Multi-agent operation management method and system based on industrial internet platform

Abstract

The invention discloses a multi-subject operation management method based on an industrial internet platform, and the method comprises the steps: designing a diversified evaluation index system covering the core tasks and key performances of different subjects according to the role difference and task difference of all subjects in the industrial internet platform; the method comprises the following steps: acquiring equipment operation data, user behavior data and transaction record data from a plurality of data sources of a platform by adopting a distributed data acquisition technology to obtain a multi-source heterogeneous data set; performing noise removal and data fusion on the multi-source heterogeneous data set to obtain a standardized data set; extracting core indexes such as equipment operation efficiency, user activeness and transaction completion rate from the standardized data set to obtain a key information set; constructing an evaluation model according to the key information set, and quantitatively associating the core indexes of different subjects with the operation performance; the influence degree of each index on the operation performance is judged through causal inference and comparative analysis; and dynamically presenting the evaluation result and the comparative analysis result in a visual mode.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial Internet, and particularly to a multi-agent operation management method and system based on an industrial Internet platform. Background Art

[0002] With the rapid development of the industrial Internet, the multi-agent operation management of industrial Internet platforms has become an important topic in current research and practice. Industrial Internet platforms involve multiple agents, such as equipment manufacturers, system integrators, service providers, end-users, etc. These agents play different roles in the platform and undertake diverse tasks and responsibilities. However, how to comprehensively, objectively, and accurately evaluate the operation effects of these different agents remains a complex technical problem to be solved urgently.

[0003] Currently, the operation evaluation of industrial Internet platforms faces challenges in many aspects. First, the operation goals and business processes of different agents vary greatly, resulting in difficulties in uniformly setting evaluation indicators, and the characteristics and demands of each agent need to be fully considered. Second, the data generated during the operation process is characterized by being massive, multi-source, and heterogeneous, which brings great difficulties to data collection, processing, and analysis. Traditional data analysis methods are difficult to effectively mine the key information in these data, and thus it is difficult to construct a scientific and reasonable evaluation model.

[0004] In addition, the interpretability and credibility of evaluation results are key factors affecting their application value. In existing technologies, methods such as causal inference and comparative analysis have been applied in some fields, but their application in the multi-agent operation evaluation of industrial Internet is still immature. How to clarify the internal relationship between various indicators and operation performance through these technologies and present the evaluation results in an intuitive way is a technical bottleneck that needs to be broken through currently.

[0005] The operation of industrial Internet platforms is a dynamic process, and evaluation should not only stay in the stage of post-event analysis. Currently, the application of the closed-loop mechanism of real-time monitoring, dynamic early warning, and continuous optimization in the operation of industrial Internet platforms is not perfect enough. How to achieve real-time monitoring of the whole process, early warning of potential risks and hidden dangers, and dynamic adjustment of operation strategies is crucial for improving the overall operation efficiency and service quality of industrial Internet platforms. Summary of the Invention

[0006] The present invention provides a multi-agent operation management method and system based on an industrial Internet platform to solve the above-mentioned existing technical problems.

[0007] The technical solution of the present invention is realized as follows:

[0008] A multi-agent operation management method based on an industrial Internet platform includes the following steps:

[0009] Design a diversified evaluation index system covering the core tasks and key performance of different entities according to the role differences and task differences of each entity in the industrial Internet platform;

[0010] Adopt distributed data acquisition technology to obtain device operation data, user behavior data, and transaction record data from multiple data sources of the platform, and obtain a multi-source heterogeneous data set;

[0011] Perform noise removal and data fusion on the multi-source heterogeneous data set to obtain a standardized data set;

[0012] Extract core indicators such as device operation efficiency, user activity, and transaction completion rate from the standardized data set to obtain a key information set;

[0013] Construct an evaluation model based on the key information set to quantitatively associate the core indicators of different entities with operation performance;

[0014] Use causal inference and comparative analysis to judge the influence degree of each indicator on operation performance;

[0015] Dynamically present the evaluation results and comparative analysis results in a visual way;

[0016] Conduct real-time monitoring of the evaluation results through time series analysis, generate dynamic warning signals in combination with warning thresholds, continuously optimize operation strategies and service quality, and obtain an optimization plan for multi-entity collaboration efficiency.

[0017] Furthermore, the process of designing the diversified evaluation index system includes:

[0018] Adopt a preset classification method to divide the role differences and task differences of entity classes in the industrial Internet platform, and obtain the characteristic data of entity classes;

[0019] According to the characteristic data of entity classes, design diversified evaluation indicators covering core tasks and key performance, and form an evaluation index framework;

[0020] Through a preset weight assignment method, conduct systematic processing on the evaluation indicators to determine the weight values of each indicator;

[0021] Combine the coverage range and performance evaluation to optimize the evaluation index framework and obtain the final evaluation index system.

[0022] Furthermore, the process of obtaining the multi-source heterogeneous data set includes:

[0023] Adopt distributed data acquisition technology to obtain device operation data, user behavior data, and transaction record data from the platform to obtain a multi-source heterogeneous data set;

[0024] For multi-source heterogeneous data sets, noise data and redundant data are removed to obtain a cleaned data set.

