalculating the feature vector and performing dimensionality reduction and normalization during the data stream processing. In Flink, functions such as CollectList can be used to collect and combine these feature vectors. Adopt the principal component analysis dimensionality reduction method to project the comprehensive feature vector into a three-dimensional space.First, calculate the covariance matrix of the comprehensive feature vector, then find the eigenvalues and eigenvectors of this matrix, select the eigenvectors corresponding to the top three largest eigenvalues as the principal components, project the original data onto these three principal components to obtain a three-dimensional space vector. A right-handed coordinate system is used to construct the three-dimensional vector space. Select three linearly independent vectors as the basis and label them as i, j, and k. Three main and independent feature directions can be first selected from the three-dimensional space vector after dimensionality reduction as the basis, and then the order of the three directions is determined according to the right-hand rule to ensure the orthogonality and direction consistency of the coordinate system. Assume that the main direction of the device state feature vector in the first principal component direction after dimensionality reduction is the main direction, use it as the x-axis direction, and select the corresponding basis vector i; then select a direction linearly independent of the x-axis from the combined direction of the enterprise financial feature vector and the customer behavior feature vector as the y-axis direction, with the basis vector j; finally, determine a direction linearly independent of both the x and y axes from the combined direction of the human resource feature vector and the external feature vector as the z-axis direction, with the basis vector k, and determine the order of the three directions as x, y, z according to the right-hand rule to construct a right-handed three-dimensional vector space with i, j, and k as the basis.

[0074] In one embodiment, in step S3, the three-dimensional vector space and Apache Flink are used, combined with a time series prediction model, to analyze and predict the comprehensive benefit index of the enterprise, including the following steps:

[0075] Through the stream processing ability of Apache Flink, visualize the vector changes in the three-dimensional vector space; according to the position and change trend of the vector in the three-dimensional vector space, evaluate the real-time cost-benefit status of the enterprise operation in real time;

[0076] Calculate the formula for the comprehensive benefit index:

[0077] Obtain the comprehensive benefit index Q of the enterprise; where, is the modulus of the current state vector, and the current state vector represents the position in the three-dimensional vector space; E(f) represents the information entropy of the vector distribution of the current state vector; f represents the current state feature vector; S t represents the revenue of the enterprise at time t; C t is the cost of the enterprise at time t; α, β, and γ respectively represent the weight coefficients, which are 0.4, 0.3, and 0.3;

[0078] Use the long short-term memory network model to predict the future comprehensive benefit index, and input the historical operation multi-source data set into the long short-term memory LSTM model for training; during the training process, input it into the LSTM model in chronological order, learn the data characteristics and change rules at different time points, and establish a mapping relationship from the past to the present and then to the future;

[0079] Deploy the trained LSTM model to the Apache Flink environment to predict the newly incoming multi-source operation data in real time; when the real-time multi-source operation data enters the Flink stream processing pipeline, according to the current input data and the comprehensive benefit index formula, the prediction result is the future comprehensive benefit index.

[0080] Specifically, through the stream processing ability of Apache Flink, access real-time data streams from data sources such as Kafka and Socket. Use Flink's DataStream API to perform operations such as transforming and aggregating real-time data, and calculate the vector changes in three-dimensional space, including: calculating the values of device status vectors, enterprise financial vectors, etc. at each time point. Use Flink's Table API or SQL API to output the calculation results to visualization tools such as Grafana and Echarts to achieve the visualization of the position and change trend of vectors in three-dimensional space. A three-dimensional line chart can be drawn to show the change trajectory of vectors over time, intuitively reflecting the real-time cost-benefit status of enterprise operations. In Flink, calculate the comprehensive benefit index Q according to the comprehensive benefit index formula; where w1, w2, and w3 are 0.4, 0.3, and 0.3 respectively, as weight coefficients, and are dynamically adjusted according to the actual application situation and historical data. Collect historical multi-source operation data sets, including device operation data, enterprise financial data, customer behavior data, etc., clean and normalize the data, and convert it into a sample or time series format suitable for input to the LSTM model. Use the data of the past n time steps as input to predict the comprehensive benefit index at the n+1th time step. Use deep learning frameworks such as TensorFlow, Keras, or PyTorch to build an LSTM model. Input the prepared data into the LSTM model for training. During the training process, input the data in chronological order, learn the data characteristics and change rules at different time points, and establish a mapping relationship from the past to the present and then to the future. Appropriate optimization algorithms such as Adam and appropriate loss functions such as mean squared error can be used to optimize the training process of the model. Export the trained LSTM model and deploy it to the Apache Flink environment. The LSTM model can be called using Flink's machine learning library or through custom functions. When the real-time multi-source operation data enters the Flink stream processing pipeline, first preprocess the data, and then use the current input data and the comprehensive benefit index formula as features to input into the LSTM model to predict the future comprehensive benefit index, and the output prediction result is the future comprehensive benefit index.

