A management method and system for improving enterprise efficiency based on artificial intelligence

Abstract

The present invention provides a management method and system for improving enterprise efficiency based on artificial intelligence, which belongs to the field of enterprise management technology. The present invention can obtain multiple clusters by clustering multiple nodes of a target enterprise, and generate data sampling strategy packages for each cluster according to the business attribute characteristics of each cluster and the influence indicators corresponding to each node, and use the sample data set obtained by sampling the central node of each cluster according to the data sampling strategy package to train a target efficiency optimization model, and then use the target efficiency optimization model to optimize the initial management strategy of each cluster output by the strategy generation model to obtain the target management strategy of each cluster, and manage each node in the cluster based on the target management strategy. In this way, the initial management strategy is optimized in advance using the target efficiency optimization model of the central node of each cluster, which can effectively and timely improve the efficiency of the enterprise.

Description

Technical Field

[0001] The present invention relates to the field of enterprise management technology, and in particular to a management method and system for improving enterprise efficiency based on artificial intelligence. Background Art

[0002] Enterprise management refers to the process of planning, organizing, coordinating, controlling and supervising all aspects of an enterprise. As the scale of an enterprise and its business continue to expand and become more complex, enterprises usually establish multiple subsidiaries or multiple regional management departments to operate different business segments as needed. The increase in the number of subsidiaries or regional management departments can easily lead to problems such as an increase in the management span of the corporate headquarters, an increase in management costs, and a decrease in decision-making efficiency, thereby reducing corporate effectiveness.

[0003] At present, when traditional enterprises formulate corporate management strategies for each subsidiary, it is usually done by the management of the corporate headquarters on a whim, and the corporate effectiveness can only be statistically obtained after the corporate management strategies of each subsidiary are actually implemented. This method cannot formulate efficient and reasonable corporate management strategies for each subsidiary in a timely manner. Summary of the invention

[0004] The embodiment of the present invention provides a management method for improving enterprise efficiency based on artificial intelligence. The present invention adopts the following technical solutions:

[0005] In a first aspect, the present invention provides a management method for improving enterprise efficiency based on artificial intelligence, which is applied to a central server, and the method comprises:

[0006] Acquire business data of multiple nodes of the target enterprise, and cluster the multiple nodes based on the business data of the multiple nodes to obtain multiple clusters; different clusters correspond to different business attribute characteristics;

[0007] For any cluster, based on the transaction data of each node in the cluster and a preset indicator recognition model, the influence index of each node output by the indicator recognition model is obtained, and the node with the highest influence index is determined as the central node of the cluster;

[0008] Based on the business attribute characteristics of each cluster and the influence index corresponding to each node in each cluster, a data sampling strategy package for each cluster is generated; the data sampling strategy package includes a data sampling strategy for each node in the cluster;

[0009] For any cluster, the data sampling strategy package and the initial performance optimization model of the cluster are sent to the central node of the cluster, so that the central node collects the historical performance-related data of each node in the cluster according to the data sampling strategy package to obtain a sample data set; and the initial performance optimization model is trained based on the sample data set to obtain a target performance optimization model corresponding to the cluster; wherein different clusters correspond to different target performance optimization models;

[0010] Inputting the business demand information and resource information uploaded by each node of each cluster into a strategy generation model, and outputting an initial management strategy for each cluster;

[0011] For any cluster, the initial management strategy of the cluster is sent to the central node of the cluster, so that the central node optimizes the initial management strategy based on the target performance optimization model, obtains the target management strategy corresponding to the cluster, and manages each of the nodes in the cluster based on the target management strategy.

[0012] In one embodiment of the present invention, based on the service data of the plurality of nodes, the plurality of nodes are clustered to obtain a plurality of clusters, including:

[0013] Preprocessing the business data of the plurality of nodes to obtain characteristic data corresponding to each of the plurality of nodes;

[0014] A clustering operation is performed on the feature data corresponding to each of the plurality of nodes, and the plurality of nodes are divided into a plurality of clusters.

[0015] In one embodiment of the present invention, preprocessing the service data of the plurality of nodes to obtain feature data corresponding to each of the plurality of nodes includes:

[0016] Performing data cleaning and standardization processing on the business data of the plurality of nodes to obtain initial feature data of the plurality of nodes;

[0017] For any node, extract the associated feature data in the initial feature data to obtain at least one associated data group, and perform feature fusion on the associated feature data in each associated data group to obtain fused feature data corresponding to each associated data group;

[0018] Based on the fused feature data and the independent feature data, feature data corresponding to the node is obtained; wherein the independent feature data is other feature data in the initial feature data except the associated feature data.

[0019] In one embodiment of the present invention, the business data includes business model data and customer attribute data;

[0020] Performing a clustering operation on the feature data corresponding to each of the plurality of nodes to divide the plurality of nodes into a plurality of clusters includes:

[0021] Based on the business category in the business model data and the customer category in the customer attribute data, a plurality of business attribute features are obtained, and the number of the plurality of business attribute features is determined as the number of clusters of the target clustering algorithm;

[0022] The target clustering algorithm is used to perform a clustering operation on the feature data according to the number of clusters, so as to divide the plurality of nodes into a plurality of clusters whose number is the same as the number of clusters.

[0023] In one embodiment of the present invention, the indicator recognition model is constructed based on a graph convolutional neural network; the indicator recognition model is used to perform the following steps:

[0024] For any cluster, each node in the cluster is mapped to a graph vertex, and based on the transaction data of each node, the edges between the graph vertices are constructed to obtain the graph structure data corresponding to the cluster;

[0025] Determine the centrality index, clustering coefficient and importance index of each graph vertex in the graph structure data;

[0026] For any graph vertex, determine the influence index of the graph vertex based on the centrality index, the clustering coefficient and the importance index of the graph vertex;

[0027] Based on the influence index of each of the graph vertices and the mapping relationship between each of the graph vertices and each of the nodes, the influence index of each of the nodes is determined.