[0025] Further, the process of obtaining the standardized data set includes:

[0026] According to the cleaned data set, a data fusion technology is used to unify the format and structure of the data to obtain a standardized data set;

[0027] It also includes a data feature extraction process. Specifically, according to the cleaned data set, key features are extracted from the operation data, behavior data, and record data to form a feature matrix.

[0028] Further, the process of obtaining the key information set includes:

[0029] An association rule mining algorithm is used to extract core indicators including equipment operation efficiency, user activity, and transaction record completion rate from the standardized data set to obtain a key information set;

[0030] It also includes, according to the key information set, using preset threshold conditions to screen the equipment operation efficiency, user activity, and transaction completion rate to obtain a screened information set;

[0031] For the screened information set, the equipment, users, and transactions are classified to obtain a classification result set.

[0032] Further, the process of constructing an evaluation model based on the key information set and quantitatively associating the core indicators of different entities with the operation performance includes:

[0033] Obtain the equipment operation efficiency, user activity, and transaction completion rate data from the key information set as the input variables of the multiple linear regression algorithm, and the operation performance as the output variable;

[0034] Use the multiple linear regression algorithm to fit the input variables and output variables, determine the contribution coefficients of the equipment operation efficiency, user activity, and transaction completion rate to the operation performance according to the fitting results, and construct an evaluation model;

[0035] Extract new data from the standardized data set, input it into the evaluation model, and calculate the weighted scores of the equipment operation efficiency, user activity, and transaction completion rate.

[0036] Further, the process of judging the influence degree of each indicator on the operation performance includes:

[0037] Use causal inference technology to analyze the causal relationship between the equipment operation efficiency, user activity, and transaction completion rate and the operation performance, and obtain the causal intensity value;

[0038] According to the causal intensity value, horizontally compare the evaluation results of equipment, users, and transactions by combining the comparative analysis technology to obtain the horizontal difference value;

[0039] According to the horizontal difference value, vertically compare the evaluation results of equipment, users, and transactions by combining the comparative analysis technology to obtain the vertical difference value;

[0040] Based on the horizontal difference value and the vertical difference value, judge the influence degree of equipment operation efficiency, user activity, and transaction completion rate on operation performance to obtain the influence degree value.

[0041] Furthermore, the process of dynamically presenting in a visual manner includes:

[0042] Adopt data visualization technology to display the evaluation results and comparative analysis results in the form of charts, generate a dynamic visualization report, and determine the intuitive performance of the evaluation results;

[0043] According to the chart data in the visualization report, extract the characteristic values of key indicators to obtain a set of characteristic values;

[0044] Classify the key indicators according to the set of characteristic values to obtain the classification result;

[0045] According to the classification result, use the regression analysis method to predict the trend of key indicators to obtain the prediction result.

[0046] The multi-agent operation management system based on the industrial Internet platform includes a data collection and preprocessing module, an evaluation index and model management module, a data analysis and visualization module, a real-time monitoring and early warning module, and an operation optimization and decision support module;

[0047] The evaluation index and model management module is used to design and manage the evaluation index system and evaluation model for multi-agent operation. Specifically, according to the differences in the roles and tasks of the agents, design diverse evaluation indicators covering core tasks and key performance; construct an evaluation model to quantify the correlation between core indicators and operation performance; perform systematic processing on the evaluation indicators through a preset weight allocation method, and optimize according to the coverage and performance evaluation;

[0048] The data analysis and visualization module extracts key information from the standardized dataset and presents the analysis results in an intuitive manner. Specifically, use the association rule mining algorithm to extract core indicators; display the evaluation results and comparative analysis in the form of dynamic charts and dashboards to enhance the interpretability of the results; perform trend prediction based on the historical data of key indicators.

[0049] The data collection and preprocessing module is responsible for collecting data from multiple data sources of the industrial Internet platform and performing preprocessing;

[0050] The real-time monitoring and early warning module monitors the operation process in real time and generates dynamic early warning signals. Specifically, it monitors the evaluation results in real time, analyzes the time-varying trend of data, and generates dynamic early warning signals in real time according to the preset threshold conditions to timely discover potential problems. Combining the real-time monitoring data, it conducts early warning of possible risk hazards and supports the dynamic adjustment of operation strategies;

[0051] The operation optimization and decision support module dynamically adjusts the operation strategy according to the early warning signal and the evaluation result to improve the operation efficiency. Combining the output of the evaluation model, it optimizes the service quality to meet the needs of different entities and provides a decision support tool based on data analysis.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] 1. By designing a diversified evaluation index system, the present invention fully considers the role and task differences of different entities in the industrial Internet platform, uses a preset classification method to divide the roles and tasks of entities, obtains feature data, designs diversified evaluation indexes covering core tasks and key performance based on these feature data, and systematically processes the indexes through a weight assignment method to determine the weight values of each index. Combining the coverage and performance evaluation, it optimizes the evaluation index framework to form the final evaluation index system, effectively solving the problem that it is difficult to uniformly set evaluation indexes due to the differences in operation objectives and business processes of different entities, and ensuring the comprehensiveness and pertinence of the evaluation indexes;