[0081] In one embodiment, the step of analyzing and evaluating the future operation trend and potential risks of the enterprise according to the prediction result in step S3 includes the following steps:

[0082] The comprehensive benefit index at different time points is plotted as a time series graph, where the abscissa represents time and the ordinate represents the comprehensive benefit index;

[0083] Using the linear regression method, with time as the independent variable and the comprehensive benefit index as the dependent variable, a straight line is fitted; the rising or falling trend of the comprehensive benefit index over time is judged by the slope of the fitted straight line; when Q shows an upward trend, it indicates that the future operation trend of the enterprise is good; when Q shows a downward trend, it indicates that the future operation trend of the enterprise is poor;

[0084] According to the historical comprehensive benefit index, different thresholds are divided: when Q > 0.8, it indicates that the operation state of the enterprise is excellent; when 0.4 ≤ Q ≤ 0.8, it indicates that the operation state of the enterprise is good; when Q < 0.4, it indicates that the operation state of the enterprise is dangerous;

[0085] The LSTM model is used to predict the future comprehensive benefit index, and whether there are potential risks in the future operation state of the enterprise is judged according to the set threshold; when the future comprehensive benefit index is less than 0.4, it is identified that there are potential risks, otherwise, there are no potential risks.

[0086] Specifically, organize the comprehensive benefit indexes at different predicted time points into a data set in chronological order, where the abscissa is time and the ordinate is the comprehensive benefit index. Use data visualization tools such as Matplotlib, Seaborn, or Plotly, etc. to draw a time series graph. In Flink, output the prediction results to an external storage system such as HDFS, MySQL, etc., and then use a scripting language such as Python to read the data and draw the graph. Observe the trend of the time series graph. If the overall curve shows an upward trend, that is, the Q value gradually increases with time, it indicates that the future operation trend of the enterprise is good; if the curve shows a downward trend, that is, the Q value gradually decreases with time, it indicates that the future operation trend of the enterprise is poor. Collect the time series data of historical comprehensive benefit indexes and organize them into a data set containing the time independent variable and the Q value dependent variable. Use a linear regression algorithm, such as the LinearRegression model in the scikit-learn library, with time as the independent variable and the Q value as the dependent variable to fit the data. Fit a straight line representing the linear trend of the Q value changing with time. Obtain the slope k of the fitted straight line. If k>0, it means that Q shows an upward trend and the future operation trend of the enterprise is good; if k<0, it means that Q shows a downward trend and the future operation trend of the enterprise is poor. According to the historical comprehensive benefit index data, divide the enterprise operation status into three levels: when Q>0.8, the enterprise operation status is excellent. When 0.4≤Q≤0.8, the enterprise operation status is good. When Q<0.4, the enterprise operation status is dangerous. The division of the threshold is dynamically adjusted according to historical data, experimental combination or actual application. Use the trained LSTM model to predict the future comprehensive benefit index. Compare the predicted future Q value with the set threshold. If the predicted future comprehensive benefit index is less than 0.4, that is, Q<0.4, identify the existence of potential risks, indicating that the future operation status of the enterprise may be in a dangerous state; otherwise, there are no potential risks and the enterprise operation status is relatively good.