[0028] In one embodiment of the present invention, for any graph vertex, determining the influence index of the graph vertex based on the centrality index, the clustering coefficient and the importance index of the graph vertex includes:

[0029] For any graph vertex, obtaining a first weight of the graph vertex for the centrality index, a second weight for the clustering coefficient, and a third weight for the importance index;

[0030] The influence index of the graph vertex is determined according to the sum of a first product of the centrality index and the first weight, a second product of the clustering coefficient and the second weight, and a third product of the importance index and the third weight.

[0031] In one embodiment of the present invention, based on the business attribute characteristics of each cluster and the influence index corresponding to each node in each cluster, a data sampling strategy package for each cluster is generated, including:

[0032] Based on the business attribute characteristics of each of the clusters, determining the data to be sampled and the initial sampling interval of each of the nodes in each of the clusters;

[0033] For any cluster, based on the influence index corresponding to each of the nodes in the cluster, the initial sampling interval corresponding to the cluster is corrected to obtain the target sampling interval corresponding to each of the nodes; wherein the influence index is negatively correlated with the target sampling interval;

[0034] Based on the to-be-sampled data and the target sampling interval of each of the nodes in each of the clusters, a data sampling strategy for each of the nodes in each of the clusters is determined to obtain the data sampling strategy package.

[0035] In one embodiment of the present invention, according to the data sampling strategy package, historical performance-related data of each node in the cluster is collected to obtain a sample data set, including:

[0036] For any node, based on the data sampling strategy corresponding to the node, multiple historical performance-related data of the node are collected; the historical performance-related data includes historical management strategies and historical performance data;

[0037] For any historical performance-related data, a labeling request for the historical performance-related data is sent to the node, and an optimization strategy label is obtained based on the optimization strategy labeling information fed back by the node, and a performance label is obtained based on the historical performance data;

[0038] Based on the management strategy, the performance label and the optimization strategy label, sample data corresponding to the historical performance association data is generated.

[0039] In one embodiment of the present invention, the initial management strategy is optimized based on the target effectiveness optimization model to obtain the target management strategy corresponding to the cluster, including:

[0040] Inputting the initial management strategy into the target performance optimization model, and outputting the predicted performance and the target optimization strategy for at least one item to be optimized in the initial management strategy;

[0041] Based on the target optimization strategy, at least one item to be optimized in the initial management strategy is optimized to obtain the target management strategy corresponding to the cluster.

[0042] In a second aspect, the present invention provides a management system for improving enterprise efficiency based on artificial intelligence, the system comprising:

[0043] A node clustering module is used to obtain business data of multiple nodes of a target enterprise, and cluster the multiple nodes based on the business data of the multiple nodes to obtain multiple clusters; different clusters correspond to different business attribute characteristics;

[0044] An indicator determination module is used to obtain, for any cluster, the influence indicators of each node output by the indicator identification model based on the transaction data of each node in the cluster and a preset indicator identification model, and determine the node with the highest influence indicator as the central node of the cluster;

[0045] A sampling strategy generation module, used to generate a data sampling strategy package for each cluster based on the business attribute characteristics of each cluster and the influence index corresponding to each node in each cluster; the data sampling strategy package includes a data sampling strategy for each node in the cluster;

[0046] A model training module is used to send the data sampling strategy package and the initial performance optimization model of any cluster cluster to the central node of the cluster cluster, so that the central node collects the historical performance-related data of each node in the cluster cluster according to the data sampling strategy package to obtain a sample data set; and train the initial performance optimization model based on the sample data set to obtain a target performance optimization model corresponding to the cluster cluster; wherein different cluster clusters correspond to different target performance optimization models;

[0047] A management strategy generation module, used to input the business demand information and resource information uploaded by each node of each cluster into a strategy generation model, and output an initial management strategy for each cluster;

[0048] The management strategy optimization module is used to send the initial management strategy of any cluster cluster to the central node of the cluster cluster, so that the central node optimizes the initial management strategy based on the target performance optimization model, obtains the target management strategy corresponding to the cluster cluster, and manages each of the nodes in the cluster cluster based on the target management strategy.

[0049] In one embodiment of the present invention, the node clustering module includes:

[0050] A preprocessing submodule, used to preprocess the business data of the plurality of nodes to obtain characteristic data corresponding to each of the plurality of nodes;

[0051] The node clustering submodule is used to perform clustering operations on the characteristic data corresponding to each of the multiple nodes, and divide the multiple nodes into multiple clusters.

[0052] In one embodiment of the present invention, the preprocessing submodule includes:

[0053] A standardization processing unit, used for performing data cleaning and standardization processing on the business data of the plurality of nodes to obtain initial feature data of the plurality of nodes;

[0054] A feature fusion unit, configured to extract, for any node, associated feature data from the initial feature data to obtain at least one associated data group, and perform feature fusion on the associated feature data in each associated data group to obtain fused feature data corresponding to each associated data group;

[0055] A feature acquisition unit is used to obtain feature data corresponding to the node based on the fused feature data and the independent feature data; wherein the independent feature data is other feature data in the initial feature data except the associated feature data.

[0056] In one embodiment of the present invention, the node clustering submodule includes:

[0057] A cluster number determination unit, configured to obtain a plurality of business attribute features based on the business category in the business model data and the customer category in the customer attribute data, and determine the number of the plurality of business attribute features as the number of clusters of the target clustering algorithm;

[0058] A node clustering unit is used to perform a clustering operation on the feature data according to the number of clusters by using the target clustering algorithm, so as to divide the plurality of nodes into a plurality of cluster clusters whose number is the same as the number of clusters.