[0054] 2. Aiming at the characteristics of massive, multi-source and heterogeneous data in the industrial Internet platform, the present invention proposes a systematic data processing method. It uses distributed data acquisition technology to obtain device operation data, user behavior data and transaction record data from multiple data sources of the platform to form a multi-source heterogeneous data set, and obtains a standardized data set. It extracts key features from the standardized data set to form a feature matrix, further improving the usability of the data, effectively solving the difficulties in data acquisition, processing and analysis, and providing high-quality data support for the subsequent construction of the evaluation model;

[0055] 3. The interpretability and credibility of the evaluation results are improved through causal inference and comparative analysis techniques. The causal inference technique is used to analyze the causal relationship between device operation efficiency, user activity and transaction completion rate and operation performance to obtain the causal intensity value. Combining the comparative analysis technique, the evaluation results of devices, users and transactions are compared horizontally and vertically to obtain the difference value, so as to judge the influence degree of each index on operation performance. In addition, through data visualization technology, the evaluation results and comparative analysis results are presented in the form of dynamic charts to intuitively display the characteristic values and trend predictions of key indicators. This process not only enhances the interpretability of the evaluation results, but also improves their application value;

[0056] 4. Real-time monitoring and dynamic optimization of the operation of the industrial Internet platform are achieved through time series analysis and a dynamic early warning mechanism. First, the time series analysis method is used to conduct real-time monitoring on the evaluation results, and dynamic early warning signals are generated in combination with preset early warning thresholds. Then, according to the early warning signals, the operation strategies and service quality are adjusted in a timely manner to form a closed-loop mechanism for dynamic optimization. This not only solves the problems of insufficient real-time monitoring and untimely early warning, but also improves the overall operation efficiency and service quality of the platform, achieving dynamic optimization of the entire process. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a flowchart of the method for multi-agent operation management method based on the industrial Internet platform in Embodiment 1;

[0058] Figure 2 It is a system framework diagram of the multi-agent operation management system based on the industrial Internet platform in Embodiment 2. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0059] In order to make the objectives, features, and advantages of the present invention more obvious and understandable, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the embodiments described below are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0060] Embodiment 1

[0061] As Figure 1 shown, this embodiment provides a multi-agent operation management method based on the industrial Internet platform, including the following steps:

[0062] Design a diversified evaluation index system covering the core tasks and key performance of different agents according to the role differences and task differences of each agent in the industrial Internet platform;

[0063] Adopt distributed data acquisition technology to obtain device operation data, user behavior data, and transaction record data from multiple data sources of the platform to obtain a multi-source heterogeneous data set;

[0064] Perform noise removal and data fusion on the multi-source heterogeneous data set to obtain a standardized data set;

[0065] Extract core indicators such as device operation efficiency, user activity, and transaction completion rate from the standardized data set to obtain a key information set;

[0066] Construct an evaluation model according to the key information set to quantitatively associate the core indicators of different agents with the operation performance;

[0067] Use causal inference and comparative analysis to judge the impact degree of each index on operation performance;

[0068] Dynamically present the evaluation results and comparative analysis results in a visual way;

[0069] Through time series analysis, conduct real-time monitoring on the evaluation results, generate dynamic warning signals in combination with warning thresholds, continuously optimize operation strategies and service quality, and obtain an optimized solution for multi-agent collaboration efficiency.

[0070] Furthermore, the process of designing a diverse evaluation index system includes:

[0071] Adopt a preset classification method to divide the role differences and task differences of the entity classes in the industrial Internet platform, and obtain the characteristic data of the entity classes;

[0072] According to the characteristic data of the entity classes, design diverse evaluation indexes covering core tasks and key performance, and form an evaluation index framework;

[0073] Through a preset weight assignment method, conduct systematic processing on the evaluation indexes to determine the weight values of each index;

[0074] Combine the coverage range and performance evaluation to optimize the evaluation index framework and obtain the final evaluation index system;

[0075] If the task differences of the entity classes change, then update the evaluation index system according to the preset dynamic adjustment rules.

[0076] , group the entity classes in the industrial Internet platform, judge the applicability of the evaluation indexes for each group, and generate evaluation reports for each entity class according to the grouping results and the evaluation index system;

[0077] Specifically, it can be as follows. Adopt a preset classification method to divide the entity classes in the industrial Internet platform into three categories: equipment, users, and transactions, and obtain the characteristic data of each category;

[0078] According to the characteristic data of the equipment category, design evaluation indexes such as equipment operation efficiency, and set the target value to 95%. Through a preset weight assignment method, use the analytic hierarchy process to determine the weight of equipment operation efficiency as 0.4, user activity as 0.3, and transaction completion rate as 0.3;

[0079] Combine the coverage range and performance evaluation to optimize the evaluation index framework and obtain the final evaluation index system. If the task differences of the user category change, then according to the dynamic adjustment rules, adjust the weight of user activity from 0.3 to 0.35 and update the evaluation index system;

[0080] The K-means clustering algorithm is used to group the main classes, and the number of clusters is set to 3 to determine the applicability of the evaluation indicators of each group.