[0087] In one embodiment, embedding the knowledge graph into a three-dimensional vector space and constructing a dynamic graph neural network in step S3 includes the following steps:

[0088] Embed the knowledge graph into a three-dimensional vector space. In the knowledge graph, nodes represent entities including: devices, customers, and employees, and edges represent relationships between entities including: the relationship between devices and customers, the relationship between devices and employees, and the relationship between customers and employees;

[0089] Use the TransE model to embed entities into a three-dimensional vector space; introduce time attributes, regard time as a dimension in the graph, convert the time stamp into a sine or cosine periodic vector, splice it with the three-dimensional space vector, and then reduce the dimension to three through a fully connected layer to form a spatio-temporal joint embedding, obtaining node representations containing spatio-temporal information;

[0090] Adopt the EvolveGCN model to evolve the parameters of the graph convolutional network through LSTM, capturing the dynamic changes of the graph structure;

[0091] In combination with the Apache Flink stream processing framework, synchronize the incremental graph structure data every 5 minutes; utilize the stream processing ability of Flink to update the new data into the knowledge graph and update the parameters of the dynamic graph neural network in real time through the EvolveGCN model.

[0092] Specifically, construct a knowledge graph, where the nodes represent entities, including: devices, customers, and employees; the edges represent the relationships between entities, including: the relationship between devices and customers, the relationship between devices and employees, and the relationship between customers and employees. Use the TransE model to embed the entities into a three-dimensional vector space. The core idea of the TransE model is to map the entities and relationships in the knowledge graph into a low-dimensional vector space, and represent the semantic connection between entities and relationships through the distance between vectors. Introduce time attributes to form spatio-temporal joint embedding. Take time as a dimension in the graph, convert the timestamp into a sine or cosine periodic vector, splice it with the spatial vector, and then reduce the dimension to three through a fully connected layer to form spatio-temporal joint embedding, obtaining node representations containing spatio-temporal information. Adopt the EvolveGCN model to evolve the parameters of the graph convolutional network, and use LSTM to evolve the parameters of the graph convolutional network to capture the dynamic changes of the graph structure. The EvolveGCN model adapts the training of the graph convolutional network model in the time dimension, uses RNN to obtain the parameters of the graph convolutional network to capture the dynamic information of the graph sequence. Specifically, in each time step, adjust the parameters of the graph convolutional network model through LSTM to improve the model's adaptability. In combination with the Apache Flink stream processing framework, synchronize the incremental graph structure data every 5 minutes. Utilize the stream processing ability of Flink to update the new data into the knowledge graph and update the parameters of the graph neural network in real time through the EvolveGCN model.

[0093] In one of the embodiments, in step S4, analyze the device anomaly offset, the supply chain anomaly offset, and the customer activity, construct a risk matrix and identify the propagation path of the comprehensive benefit index decline, including the following steps:

[0094] Calculate the average value of each feature parameter in the normal device operation feature vector as the reference value to obtain the reference vector; use the Euclidean distance formula to calculate the offset between the current device state vector and the reference vector in real time to obtain the device anomaly offset;

[0095] Combine the external feature vector, the enterprise financial feature vector, and the human resource feature vector to generate a supply chain state feature vector, and obtain the supply chain anomaly offset in the same way as the calculation method of the device anomaly offset;

[0096] Obtain the activity data of customer nodes, calculate the difference between the activity in the previous period and the activity in the current period, and then divide it by the activity in the previous period to obtain the activity decline rate;

[0097] Take the device anomaly offset, the supply chain anomaly offset, the customer activity, and the timestamp as input signals; introduce a graph attention network model to construct a dynamic graph neural network, and use the graph attention mechanism to automatically learn the relationships between nodes;

[0098] Input the input signals into the constructed dynamic graph neural network for training, calculate the influence degree of the critical path on the potential risk probability through the graph attention mechanism, and output the potential risk probability;

[0099] According to the output results of the dynamic graph neural network, evaluate the risk influence degrees of different nodes and paths, construct a risk conduction matrix, and the elements in the matrix represent the risk conduction intensities between different nodes or among the three business links; the three business links include: the device operation business link, the customer activity business link, and the supply chain business link;

[0100] When the comprehensive benefit index shows a downward trend, use the risk conduction matrix to identify the risk propagation path and the risk influence degree.