[0059] In one embodiment of the present invention, the indicator recognition model is constructed based on a graph convolutional neural network; the indicator determination module includes:

[0060] A node mapping submodule is used to map each node in any cluster to a graph vertex, and to construct edges between each of the graph vertices based on the transaction data of each node to obtain graph structure data corresponding to the cluster;

[0061] A first index determination submodule, used to determine the centrality index, clustering coefficient and importance index of each graph vertex in the graph structure data;

[0062] A second indicator determination submodule, for determining, for any graph vertex, an influence indicator of the graph vertex based on the centrality indicator, the clustering coefficient and the importance indicator of the graph vertex;

[0063] The third indicator determination submodule is used to determine the influence indicator of each of the nodes based on the influence indicator of each of the graph vertices and the mapping relationship between each of the graph vertices and each of the nodes.

[0064] In one embodiment of the present invention, the second indicator determination submodule includes:

[0065] A weight acquisition unit, used for acquiring, for any graph vertex, a first weight of the graph vertex for the centrality index, a second weight for the clustering coefficient, and a third weight for the importance index;

[0066] An influence index determination unit is used to determine the influence index of the graph vertex based on the sum of a first product of the centrality index and the first weight, a second product of the clustering coefficient and the second weight, and a third product of the importance index and the third weight.

[0067] In one embodiment of the present invention, the sampling strategy generation module includes:

[0068] A sampling parameter determination submodule, used to determine the data to be sampled and the initial sampling interval of each of the nodes in each of the clusters based on the business attribute characteristics of each of the clusters;

[0069] A sampling interval correction submodule is used to correct the initial sampling interval corresponding to any cluster cluster based on the influence index corresponding to each node in the cluster cluster to obtain the target sampling interval corresponding to each node; wherein the influence index is negatively correlated with the target sampling interval;

[0070] The sampling strategy determination submodule is used to determine the data sampling strategy of each of the nodes in each of the clusters based on the to-be-sampled data and the target sampling interval of each of the nodes in each of the clusters to obtain the data sampling strategy package.

[0071] In one embodiment of the present invention, the model training module includes:

[0072] A data sampling submodule, for collecting, for any node, a plurality of historical performance-related data of the node based on a data sampling strategy corresponding to the node; the historical performance-related data includes a historical management strategy and historical performance data;

[0073] A label determination submodule, for sending a labeling request for any historical performance-related data to the node, and obtaining an optimization strategy label based on the optimization strategy labeling information fed back by the node, and obtaining a performance label based on the historical performance data;

[0074] The sample data generation submodule is used to generate sample data corresponding to the historical performance association data based on the management strategy, the performance label and the optimization strategy label.

[0075] In one embodiment of the present invention, the management strategy optimization module includes:

[0076] An output submodule, used for inputting the initial management strategy into the target performance optimization model, and outputting the predicted performance and the target optimization strategy for at least one item to be optimized in the initial management strategy;

[0077] The optimization submodule is used to optimize at least one item to be optimized in the initial management strategy based on the target optimization strategy to obtain the target management strategy corresponding to the cluster.

[0078] In summary, the above-mentioned management method and system for improving enterprise efficiency based on artificial intelligence have the following technical effects:

[0079] An embodiment of the present invention provides a management method for improving enterprise efficiency based on artificial intelligence. First, multiple nodes of a target enterprise are clustered to obtain multiple cluster clusters with different business attribute characteristics. Then, the influence indicator of each node is identified by using an indicator recognition model to determine the central node of each cluster cluster. Based on the business attribute characteristics of each cluster cluster and the influence indicator corresponding to each node in each cluster cluster, a data sampling strategy package for each cluster cluster is generated, and the data sampling strategy package and the initial efficiency optimization model of the cluster cluster are sent to the central node of the cluster cluster. Subsequently, each central node is used to collect historical efficiency-related data of each node in the cluster cluster according to the data sampling strategy package, and a sample data set is obtained. The initial efficiency optimization model is trained based on the sample data set to obtain a target efficiency optimization model corresponding to the cluster cluster. Using the target efficiency optimization model, the initial management strategy of each cluster cluster output by the strategy generation model can be optimized to obtain a target management strategy corresponding to the cluster cluster, and each node in the cluster cluster is managed based on the target management strategy. In this way, by specifically configuring the target performance optimization model for the central node of each cluster, the central node can optimize the performance of the initial management strategy output by the strategy generation model in advance, thereby timely and comprehensively improving the enterprise management efficiency and overall effectiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 A flowchart of a management method for improving enterprise efficiency based on artificial intelligence provided by an embodiment of the present invention.

[0081] Figure 2A schematic diagram of functional modules of a management system for improving enterprise efficiency based on artificial intelligence provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0082] The technical solution of the present invention will be described below in conjunction with the accompanying drawings.

[0083] It should be noted that currently, medium and large enterprises usually establish multiple subsidiaries to operate different business segments. As the number of subsidiaries increases, the management span of the corporate headquarters also increases, which may lead to a larger management span and more management affairs; at the same time, the increase in subsidiaries not only brings an increase in the number of management personnel, but also involves an increase in management costs such as legal affairs, compliance, control, human resources, and finance; too many management levels may lead to coordination difficulties and low management efficiency. Therefore, as the scale of enterprises and their businesses continue to expand and become more complex, it will bring a series of challenges to the efficiency management of enterprises.

[0084] In traditional performance management, it is usually necessary to evaluate the actual performance data using relevant performance indicators after the enterprise management strategies of each subsidiary are actually implemented. If the enterprise performance is poor, the enterprise management strategy is passively optimized. This method is post-optimization, and it is impossible to formulate efficient and reasonable enterprise management strategies for each subsidiary in a timely manner.