[0081] Furthermore, the process of obtaining a multi-source heterogeneous data set includes:

[0082] Distributed data collection technology is used to obtain device operation data, user behavior data, and transaction record data from the platform to obtain multi-source heterogeneous data sets;

[0083] For multi-source heterogeneous data sets, remove noise data and redundant data to obtain the cleaned data set.

[0084] Furthermore, the process of obtaining the standardized data set includes:

[0085] According to the cleaned data set, data fusion technology is used to unify the data format and structure to obtain a standardized data set; if there is abnormal data in the cleaned data set, data cleaning technology is used to further remove the abnormal data to obtain a standardized data set, and data fusion technology is used to integrate the standardized data set into a unified format and structure to obtain a fused data set;

[0086] It also includes a data feature extraction process, specifically, extracting key features from the operation data, behavior data and record data according to the cleaned data set to form a feature matrix;

[0087] If the dimension of the feature matrix is ​​high, the principal component analysis technique is used to reduce the dimension of the high-dimensional features to obtain the feature matrix after dimension reduction. According to the feature matrix after dimension reduction, the subjects in the data set are classified to obtain the subject grouping result;

[0088] Specifically, as described below, distributed data collection technology can be used to obtain device operation data, user behavior data, and transaction record data from multiple data sources on the platform to obtain multi-source heterogeneous data sets. For example, information such as temperature, click events, and order amounts can be collected from sensors, log systems, and transaction databases respectively;

[0089] According to the multi-source heterogeneous data set, data cleaning technology is used to remove noise data and redundant data, such as using regular expressions to filter out irrelevant characters in the log and deleting duplicate transaction records to obtain a cleaned data set. If there is abnormal data in the cleaned data set, data cleaning technology is used to further remove the abnormal data, such as using box plots to detect and remove outliers in temperature data to obtain a standardized data set.

[0090] Use data fusion technology to integrate standardized data sets into a unified format and structure. For example, align the timestamp of device operation data with the timestamp of user behavior data and convert them into a unified JSON format to obtain a fused data set.

[0091] Based on the fused dataset, key features are extracted from operation data, behavior data, and record data using feature extraction techniques. For example, the TF-IDF algorithm is used to extract high-frequency keywords from user behavior logs, and the mean and variance of transaction amounts are calculated to form a feature matrix. If the dimension of the feature matrix is high, principal component analysis technology is used to reduce the dimension of the high-dimensional features. For example, the 100-dimensional features are reduced to 10 dimensions through the PCA algorithm to obtain the dimension-reduced feature matrix;

[0092] Based on the dimension-reduced feature matrix, the entities in the dataset are classified. For example, the K-Means algorithm is used to divide users into 5 groups to obtain the entity grouping result.

[0093] Furthermore, the process of obtaining the key information set includes:

[0094] The association rule mining algorithm is used to extract core indicators including device operation efficiency, user activity, and transaction record completion rate from the standardized dataset to obtain the key information set;

[0095] Data on device operation efficiency, user activity, and transaction completion rate are extracted from the key information set as input variables for the multiple linear regression algorithm. The multiple linear regression algorithm is used to fit the input variables and operation performance, and the goodness of fit is calculated. If the goodness of fit is lower than the preset threshold, the variable combination or weight of device operation efficiency, user activity, and transaction completion rate is adjusted, and the multiple linear regression fitting is performed again;

[0096] According to the fitting results, the contribution coefficients of device operation efficiency, user activity, and transaction completion rate to operation performance are determined, an evaluation model is constructed, new data is extracted from the standardized dataset, input into the evaluation model, the weighted scores of device operation efficiency, user activity, and transaction completion rate are calculated, and the weighted scores of devices, users, and transactions are classified to generate classification labels;

[0097] It also includes screening device operation efficiency, user activity, and transaction completion rate according to preset threshold conditions based on the key information set to obtain the screened information set;

[0098] For the screened information set, devices, users, and transactions are classified to obtain the classification result set;

[0099] Specifically, it can be described as follows. The association rule mining algorithm is used to perform data mining on the initial dataset to extract the association rules between device operation efficiency, user activity, and transaction completion rate. For example, the Apriori algorithm is used to set the minimum support degree to 0.5 and the minimum confidence degree to 0.7 to mine the strong association rules between device operation efficiency and user activity;

[0100] Generate a key information set including device operation efficiency, user activity, and transaction completion rate according to the mining results of association rules. For example, filter out records where the device operation efficiency is greater than 80%, the user activity is greater than 70%, and the transaction completion rate is greater than 85%.

[0101] Extract data on device operation efficiency, user activity, and transaction completion rate from the key information set as input variables for the multiple linear regression algorithm. For example, the device operation efficiency is 85%, the user activity is 72%, and the transaction completion rate is 90%.