[0101] Specifically, collect the data of each characteristic parameter in the normal device operation characteristic vector, calculate the average value of each characteristic parameter, and form a vector with these average values as the reference vector. During the device operation process, obtain the current device state vector in real time, and this vector contains the same characteristic parameters as the reference vector. Then, use the Euclidean distance formula to calculate the distance between the current device state vector and the reference vector, which is the device anomaly offset. The Euclidean distance formula is: Assume the reference vector is (a1, a2,..., an), and the current device state vector is (b1, b2,..., bn), then the device anomaly offset is Collect data such as external feature vectors, enterprise financial feature vectors, and human resource feature vectors, integrate and process these vectors to generate a supply chain status feature vector. After obtaining the supply chain status feature vector, use the same calculation method of the above Euclidean distance formula to obtain the supply chain abnormal offset. Obtain the customer node activity data, and count the activity values in the previous cycle and the current cycle. Calculate the difference in activity between the two cycles, and then divide this difference by the activity in the previous cycle. The resulting value is the activity decline rate, which is used to measure the change degree of customer activity. Take the calculated equipment abnormal offset, supply chain abnormal offset, customer activity, and the corresponding timestamps as input signals. The graph attention network model can automatically learn the complex relationships and interactions between nodes. By constructing a completed dynamic graph neural network and then applying the graph attention mechanism, the model can better capture the dynamic associations and risk transmission paths between different business processes and nodes. Input the above input signals into the constructed dynamic graph neural network for training. During the training process, through the graph attention mechanism, the model can automatically learn and calculate the influence degree of the key path on the potential risk probability, and finally output the potential risk probability, providing a quantitative basis for risk assessment. According to the output results of the dynamic graph neural network, evaluate the risk influence degree of different nodes and paths, and determine which nodes and paths play a key role in risk propagation. Construct a matrix, and the elements in the matrix represent the intensity of risk conduction between different nodes or the three business processes. The three business processes include the equipment operation business process, the customer activity business process, and the supply chain business process. Through this matrix, the conduction situation of risks between each business process and node can be intuitively displayed. When the comprehensive benefit index shows a downward trend, use the constructed risk conduction matrix. By analyzing the risk conduction intensity of each node and path in the matrix, determine the risk propagation direction, key path, and risk influence degree, so as to help the enterprise take effective risk control measures in advance. The key path refers to the propagation path in the risk conduction matrix constructed by the dynamic graph neural network that has a greater influence on the potential risk probability. The nodes and edges on these paths represent the key links of risk propagation, and have a direct or important impact on the decline of the comprehensive benefit index.

[0102] In one embodiment, the step S4 of dynamically allocating business process resources by using the analysis result of the risk conduction matrix includes the following steps:

[0103] Adopt the feature vector weighting method to obtain the contribution weights of the three business processes of equipment operation, customer activity, and supply chain to the comprehensive benefit index respectively;

[0104] According to the intensity of risk conduction between the three business processes in the risk conduction matrix, calculate the sensitivity of the three business processes during the risk conduction process respectively;

[0105] Automatically learn the relationships between nodes through the graph attention mechanism, calculate the influence degree of the critical path on the potential risk probability, and determine the influence rankings of the three business processes;

[0106] Combining the contribution weights of the three business processes to the comprehensive benefit index, the risk conduction sensitivity, and the influence rankings obtained from the graph attention network, construct a resource allocation weight function to determine the weight coefficients of resource allocation;

[0107] After normalizing the obtained resource allocation weight coefficients of the three business processes respectively, sort the three business processes respectively to determine the resource allocation response priorities of the three business processes.

[0108] Specifically, adopt the eigenvector weighting method to calculate the contribution weights of the three business processes of equipment operation, customer activity, and supply chain to the comprehensive benefit index respectively, and determine the importance of each business process in the comprehensive benefit index. Collect the relevant historical data of the three business processes of enterprise equipment operation, customer activity, and supply chain, and clean the collected historical data to remove outliers, missing values, and duplicate values, etc. Then, perform normalization processing on the data with different dimensions and magnitudes, and map them to the same scale range, such as the [0,1] interval. Construct a data matrix from the preprocessed historical data of the three business processes and the comprehensive benefit index data, where each row represents a time point or sample, and each column corresponds to the index data of the three business processes of equipment operation, customer activity, and supply chain, as well as the comprehensive benefit index; calculate the covariance matrix between the index data of the three business processes in the data matrix and the comprehensive benefit index to measure the strength of their linear relationship and the direction of correlation; perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues and corresponding eigenvectors. Sort the eigenvectors in descending order of eigenvalues, and select the first three eigenvectors to form an eigenvector matrix; project the original data matrix onto the eigenvector matrix to obtain the principal component score matrix. Determine the contribution weights of each business process to the comprehensive benefit index according to the variance contribution rate of each principal component, that is, the proportion of the eigenvalue to the total eigenvalue, that is, the contribution weight is equal to the sum of the squares of the coefficients of the corresponding business process in each principal component multiplied by the variance contribution rate of the principal component, and then normalize the contribution weights of the three business processes to make their sum equal to 1.