[0085] Based on this, the present invention aims to provide a management method for improving enterprise efficiency based on artificial intelligence. First, multiple nodes of the target enterprise are clustered to obtain multiple clusters with different business attribute characteristics; then, based on the business attribute characteristics of each cluster and the influence indicators corresponding to each node in each cluster, a data sampling strategy package for each cluster is generated; then, each central node is used to collect the historical efficiency-related data of each node in the cluster according to the data sampling strategy package, and a sample data set is obtained, and the initial efficiency optimization model is trained based on the sample data set to obtain a target efficiency optimization model corresponding to the cluster; using the target efficiency optimization model, the initial management strategy of each cluster output by the strategy generation model can be optimized to obtain the target management strategy corresponding to the cluster, and each node in the cluster is managed based on the target management strategy. In this way, by configuring the target efficiency optimization model for the central node of each cluster, the central node can optimize the efficiency of the initial management strategy in advance, thereby effectively and comprehensively improving the management efficiency and overall efficiency of the enterprise.

[0086] Reference Figure 1 The embodiment of the present invention provides a management method for improving enterprise efficiency based on artificial intelligence, which is applied to a central server. The method may include the following steps:

[0087] S101: Obtain business data of multiple nodes of a target enterprise, and cluster the multiple nodes based on the business data of the multiple nodes to obtain multiple clusters.

[0088] It should be noted that the central server is a server configured at the headquarters of the target enterprise, and the central server is communicated with multiple edge servers configured at multiple nodes, wherein the multiple nodes correspond one-to-one to the multiple edge servers; the node of the target enterprise represents the business unit of the target enterprise, for example, the node can be a subsidiary or management department of the target enterprise in charge of a certain business, and one or more of the multiple nodes can carry out different or the same businesses.

[0089] In this embodiment, the central server can obtain the business attribute characteristics of each node by analyzing the business data of multiple nodes, and can automatically classify the nodes with the same business attribute characteristics into the same cluster by clustering the multiple nodes. The business data may include but is not limited to financial data, market data, business model data, and customer attribute data.

[0090] In this embodiment, different clusters correspond to different business attribute characteristics, and each node under the same cluster has the same business attribute characteristics. Each node under the same cluster can carry out the same business respectively, or carry out a business together as a supply chain.

[0091] In this embodiment, by automatically grouping nodes with the same business attribute characteristics into the same cluster, different management strategies and goals can be formulated for nodes in different clusters, thereby performing more targeted management and improving management efficiency.

[0092] S102: For any cluster, based on the transaction data of each node in the cluster and a preset indicator recognition model, obtain the influence index of each node output by the indicator recognition model, and determine the node with the highest influence index as the central node of the cluster.

[0093] In this embodiment, considering that there is usually a central node with the greatest influence among multiple nodes in the same cluster, the central node may be the node in the cluster that is most important to the target enterprise, has the greatest contribution, or has the highest competitiveness. For example, the central node may be a node configured by the target enterprise to manage other nodes in the cluster, or may be the node with the largest scale, the largest market share, or the highest enterprise level among all nodes.

[0094] In this embodiment, the indicator identification model can comprehensively evaluate the influence of nodes from multiple dimensions based on the transaction data of each node, thereby obtaining the influence index of each node, and determining the node with the highest influence index as the central node of the cluster. The transaction data may include but is not limited to financial data, management data, market performance data, operational data, and human resources data.

[0095] In this implementation, by identifying the central node of each cluster and then utilizing the central node, unified energy efficiency management of each node in the cluster can be achieved.

[0096] S103: Generate a data sampling strategy package for each cluster based on the business attribute characteristics of each cluster and the influence index corresponding to each node in each cluster.

[0097] It should be noted that the data sampling strategy package includes a data sampling strategy for each node in the cluster.

[0098] In this embodiment, the central server can formulate corresponding data sampling strategies for each node in the cluster by comprehensively considering the business attribute characteristics of the cluster and the influence index corresponding to each node in each cluster, thereby effectively ensuring the comprehensiveness and accuracy of data when collecting data for important nodes. Among them, the influence index of the node represents the degree of influence of the node on the cluster, that is, the higher the influence index of a node, the greater the influence of the node on the business corresponding to the cluster.

[0099] In a specific implementation, S103 may include the following sub-steps:

[0100] S103 - 1 : Based on the service attribute characteristics of each cluster, determine the data to be sampled and the initial sampling interval of each node in each cluster.

[0101] In this embodiment, the central server presets different benchmark sampling strategies for clusters with different business attribute characteristics, and the benchmark sampling strategies include the data to be sampled and the initial sampling interval of each node in the cluster. In other words, different business attribute characteristics correspond to different data to be sampled and different initial sampling intervals.

[0102] It should be noted that the more important the business corresponding to the business attribute characteristics is to the target enterprise, the more data types there are to be sampled, and the shorter the initial sampling interval is. The shorter the initial sampling interval is, the more data can be collected in the same time.

[0103] S103-2: For any cluster, based on the influence index corresponding to each node in the cluster, the initial sampling interval corresponding to the cluster is corrected to obtain the target sampling interval corresponding to each node.

[0104] In this embodiment, the influence index is negatively correlated with the target sampling interval. That is, different influence indexes correspond to different correction amplitudes. The higher the influence index corresponding to a node, the greater the correction amplitude of the node, and the shorter the corresponding target sampling interval. In this way, more comprehensive and accurate historical performance-related data of the node can be collected.

[0105] Exemplarily, a cluster includes node A, node B and node C, among which, since node A has the largest business scale or is in a dominant position in a business chain, node A has the highest influence index, so the initial sampling interval of node A will be corrected according to the maximum correction amplitude to obtain the target sampling interval of node A.