[0102] Use the multiple linear regression algorithm to fit the input variables and operation performance, and calculate the goodness of fit. For example, the goodness of fit is 0.85. If the goodness of fit is lower than the preset threshold of 0.9, adjust the variable combination or weight of device operation efficiency, user activity, and transaction completion rate. For example, adjust the weight of device operation efficiency from 0.4 to 0.5 and perform multiple linear regression fitting again.

[0103] According to the fitting results, determine the contribution coefficients of device operation efficiency, user activity, and transaction completion rate to operation performance. For example, the contribution coefficient of device operation efficiency is 0.5, the user activity is 0.3, and the transaction completion rate is 0.2, and construct an evaluation model.

[0104] Extract new data from the standardized data set, input it into the evaluation model, and calculate the weighted scores of device operation efficiency, user activity, and transaction completion rate. For example, the score of device operation efficiency is 85, the score of user activity is 72, and the score of transaction completion rate is 90.

[0105] Classify the weighted scores of devices, users, and transactions. For example, use the K-means algorithm with K = 3 to generate classification labels.

[0106] Furthermore, the process of constructing an evaluation model based on the key information set and quantitatively associating the core indicators of different entities with operation performance includes:

[0107] Obtain data on device operation efficiency, user activity, and transaction completion rate from the key information set as input variables for the multiple linear regression algorithm, and operation performance as the output variable.

[0108] Use the multiple linear regression algorithm to fit the input variables and output variables, and determine the contribution coefficients of device operation efficiency, user activity, and transaction completion rate to operation performance according to the fitting results, and construct an evaluation model.

[0109] Extract new data from the standardized data set, input it into the evaluation model, and calculate the weighted scores of device operation efficiency, user activity, and transaction completion rate.

[0110] Classify the weighted scores of devices, users, and transactions to obtain classification labels. For the classification labels, use principal component analysis technology to reduce the dimensions of the scores of device operation efficiency, user activity, and transaction completion rate, and obtain the dimensionality reduction result;

[0111] According to the dimensionality reduction result, use visualization technology to generate the index distribution maps of devices, users, and transactions, and obtain the distribution result;

[0112] According to the distribution result, use association rule mining technology to mine the association relationships among device operation efficiency, user activity, and transaction completion rate, and obtain the association rule set.

[0113] Specifically, as described below, extract the data of device operation efficiency, user activity, and transaction completion rate from the key information set. For example, the device operation efficiency is 85%, the user activity is 60%, and the transaction completion rate is 90%, and use them as the input variables of the multiple linear regression algorithm;

[0114] Use the multiple linear regression algorithm to fit the input variables and operation performance. For example, the input variables are X1, X2, X3, and the operation performance is Y, and the fitting formula is Y = 0.3X1 + 0.5X2 + 0.2X3, and obtain the preliminary fitting result. If the fitting degree is lower than the preset threshold of 0.8, adjust the variable combination or weight. For example, adjust the formula to Y = 0.4X1 + 0.4X2 + 0.2X3, and perform multiple linear regression fitting again to obtain the optimized fitting result;

[0115] According to the optimized fitting result, determine that the contribution coefficients of device operation efficiency, user activity, and transaction completion rate to operation performance are 0.4, 0.4, and 0.2 respectively, and construct an evaluation model;

[0116] Extract new data from the standardized data set. For example, the device operation efficiency is 88%, the user activity is 65%, and the transaction completion rate is 92%. Input them into the evaluation model and calculate the weighted scores of device operation efficiency, user activity, and transaction completion rate to be 35.2, 26, and 18.4 respectively;

[0117] Classify the weighted scores of devices, users, and transactions. For example, use the K-means algorithm to classify the scores into high, medium, and low categories to obtain classification labels. For the classification labels, use principal component analysis technology to reduce the dimensions of the scores of device operation efficiency, user activity, and transaction completion rate. For example, reduce the three-dimensional data to two-dimensional to obtain the dimensionality reduction result;

[0118] According to the dimensionality reduction result, use visualization technology to generate the index distribution maps of devices, users, and transactions. For example, draw a scatter plot to obtain the distribution result;

[0119] According to the distribution results, association rule mining technology is used to mine the association relationship among the equipment operation efficiency, user activity, and transaction completion rate. For example, the association rule between the equipment operation efficiency and the transaction completion rate is mined to obtain the association rule set.