[0109] According to the intensity of risk transmission between the three business processes in the risk transmission matrix, calculate the sensitivity of each of the three business processes during the risk transmission process. For each business process, calculate the average value of the risk transmission intensity between it and the other two business processes. The contribution weight of each business process to the comprehensive benefit index obtained by using the eigenvector weighting method is the weight value calculated in the previous step. The formula for the risk transmission sensitivity coefficient is equal to the average value of the risk transmission intensity multiplied by the business process weight multiplied by the risk impact coefficient of the business process, and calculate the risk transmission sensitivity coefficient of each business process;

[0110] Automatically learn the relationships between nodes through the graph attention mechanism, calculate the degree of influence of the critical path on the potential risk probability, determine the influence rankings of the three business processes based on the importance and roles of the business processes on the critical path, and clarify the priorities of each business process in risk propagation. According to the feature representations of each node output by the graph attention network, calculate the importance score of each node. For example, a certain pooling operation or readout function can be used to map the node features to a scalar value, indicating the influence of the node in the entire graph. Then rank the three business processes according to the influence scores to determine their criticality in the risk transmission network.

[0111] Combining the contribution weights of the three business processes to the comprehensive benefit index, the risk transmission sensitivity, and the influence rankings obtained from the graph attention network, construct a resource allocation weight function. Using this function, comprehensively consider various factors to determine the resource allocation weight coefficients of each business process to quantify the relative importance of each business process in resource allocation. A resource allocation weight function can be constructed. For example, multiply the contribution weight, the risk transmission sensitivity coefficient, and the influence ranking score by their respective influence coefficients, and then add them up to obtain the comprehensive resource allocation weight of each business process. Among them, the influence coefficient of the contribution weight is 0.4, the influence coefficient of the risk transmission sensitivity coefficient is 0.4, and the influence coefficient of the influence ranking score is 0.2; the determination of the influence coefficients can be adjusted according to the enterprise's emphasis on different factors and obtained through optimization algorithms or expert experience. Normalize the resource allocation weight coefficients of the three business processes obtained respectively to make them in the same dimension and range for easy comparison and allocation. Then rank the three business processes respectively. The business process with a larger resource allocation weight function is given priority in resource allocation response, and resources are reasonably allocated according to the priority order. Develop an adaptive scheduling engine to automatically trigger emergency plans and dynamically adjust resource allocation according to real-time risk changes and resource usage conditions.

[0112] The above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A real-time data analysis method for enterprise digital operations based on multi-source data fusion, characterized in that: The following steps are involved: S1: Real-time collection of multi-source operational data during enterprise operations, construction of multi-source operational data sets, and pre-processing; S2: Extract features based on the preprocessed operational multi-source dataset and construct a three-dimensional vector space; S3: Utilizes 3D vector space and Apache Flink, combined with a time series forecasting model, to analyze and predict the comprehensive efficiency index of an enterprise. Based on the forecast results, it analyzes and evaluates the trends and potential risks of future operations. It also embeds the knowledge graph into the 3D vector space to construct a dynamic graph neural network. S4: Analyze abnormal equipment deviations, abnormal supply chain deviations, and customer activity, build a risk matrix, and identify the propagation path of the decline in the comprehensive benefit index; use the analysis results of the risk transmission matrix to dynamically allocate business link resources; S5: Use the dashboard to visualize the company's future operating trends, comprehensive benefit index, and the path of the decline in the comprehensive benefit index.