[0106] S103 - 3 : Based on the to-be-sampled data of each node in each cluster and the target sampling interval, determine the data sampling strategy of each node in each cluster to obtain a data sampling strategy package.

[0107] It should be noted that the data sampling strategy of a node includes the data to be sampled and the target sampling interval corresponding to the node.

[0108] In this implementation, the central server integrates the data sampling strategies of the nodes in each cluster, so as to obtain the data sampling strategy package corresponding to each cluster.

[0109] In this embodiment, compared with the method of balanced sampling of each node in the cluster, by assigning different sampling weights to the nodes according to the influence indicators corresponding to the nodes, it can ensure that the performance benefits of the nodes with higher influence indicators are greater, thereby improving the overall performance benefits of the cluster.

[0110] S104: For any cluster, the data sampling strategy package and the initial performance optimization model of the cluster are sent to the central node of the cluster, so that the central node collects the historical performance-related data of each node in the cluster according to the data sampling strategy package to obtain a sample data set; and the initial performance optimization model is trained based on the sample data set to obtain a target performance optimization model corresponding to the cluster.

[0111] In this embodiment, since the target performance optimization models of different central nodes are trained based on the historical performance association data of nodes in different clusters, and different clusters correspond to different business attribute characteristics, different clusters correspond to different target performance optimization models. In this way, based on the target performance optimization model, the management strategy of the cluster can be predicted and optimized in a targeted manner.

[0112] It should be noted that the initial performance optimization model can be an untrained initial neural network model; or, in order to increase the model training speed, the initial neural network model can be preliminarily trained based on general training samples to obtain a general performance optimization model with lower prediction accuracy.

[0113] In the specific implementation, the central server can send the data sampling strategy package and the initial performance optimization model to the edge server of the central node of the cluster, so that the edge server can train the initial performance optimization model locally. In this way, the computing pressure of the central server can be effectively reduced, and the risk of data leakage during data transmission can be avoided, while improving the training efficiency of the initial performance optimization model.

[0114] In this embodiment, the edge server can generate sample data corresponding to the historical performance association data by collecting the historical performance association data of each node in the cluster, and then obtain a sample data set composed of a plurality of sample data.

[0115] In this embodiment, in order to enable the target performance optimization model to output the corresponding optimization strategy while predicting the performance, the step of collecting the historical performance association data of each node in the cluster according to the data sampling strategy package in S104 to obtain the sample data set may specifically include the following sub-steps:

[0116] S104-1: For any node, based on the data sampling strategy corresponding to the node, multiple historical performance-related data of the node are collected.

[0117] In a specific implementation, after receiving the data sampling strategy sent by the central server, the edge server configured at the central node will perform data sampling on the corresponding node based on the data to be sampled and the target sampling interval indicated by the data sampling strategy to obtain multiple historical performance-related data of the node. The historical performance-related data includes historical management strategies and historical performance data.

[0118] It should be noted that based on historical performance data, corresponding performance labels can be obtained, and then sample data with performance labels can be obtained. The target performance optimization model trained based on the sample data can output predicted performance based on the input management strategy.

[0119] S104-2: For any historical performance-related data, a labeling request for the historical performance-related data is sent to the node, and based on the optimization strategy labeling information fed back by the node, an optimization strategy label is obtained, and based on the historical performance data, a performance label is obtained.

[0120] In this embodiment, in order to enable the target performance optimization model to further output the corresponding optimization strategy based on the predicted performance, the edge server of the central node will send a labeling request for the historical performance-related data to the management terminal of each node, so that the corresponding management personnel can feedback the optimization strategy labeling information to the edge server through the management terminal. The edge server can obtain the optimization strategy label of the historical performance-related data based on the optimization strategy in the optimization strategy labeling information.

[0121] S104-3: Generate sample data corresponding to the historical performance-related data based on the management strategy, performance label, and optimization strategy label.

[0122] In this embodiment, after obtaining the management policy, performance label and optimization strategy label corresponding to any historical performance association data, the edge server of the central node can generate sample data corresponding to the historical performance association data, and then obtain a sample data set by integrating the sample data corresponding to multiple historical performance association data.

[0123] In a specific implementation, after the edge server of the central node obtains the sample data set, it can start the training process of the initial performance optimization model. Specifically, the target performance optimization model can be trained in the following ways: extract features from the sample data to obtain performance feature data of the sample data, input the performance feature data of the sample data into the initial performance optimization model for training, and obtain the predicted performance and optimization strategy output by the initial performance optimization model; then calculate the first loss function value based on the predicted performance and performance label, and calculate the second loss function value based on the optimization strategy and optimization strategy label; then iteratively update the network parameters of the initial performance optimization model based on the first loss function value and the second loss function value until the initial performance optimization model meets the training cutoff condition, and obtain the pre-trained target performance optimization model.

[0124] S105: Inputting the business demand information and resource information uploaded by each node of each cluster into the policy generation model, and outputting the initial management policy of each cluster.

[0125] It should be noted that the policy generation model is used to generate an initial management policy that meets business demand information, and the initial management policy may not be the management policy with the best performance. Therefore, after the central server generates the initial management policy for each cluster using the policy generation model, it will send the initial management policy to the central node of the corresponding cluster, so that the target performance optimization model of the central node can further optimize the performance of the initial management policy, and then obtain a target management policy that can meet business demand information and further improve performance.

[0126] It should be further explained that the business demand information of a node represents the business indicators that the node needs to achieve, for example, the operating profit needs to reach the target profit within a specified time, and the resource information represents the personnel resource information and equipment resource information possessed by the node.