[0120] Further, the process of determining the influence degree of each indicator on the operation performance includes:

[0121] Causal inference technology is used to analyze the causal relationship between the equipment operation efficiency, user activity, transaction completion rate, and operation performance to obtain the causal strength value;

[0122] According to the causal strength value, combined with the comparative analysis technology, the evaluation results of equipment, users, and transactions are compared horizontally to obtain the horizontal difference value;

[0123] According to the horizontal difference value, combined with the comparative analysis technology, the evaluation results of equipment, users, and transactions are compared longitudinally to obtain the longitudinal difference value;

[0124] According to the horizontal difference value and the longitudinal difference value, the influence degree of the equipment operation efficiency, user activity, and transaction completion rate on the operation performance is judged to obtain the influence degree value;

[0125] Specifically, it can be described as follows. Using causal inference technology, through the causal analysis method based on the Bayesian network, the causal strength values between the equipment operation efficiency, user activity, transaction completion rate, and operation performance are calculated. For example, the causal strength value of the equipment operation efficiency is 0.75, the user activity is 0.68, and the transaction completion rate is 0.82;

[0126] According to the causal strength value, the T-test is used for horizontal comparative analysis to calculate the horizontal difference values of the evaluation results of equipment, users, and transactions. For example, the horizontal difference value of the equipment operation efficiency is 0.12, the user activity is 0.15, and the transaction completion rate is 0.10;

[0127] According to the horizontal difference value, the time series analysis method is used for longitudinal comparison to calculate the longitudinal difference values of the evaluation results of equipment, users, and transactions. For example, the longitudinal difference value of the equipment operation efficiency is 0.08, the user activity is 0.12, and the transaction completion rate is 0.06;

[0128] According to the horizontal difference value and the longitudinal difference value, the influence degree values of the equipment operation efficiency, user activity, and transaction completion rate on the operation performance are calculated by the weighted average method. For example, the influence degree value of the equipment operation efficiency is 0.40, the user activity is 0.35, and the transaction completion rate is 0.25.

[0129] Further, the process of dynamically presenting in a visual manner includes:

[0130] Using data visualization technology, the evaluation results and comparative analysis results are presented in the form of charts to generate a dynamic visualization report, and the intuitive performance of the evaluation results is determined;

[0131] According to the chart data in the visualization report, the eigenvalue of the key indicator is extracted to obtain an eigenvalue set;

[0132] According to the eigenvalue set, the key indicators are classified to obtain a classification result;

[0133] According to the classification result, the regression analysis method is used to predict the trend of the key indicator to obtain a prediction result;

[0134] Specifically, as described below, using data visualization technology, the evaluation results and comparative analysis results are presented in the form of charts to generate a dynamic visualization report, and the intuitive performance of the evaluation results is determined. For example, the D3.js library is used to present the device operation efficiency, user activity, and transaction completion rate in the form of line charts, bar charts, and pie charts, and dynamic interaction is supported;

[0135] According to the chart data in the visualization report, the eigenvalue of the key indicator is extracted to obtain an eigenvalue set. For example, the average value of the device operation efficiency, the peak value of the user activity, and the lowest value of the transaction completion rate are extracted from the line chart as eigenvalues;

[0136] The clustering algorithm is used to classify the key indicators in the eigenvalue set to obtain a classification result. For example, the K-means algorithm is used to classify the device, user, and transaction data into high, medium, and low categories, and the clustering centers are [0.8, 0.6, 0.9], [0.5, 0.4, 0.7], and [0.3, 0.2, 0.5] respectively;

[0137] According to the classification result, the regression analysis method is used to predict the trend of the key indicator to obtain a prediction result. For example, the linear regression model is used to predict that the device operation efficiency will increase by 5% in the next three months, the user activity will decrease by 3%, and the transaction completion rate will remain stable.

[0138] Through the time series analysis method, the evaluation results are monitored in real time, combined with the preset warning threshold to generate a dynamic warning signal, and according to the dynamic warning signal, the operation strategy and service quality are continuously optimized to obtain an optimization plan for the multi-agent collaboration efficiency;

[0139] Specifically, the time series analysis method is used to monitor the evaluation results in real time and collect time series data. According to the preset warning threshold, the time series data is compared to generate a dynamic warning signal. If the dynamic warning signal is triggered, then combined with the historical data of the operation strategy and service quality, the warning reason is analyzed, and according to the warning reason, the parameters of the operation strategy and service quality are adjusted to generate an optimization plan;

[0140] Through a multi-agent collaboration mechanism, apply the optimization plan to agents such as devices, users, and transactions to obtain collaborative efficiency data;

[0141] According to the collaborative efficiency data, use data visualization technology to generate a dynamic visualization report to display the optimization effect;

[0142] Extract the key index eigenvalue in the visualization report to form an eigenvalue set;

[0143] Use a clustering algorithm to classify the eigenvalue set to obtain a classification result;

[0144] According to the classification result, use the regression analysis method to predict the trend of the key index to obtain an optimized multi-agent collaborative efficiency plan;

[0145] Specifically, it can be described as follows. Use the time series analysis method to monitor the evaluation result in real time and collect time series data. For example, use the ARIMA model to predict and monitor the daily data of the device operation efficiency;

[0146] According to the preset warning threshold, compare the time series data to generate a dynamic warning signal. For example, trigger a warning when the user activity is lower than the threshold of 70%. If the dynamic warning signal is triggered, combine the historical data of the operation strategy and service quality to analyze the warning reason. For example, use the KPI analysis tool to find that the decrease in the transaction completion rate is due to the increase in the device failure rate;