2. The method for real-time data analysis of enterprise digital operations based on multi-source data fusion according to claim 1 is characterized in that: In step S1, the multi-source data set for operation includes: equipment operation data, enterprise financial data, customer behavior data, human resources data and external data.

3. The method for real-time data analysis of enterprise digital operations based on multi-source data fusion according to claim 1 is characterized in that: The step S2 comprises the following steps: Through Apache Flink, we access pre-processed operational multi-source data sets and extract the maximum, minimum, and average values of equipment operation data, enterprise financial data, customer behavior data, and human resources data as statistical feature values. Extract the difference between capital inflow and outflow and capital turnover rate from corporate financial data; extract customer retention rate and customer activity from customer behavior data; extract employee turnover rate from human resources data; The feature values extracted from external data include: the busyness and traffic congestion level of the enterprise's geographical area; Based on the results of feature extraction of the above multi-source operational data sets, the feature vectors are combined into equipment status feature vectors, enterprise financial feature vectors, customer behavior feature vectors, human resource feature vectors and external feature vectors; The five constructed eigenvectors are combined to form a comprehensive eigenvector matrix. Principal component analysis is used for dimensionality reduction, with the principal component retaining 95% of the variance to normalize the fused eigenvectors. The eigenvectors are then updated in real time within the Apache Flink stream processing framework. The comprehensive eigenvector matrix is projected into three-dimensional space, and a right-hand coordinate system is used to construct a three-dimensional space vector, where the x-axis is the equipment status eigenvector, the y-axis is the enterprise financial eigenvector and the customer behavior eigenvector, and the z-axis is the human resources eigenvector and the external eigenvector; three linearly independent vectors are selected as the basis, marked as i, j, and k, and the order of the three directions is determined according to the right-hand rule.

4. The method for real-time data analysis of enterprise digital operations based on multi-source data fusion according to claim 3 is characterized in that: In step S3, the three-dimensional vector space and Apache Flink are used in combination with a time series prediction model to analyze and predict the comprehensive benefit index of the enterprise, including the following steps: By leveraging Apache Flink's stream processing capabilities, we can visualize vector changes in a three-dimensional vector space. Based on the position and change trends of vectors in the three-dimensional vector space, we can assess the cost-effectiveness of enterprise operations in real time. Formula for calculating comprehensive benefit index: Get the comprehensive benefit index Q of the enterprise; among them, is the modulus of the current state vector, which represents the position in the three-dimensional vector space; E(f) represents the information entropy of the vector distribution of the current state vector; f represents the current state eigenvector; S t represents the profit of the enterprise at time t; C t is the enterprise's cost at time t; α, β, and γ represent weight coefficients, which are 0.4, 0.3, and 0.3 respectively; Using a long-short-term memory (LSTM) model to predict future comprehensive benefit indices, the multi-source historical operational data set is fed into the LSTM model for training. During training, the data is fed into the LSTM model in chronological order, learning the data characteristics and change patterns at different time points and establishing a mapping relationship from the past to the present and then to the future. The trained LSTM model is deployed in the Apache Flink environment to make real-time predictions on newly incoming multi-source operational data. When the real-time multi-source operational data enters the Flink stream processing pipeline, the output prediction result is the future comprehensive benefit index based on the current input data and the comprehensive benefit index formula.

5. The method for real-time data analysis of enterprise digital operations based on multi-source data fusion according to claim 4 is characterized in that: The step S3 analyzes and evaluates the trend and potential risks of the enterprise's future operations based on the forecast results, including the following steps: The comprehensive benefit index at different time points is predicted to be drawn into a time series graph, where the horizontal axis represents time and the vertical axis represents the comprehensive benefit index; Using the linear regression method, with time as the independent variable and the comprehensive benefit index as the dependent variable, a straight line is fitted. The slope of the fitted line is used to determine whether the comprehensive benefit index is increasing or decreasing over time. When Q shows an upward trend, it indicates that the company's future operating trend is positive; when Q shows a downward trend, it indicates that the company's future operating trend is negative. Based on the historical comprehensive benefit index, different thresholds are set: when Q>0.8, the enterprise's operating status is excellent; when 0.4≤Q≤0.8, the enterprise's operating status is good; when Q<0.4, the enterprise's operating status is dangerous; The LSTM model is used to predict the future comprehensive benefit index and determine whether there are potential risks in the future enterprise operation status based on the set threshold. When the future comprehensive benefit index is less than 0.4, the potential risk is identified; otherwise, there is no potential risk.