[0127] In a specific implementation, the policy generation model can adopt a large language model. Specifically, the central server can input the business demand information, resource information and preset format constraint text uploaded by each node of each cluster into the policy generation model, and output the initial management policy of each cluster; wherein the initial management policy has the target format specified by the format constraint text. In order to enable the large language model to output a more accurate initial management policy, preset policy auxiliary information can also be input into the large language model, and the policy auxiliary information can include market environment data and pre-screened preferred policy templates, etc.

[0128] In the specific implementation, the policy generation model can use a pre-trained deep learning model. Specifically, a training data set containing enterprise business demand information, resource information and corresponding management policies can be first constructed, and the business demand information and resource information are used as the input of the initial deep learning model, and the management policy is used as the output of the initial deep learning model. The initial deep learning model is iteratively trained so that the trained deep learning model can learn the relationship between business demand information, resource information and management policy.

[0129] S106: For any cluster, the initial management strategy of the cluster is sent to the central node of the cluster, so that the central node optimizes the initial management strategy based on the target performance optimization model, obtains the target management strategy corresponding to the cluster, and manages each node in the cluster based on the target management strategy.

[0130] In this embodiment, after the central server sends the initial management strategy of each cluster to the central node of the cluster, the target performance optimization model trained in the edge server of the central node can be used to optimize the initial management strategy to obtain the optimized target management strategy. The target management strategy can specifically include management sub-strategies for each node in the cluster, and the management sub-strategy of the node is used to instruct the node to use the target resources to achieve the target demand according to the target strategy.

[0131] In the specific implementation, the edge server first inputs the initial management strategy into the target performance optimization model, and outputs the predicted performance and the target optimization strategy for at least one item to be optimized in the initial management strategy; then, based on the target optimization strategy, at least one item to be optimized in the initial management strategy is optimized to obtain the target management strategy corresponding to the clustering cluster.

[0132] In this embodiment, by cooperating with the policy generation model of the central server and the target performance optimization model of each central node, on the one hand, the policy generation model of the central server can automatically generate the initial management strategy of each cluster by comprehensively analyzing the business demand information and resource information of each node. The initial management strategy can meet the business needs of the node. On the other hand, the target performance optimization model of the central node can perform targeted performance optimization on the initial management strategy, and then configure reasonable target management strategies for each node of each cluster, further improving the management efficiency and overall performance of the enterprise.

[0133] In a feasible implementation manner, the step of clustering the multiple nodes based on the business data of the multiple nodes to obtain multiple clusters in S101 may specifically include the following sub-steps:

[0134] S101 - 1 : Preprocessing the business data of multiple nodes to obtain feature data corresponding to each of the multiple nodes.

[0135] In this implementation, by preprocessing the business data, the quality of the feature data can be effectively guaranteed, thereby improving the accuracy of clustering.

[0136] In some implementations, S101-1 may specifically include the following sub-steps:

[0137] S101-1-1: Perform data cleaning and standardization on the business data of multiple nodes to obtain initial feature data of the multiple nodes.

[0138] In the specific implementation, the missing data of the business data can be filled by means of mean filling, median filling, mode filling or model prediction filling to ensure the integrity of the business data; then the text data in the business data can be converted into numerical data by means of one-hot encoding, label encoding and the like; finally, the numerical data is standardized to convert the numerical data to the same range, for example, converting the data to [0, 1], thereby obtaining the initial feature data corresponding to each of the multiple nodes.

[0139] S101-1-2: for any node, extract associated feature data from the initial feature data to obtain at least one associated data group, and perform feature fusion on the associated feature data in each associated data group to obtain fused feature data corresponding to each associated data group;

[0140] In this embodiment, considering that there are many types of business data, the corresponding initial feature data has many dimensions. Therefore, in order to reduce the feature dimensions of the initial feature data and improve the clustering efficiency and clustering effect, the associated feature data in the initial feature data will be fused to reduce the number of features while retaining the feature information of the associated feature data. In a specific implementation, the fused feature data can be obtained by combining the product or ratio of the associated feature data.

[0141] For example, the "market share" and "customer satisfaction" in the business data can be used as a set of related feature data. By fusing the related feature data, the fused feature data representing the "customer satisfaction market index" can be obtained; or the "customer geographic location" and "customer satisfaction score" can be fused to obtain the fused feature data representing the "satisfaction score of the customer's region".

[0142] S101-1-3: Obtain feature data corresponding to the node based on the fused feature data and the independent feature data.

[0143] It should be noted that the independent feature data refers to other feature data in the initial feature data except the associated feature data.

[0144] In this embodiment, by merging the associated feature data of each node, it is possible to reduce the feature dimension while retaining important information.

[0145] S101 - 2 : performing a clustering operation on the feature data corresponding to each of the multiple nodes, and dividing the multiple nodes into multiple clusters.

[0146] In a specific implementation, a K-MEANS clustering algorithm or a Fuzzy C-Means Clustering (FCM) algorithm may be used to perform clustering operations on multiple nodes to obtain multiple clusters.

[0147] It should be noted that the K-MEANS clustering algorithm is a basic partitioning algorithm with a known number of clustering categories. It is a typical distance-based clustering algorithm that uses distance as an evaluation index of similarity, that is, it is believed that the closer the distance between two objects, the greater their similarity. The algorithm believes that clusters are composed of objects that are close in distance, so obtaining compact and independent clusters is the ultimate goal. The clustering result is K clusters divided into K categories. The fuzzy C-means clustering algorithm is a type of fuzzy clustering algorithm. It obtains the membership of each node's feature data to the center of the c class by optimizing the objective function, and determines the cluster to which the node belongs, so as to achieve the purpose of automatically classifying each node.