[0147] According to the warning reason, adjust the parameters of the operation strategy and service quality to generate an optimization plan. For example, adjust the device maintenance frequency from once a week to once every three days;

[0148] Through a multi-agent collaboration mechanism, apply the optimization plan to agents such as devices, users, and transactions to obtain collaborative efficiency data. For example, compare the user retention rate before and after optimization through A / B testing;

[0149] According to the collaborative efficiency data, use data visualization technology to generate a dynamic visualization report to display the optimization effect. For example, use Tableau to draw a trend comparison chart of the device operation efficiency and user activity;

[0150] Extract the key index eigenvalue in the visualization report to form an eigenvalue set. For example, extract the mean and variance of the device operation efficiency from the chart;

[0151] Use a clustering algorithm to classify the eigenvalue set to obtain a classification result. For example, use the K-means algorithm to classify the device operation efficiency into three categories: high, medium, and low;

[0152] According to the classification results, a regression analysis method is used to predict the trends of key indicators, and an optimized multi-agent collaborative efficiency plan is obtained. For example, the growth trend of user activity in the next three months is predicted through linear regression.

[0153] Example 2

[0154] As Figure 2 shown, this embodiment provides a data collection and preprocessing module, an evaluation index and model management module, a data analysis and visualization module, a real-time monitoring and early warning module, and an operation optimization and decision support module;

[0155] The evaluation index and model management module is used to design and manage the evaluation index system and evaluation models for multi-agent operations. Specifically, according to the differences in agent roles and tasks, diverse evaluation indexes covering core tasks and key performance are designed; an evaluation model is constructed to quantify the correlation between core indexes and operation performance; the evaluation indexes are systematically processed through a preset weight allocation method and optimized according to the coverage range and performance evaluation;

[0156] The data analysis and visualization module extracts key information from the standardized data set and presents the analysis results in an intuitive way. Specifically, the association rule mining algorithm is used to extract core indexes; the evaluation results and comparative analysis are displayed in forms including dynamic charts and dashboards to enhance the interpretability of the results; trend prediction is carried out based on the historical data of key indexes.

[0157] The data collection and preprocessing module is responsible for collecting data from multiple data sources of the industrial Internet platform and performing preprocessing;

[0158] The real-time monitoring and early warning module conducts real-time monitoring of the operation process and generates dynamic early warning signals. Specifically, the evaluation results are monitored in real time, the time change trend of the data is analyzed, and dynamic early warning signals are generated in real time according to the preset threshold conditions to timely discover potential problems. Combining the real-time monitoring data, early warning of possible risk hazards is carried out to support the dynamic adjustment of operation strategies;

[0159] The operation optimization and decision support module dynamically adjusts operation strategies according to the early warning signals and evaluation results to improve operation efficiency, optimizes service quality in combination with the output of the evaluation model to meet the needs of different agents, and provides a decision support tool based on data analysis.

[0160] The multi-agent operation management system based on the industrial Internet platform realizes the accurate evaluation and collaborative optimization of the operation effects of different agents in the platform through a comprehensive evaluation index system, efficient data processing and analysis capabilities, real-time monitoring and dynamic warning mechanisms, intuitive visualization, and operation optimization and decision-making support functions. It can effectively improve the operation efficiency, service quality, and competitiveness of the platform, reduce operation costs, enhance the sustainable development ability of the platform, and provide strong support for the efficient management and innovative development of the industrial Internet platform.

Claims

1. A multi-agent operation management method based on an industrial Internet platform, characterized in that, It includes the following steps: Design a diversified evaluation index system covering the core tasks and key performances of different entities according to the role differences and task differences of each entity in the industrial Internet platform; Adopt distributed data acquisition technology to obtain device operation data, user behavior data, and transaction record data from multiple data sources of the platform, and obtain a multi-source heterogeneous data set; Remove noise and fuse the multi-source heterogeneous data set to obtain a standardized data set; Extract core indicators including device operation efficiency, user activity, and transaction completion rate from the standardized data set to obtain a key information set; Construct an evaluation model based on the key information set to quantitatively associate the core indicators of different entities with operation performance; Use causal inference and comparative analysis to judge the influence degree of each indicator on operation performance; Dynamically present the evaluation results and comparative analysis results in a visual manner; Conduct real-time monitoring of the evaluation results through time series analysis, generate dynamic warning signals in combination with warning thresholds, continuously optimize operation strategies and service quality, and obtain an optimization plan for multi-entity collaboration efficiency.

2. The multi-agent operation management method based on an industrial Internet platform according to claim 1, characterized in that, The process of designing the diversified evaluation index system includes: Adopt a preset classification method to divide the role differences and task differences of entity classes in the industrial Internet platform, and obtain the characteristic data of entity classes; Design diversified evaluation indicators covering core tasks and key performances according to the characteristic data of entity classes to form an evaluation index framework; Through a preset weight allocation method, conduct systematic processing on the evaluation indicators to determine the weight values of each indicator; Optimize the evaluation index framework in combination with the coverage range and performance evaluation to obtain the final evaluation index system.