6. The method for real-time data analysis of enterprise digital operations based on multi-source data fusion according to claim 1 is characterized in that: In step S3, the knowledge graph is embedded in the three-dimensional vector space to construct a dynamic graph neural network, which includes the following steps: The knowledge graph is embedded into a three-dimensional vector space. In the knowledge graph, nodes represent entities, including devices, customers, and employees. Edges represent relationships between entities, including the relationship between devices and customers, the relationship between devices and employees, and the relationship between customers and employees. The TransE model is used to embed entities into a three-dimensional vector space. The time attribute is introduced, and time is used as a dimension in the graph. The timestamp is converted into a sine or cosine periodic vector. After being concatenated with the three-dimensional space vector, the dimensionality is reduced to three dimensions through a fully connected layer to form a joint spatiotemporal embedding, resulting in a node representation containing spatiotemporal information. The Evo lveGCN model is used to evolve the graph convolutional network parameters through LSTM to capture the dynamic changes of the graph structure; Combined with the Apache Flink stream processing framework, incremental graph structure data is synchronized every 5 minutes. By leveraging Flink's stream processing capabilities, new data is updated to the knowledge graph, and the parameters of the dynamic graph neural network are updated in real time through the EvolveGCN model.

7. The method for real-time data analysis of enterprise digital operations based on multi-source data fusion according to claim 6 is characterized in that: The step S4 analyzes the abnormal device offset, the abnormal supply chain offset, and the customer activity, constructs a risk matrix, and identifies the propagation path of the decline in the comprehensive benefit index, including the following steps: The reference vector is obtained by calculating the average value of each characteristic parameter in the normal equipment operation characteristic vector as the reference value; the offset between the current equipment state vector and the reference vector is calculated in real time using the Euclidean distance formula to obtain the equipment abnormality offset; Combining the external feature vector, enterprise financial feature vector and human resource feature vector, the supply chain status feature vector is generated. Similarly, the calculation method of equipment abnormal offset is used to obtain the supply chain abnormal offset. Obtain client node activity data, calculate the difference between the activity of the previous cycle and the activity of the current cycle, and then divide it by the activity of the previous cycle to obtain the activity decline rate; The device anomaly offset, supply chain anomaly offset, customer activity, and timestamp are used as input signals. The graph attention network model is introduced to construct a dynamic graph neural network, and the graph attention mechanism is used to automatically learn the relationship between nodes. The input signal is fed into the constructed dynamic graph neural network for training. The graph attention mechanism is used to calculate the impact of the critical path on the potential risk probability and output the potential risk probability. Based on the output of the dynamic graph neural network, the risk impact of different nodes and paths is evaluated, and a risk transmission matrix is constructed. The elements in the matrix represent the intensity of risk transmission between different nodes or three business links; the three business links include: equipment operation business link, customer activity business link, and supply chain business link; When the comprehensive benefit index shows a downward trend, the risk transmission matrix is used to identify the risk transmission path and the degree of risk impact.

8. The method for real-time data analysis of enterprise digital operations based on multi-source data fusion according to claim 7 is characterized in that: In step S4, the analysis results of the risk transmission matrix are used to dynamically allocate business link resources, including the following steps: The characteristic vector weighting method is used to obtain the contribution weights of the three business links of equipment operation, customer activity, and supply chain to the comprehensive benefit index; According to the intensity of risk transmission between the three business links in the risk transmission matrix, calculate the sensitivity of the three business links in the risk transmission process respectively; The graph attention mechanism automatically learns the relationship between nodes, calculates the impact of the critical path on the potential risk probability, and determines the influence ranking of the three business links; Combining the contribution weights of the three business links to the comprehensive benefit index, as well as the risk transmission sensitivity and the influence ranking derived from the graph attention network, a resource allocation weight function is constructed to determine the weight coefficient of resource allocation; After normalizing the obtained resource allocation weight coefficients of the three business links, the three business links are sorted respectively to determine the resource allocation response priorities of the three business links.

Images