[0148] In this implementation, to improve the accuracy of the clustering algorithm, S101-2 may specifically include the following sub-steps:

[0149] S101-2-1: Based on the business categories in the business model data and the customer categories in the customer attribute data, a plurality of business attribute features are obtained, and the number of the plurality of business attribute features is determined as the number of clusters of the target clustering algorithm.

[0150] In this embodiment, since the K-MEANS clustering algorithm or the fuzzy C-means clustering algorithm requires the number of clusters to be specified in advance, in order to achieve a reasonable division of multiple nodes, the business category and customer category of the nodes will be comprehensively considered to determine the appropriate number of clusters.

[0151] In a specific implementation, the number of customer categories under each business category may be first counted, and then the number of customer categories under all business categories may be added up to obtain a corresponding number of business attribute features, and each business attribute feature will correspond to a cluster.

[0152] S101-2-2: performing a clustering operation on the feature data according to the number of clusters by using a target clustering algorithm, so as to divide the plurality of nodes into a plurality of clusters whose number is the same as the number of clusters.

[0153] Exemplarily, let the number of clusters be a. If the target clustering algorithm adopts the K-MEANS clustering algorithm, let the K value of the K-MEANS clustering algorithm be a; if the target clustering algorithm is the fuzzy C-means clustering algorithm, let the c value in the fuzzy C-means clustering algorithm be a.

[0154] In this embodiment, by determining a suitable number of clusters for the target clustering algorithm, the quality and interpretability of the clustering results can be effectively improved, thereby better dividing the nodes into meaningful clusters and avoiding mixing the nodes together incorrectly.

[0155] In a feasible implementation, the indicator recognition model is constructed based on a graph convolutional neural network; the indicator recognition model is used to perform the following steps:

[0156] S102-1: For any cluster, map each node in the cluster to a graph vertex, and construct edges between each graph vertex based on the transaction data of each node to obtain graph structure data corresponding to the cluster.

[0157] In this embodiment, considering that there are usually business dealings or communication collaborations between the nodes in the same cluster, the nodes can together form a business network, and a graph structure data can be constructed based on the correlation between the nodes. It should be noted that the graph structure is a nonlinear structure that is more complex than the tree structure. Among them, all vertices constitute a vertex set, all edges constitute an edge set, and a complete graph structure consists of a vertex set and an edge set. In other words, the graph vertices are the smallest units in the graph structure data, and the graph nodes in the graph structure data are not in independent and identical distribution, but are usually interrelated.

[0158] It should be noted that the Graph Convolutional Network (GCN) is a deep learning model specifically designed to process graph structured data. Unlike the traditional Convolutional Neural Network (CNN) used to process image data, GCN can process graph data of arbitrary structure. GCN can propagate information in the graph structure, thereby capturing the dependencies between nodes.

[0159] In this embodiment, by analyzing the transaction data of each node, the data dependency and / or communication relationship between each node can be effectively extracted, and then the edges between each graph vertex can be constructed. Specifically, weights or other attributes can be assigned to the edges to express the strength of data dependency or communication frequency between adjacent graph vertices. At the same time, attribute information such as node name, location, business type, etc. can also be assigned to the corresponding graph vertex based on the transaction data of each node. In this way, each node in the same cluster can be mapped to a complete graph structure data.

[0160] S102-2: Determine the centrality index, clustering coefficient and importance index of each graph vertex in the graph structure data.

[0161] In this embodiment, based on the characteristics of the graph structure data, multiple dimensional indicators including centrality index, clustering coefficient and importance index of each graph vertex can be quickly analyzed and obtained.

[0162] It should be noted that the centrality index is an index used to measure the importance and influence of graph vertices in the graph structure. In the specific implementation, the degree centrality, betweenness centrality and closeness centrality of the graph vertex can be comprehensively considered to obtain the centrality index of the graph vertex.

[0163] It should be noted that the clustering coefficient is an indicator that measures the degree of interconnection between the neighbors of a graph vertex. In the specific implementation, the node clustering coefficient (Node Clustering Coefficient) and the global clustering coefficient (Global Clustering Coefficient) of the graph vertex can be considered comprehensively to obtain the clustering coefficient of the graph vertex. Among them, the node clustering coefficient measures the number of ternary closures between a graph vertex and its neighbors. The more edges there are between the neighbors of a graph vertex, the higher the clustering coefficient of the graph vertex; the global clustering coefficient measures the proportion of ternary closures between vertex neighbors in the entire graph structure data.

[0164] It should be noted that the importance index is an index used to measure the importance of a graph vertex in a graph structure. In a specific implementation, the PageRank index of a graph vertex can be determined as the importance index of the graph vertex. The larger the PageRank index value of a graph vertex, the greater the influence of the graph vertex on the other graph nodes in the graph data, and the higher the importance of the graph node in the graph structure data.

[0165] S102-3: For any graph vertex, determine the influence index of the graph vertex based on the centrality index, clustering coefficient and importance index of the graph vertex.

[0166] In this embodiment, by comprehensively considering the centrality index, clustering coefficient and importance index of the graph vertices, accurate calculation of the influence index can be achieved.

[0167] In a specific implementation, S102-3 may specifically include the following sub-steps:

[0168] S102-3-1: For any graph vertex, obtain a first weight of the graph vertex for a centrality index, a second weight for a clustering coefficient, and a third weight for an importance index.

[0169] In the specific implementation, considering that different nodes may have different characteristics (such as scale, corporate administrative level, historical performance, etc.), characteristic labels can be set for different nodes, and different characteristic labels correspond to different weight combinations, which include a first weight for the centrality index, a second weight for the clustering coefficient, and a third weight for the importance index. In this way, for any graph vertex, the corresponding weight combination can be assigned to the graph vertex based on the characteristic label of the node corresponding to the graph vertex, so as to achieve accurate calculation of the influence index of the graph vertex.