3. The multi-agent operation management method based on an industrial Internet platform according to claim 1, characterized in that The process of obtaining the multi-source heterogeneous data set includes: Adopt distributed data acquisition technology to obtain device operation data, user behavior data, and transaction record data from the platform to obtain a multi-source heterogeneous data set; For the multi-source heterogeneous data set, remove noise data and redundant data to obtain a cleaned data set.

4. The multi-agent operation management method based on an industrial Internet platform according to claim 1, characterized in that The process of obtaining the standardized data set includes: According to the cleaned data set, adopt data fusion technology to unify the data format and structure to obtain a standardized data set; It also includes a data feature extraction process. Specifically, according to the cleaned data set, extract key features from operation data, behavior data, and record data to form a feature matrix.

5. The multi-agent operation management method based on an industrial Internet platform according to claim 1, wherein, The process of obtaining the key information set includes: Adopt an association rule mining algorithm to extract core indicators including device operation efficiency, user activity, and transaction record completion rate from the standardized data set to obtain a key information set; It also includes, according to the key information set, using preset threshold conditions to screen device operation efficiency, user activity, and transaction completion rate to obtain a screened information set; For the screened information set, classify devices, users, and transactions to obtain a classification result set.

6. The multi-agent operation management method based on an industrial Internet platform according to claim 1, characterized in that The process of constructing an evaluation model based on the key information set to quantitatively associate the core indicators of different entities with operation performance includes: Obtain device operation efficiency, user activity, and transaction completion rate data from the key information set as input variables of the multiple linear regression algorithm, and operation performance as the output variable; The multiple linear regression algorithm is used to fit the input variables and output variables, and the contribution coefficients of equipment operation efficiency, user activity, and transaction completion rate to operation performance are determined according to the fitting results to construct an evaluation model; Extract new data from the standardized dataset, input it into the evaluation model, and calculate the weighted scores of equipment operation efficiency, user activity, and transaction completion rate.

7. The multi-agent operation management method based on an industrial Internet platform according to claim 1, characterized in that The process of judging the influence degree of each index on operation performance includes: Use causal inference technology to analyze the causal relationship between equipment operation efficiency, user activity, and transaction completion rate and operation performance, and obtain the causal intensity value; According to the causal intensity value, combine the comparative analysis technology to conduct a horizontal comparison of the evaluation results of equipment, users, and transactions, and obtain the horizontal difference value; According to the horizontal difference value, combine the comparative analysis technology to conduct a vertical comparison of the evaluation results of equipment, users, and transactions, and obtain the vertical difference value; According to the horizontal difference value and the vertical difference value, judge the influence degree of equipment operation efficiency, user activity, and transaction completion rate on operation performance, and obtain the influence degree value.

8. The multi-agent operation management method based on an industrial Internet platform according to claim 1, wherein The process of dynamically presenting in a visual manner includes: Use data visualization technology to display the evaluation results and comparative analysis results in the form of charts, generate a dynamic visualization report, and determine the intuitive performance of the evaluation results; Extract the characteristic values of key indicators according to the chart data in the visualization report to obtain a set of characteristic values; Classify the key indicators according to the set of characteristic values to obtain the classification result; According to the classification result, use the regression analysis method to predict the trend of key indicators and obtain the prediction result.

9. A multi-agent operation management system based on an industrial Internet platform, characterized in that: It includes a data collection and preprocessing module, an evaluation index and model management module, a data analysis and visualization module, a real-time monitoring and early warning module, and an operation optimization and decision support module; The evaluation index and model management module is used to design and manage the evaluation index system and evaluation model for multi-agent operation. Specifically, according to the differences in agent roles and tasks, design diversified evaluation indexes covering core tasks and key performance; construct an evaluation model to quantify the correlation between core indexes and operation performance; Conduct a systematic processing of the evaluation indexes through a preset weight allocation method, and optimize according to the coverage range and performance evaluation; The data analysis and visualization module extracts key information from the standardized dataset and presents the analysis results in an intuitive manner. Specifically, use the association rule mining algorithm to extract core indexes; display the evaluation results and comparative analysis in the form of dynamic charts and dashboards to enhance the interpretability of the results; Conduct trend prediction based on the historical data of key indicators.

10. The multi-agent operation management system based on the industrial Internet platform according to claim 9, characterized in that: The data collection and preprocessing module is responsible for collecting data from multiple data sources of the industrial Internet platform and performing preprocessing; The real-time monitoring and early warning module monitors the operation process in real time and generates dynamic early warning signals. Specifically, it monitors the evaluation results in real time, analyzes the time change trend of the data, and generates dynamic early warning signals in real time according to the preset threshold conditions to timely discover potential problems. Combining the real-time monitoring data, it conducts early warning of possible risk hazards and supports the dynamic adjustment of operation strategies. The operation optimization and decision support module dynamically adjusts the operation strategies according to the early warning signals and evaluation results to improve the operation efficiency. Combining the output of the evaluation model, it optimizes the service quality, meets the needs of different entities, and provides decision support tools based on data analysis.

Images