[0170] It should be noted that the sum of the first weight, the second weight and the third weight in any weight combination is 1.

[0171] S102-3-2: Determine the influence index of the graph vertex according to the sum of the first product of the centrality index and the first weight, the second product of the clustering coefficient and the second weight, and the third product of the importance index and the third weight.

[0172] In this implementation, based on the matching weight combination, the influence index of the graph vertex can be solved by performing weighted summation on the centrality index, the clustering coefficient and the importance index.

[0173] S102-4: Determine the influence index of each node based on the influence index of each graph vertex and the mapping relationship between each graph vertex and each node.

[0174] In this embodiment, since there is a one-to-one mapping relationship between each graph vertex and each node, after the influence index of each graph vertex is calculated, the influence index of each node can be obtained.

[0175] In the second aspect, based on the same inventive concept, refer to Figure 2 , shows a management system 200 for improving enterprise efficiency based on artificial intelligence provided by an embodiment of the present invention, the system comprising:

[0176] The node clustering module 201 is used to obtain the business data of multiple nodes of the target enterprise, and cluster the multiple nodes based on the business data of the multiple nodes to obtain multiple clusters; different clusters correspond to different business attribute characteristics;

[0177] The indicator determination module 202 is used to obtain the influence indicator of each node output by the indicator identification model for any cluster based on the transaction data of each node in the cluster and the preset indicator identification model, and determine the node with the highest influence indicator as the central node of the cluster;

[0178] The sampling strategy generation module 203 is used to generate a data sampling strategy package for each cluster based on the business attribute characteristics of each cluster and the influence index corresponding to each node in each cluster; the data sampling strategy package includes a data sampling strategy for each node in the cluster;

[0179] The model training module 204 is used to send the data sampling strategy package and the initial performance optimization model of the cluster cluster to the central node of the cluster cluster for any cluster cluster, so that the central node collects the historical performance-related data of each node in the cluster cluster according to the data sampling strategy package to obtain a sample data set; and train the initial performance optimization model based on the sample data set to obtain a target performance optimization model corresponding to the cluster cluster; wherein different cluster clusters correspond to different target performance optimization models;

[0180] The management strategy generation module 205 is used to input the business demand information and resource information uploaded by each node of each cluster into the strategy generation model, and output the initial management strategy of each cluster;

[0181] The management strategy optimization module 206 is used to send the initial management strategy of any cluster cluster to the central node of the cluster cluster, so that the central node optimizes the initial management strategy based on the target performance optimization model, obtains the target management strategy corresponding to the cluster cluster, and manages each node in the cluster cluster based on the target management strategy.

[0182] In one embodiment of the present invention, the node clustering module 201 includes:

[0183] A preprocessing submodule is used to preprocess the business data of multiple nodes to obtain feature data corresponding to each of the multiple nodes;

[0184] The node clustering submodule is used to perform clustering operations on the feature data corresponding to multiple nodes and divide the multiple nodes into multiple clusters.

[0185] In one embodiment of the present invention, the preprocessing submodule includes:

[0186] A standardization processing unit, used for performing data cleaning and standardization processing on the business data of multiple nodes to obtain initial feature data of multiple nodes;

[0187] A feature fusion unit is used to extract the associated feature data in the initial feature data for any node to obtain at least one associated data group, and perform feature fusion on the associated feature data in each associated data group to obtain fused feature data corresponding to each associated data group;

[0188] The feature acquisition unit is used to obtain feature data corresponding to the node based on the fused feature data and the independent feature data; wherein the independent feature data is other feature data in the initial feature data except the associated feature data.

[0189] In one embodiment of the present invention, the node clustering submodule includes:

[0190] A cluster number determination unit, configured to obtain a plurality of business attribute features based on the business categories in the business model data and the customer categories in the customer attribute data, and determine the number of the plurality of business attribute features as the number of clusters of the target clustering algorithm;

[0191] The node clustering unit is used to perform clustering operations on the feature data according to the number of clusters through a target clustering algorithm, so as to divide multiple nodes into multiple clusters whose number is the same as the number of clusters.

[0192] In one embodiment of the present invention, the indicator recognition model is constructed based on a graph convolutional neural network; the indicator determination module 202 includes:

[0193] The node mapping submodule is used to map each node in any cluster to a graph vertex, and to construct edges between each graph vertex based on the transaction data of each node to obtain the graph structure data corresponding to the cluster;

[0194] The first indicator determination submodule is used to determine the centrality index, clustering coefficient and importance index of each graph vertex in the graph structure data;

[0195] The second indicator determination submodule is used to determine the influence indicator of any graph vertex based on the centrality indicator, clustering coefficient and importance indicator of the graph vertex;

[0196] The third indicator determination submodule is used to determine the influence indicator of each node based on the influence indicator of each graph vertex and the mapping relationship between each graph vertex and each node.

[0197] In one embodiment of the present invention, the second indicator determination submodule includes:

[0198] A weight acquisition unit, used for acquiring, for any graph vertex, a first weight for a centrality index, a second weight for a clustering coefficient, and a third weight for an importance index of the graph vertex;

[0199] The influence index determination unit is used to determine the influence index of the graph vertex according to the sum of a first product of a centrality index and a first weight, a second product of a clustering coefficient and a second weight, and a third product of an importance index and a third weight.

[0200] In one embodiment of the present invention, the sampling strategy generation module 203 includes:

[0201] The sampling parameter determination submodule is used to determine the data to be sampled and the initial sampling interval of each node in each cluster based on the business attribute characteristics of each cluster;

[0202] The sampling interval correction submodule is used to correct the initial sampling interval corresponding to any cluster cluster based on the influence index corresponding to each node in the cluster cluster to obtain the target sampling interval corresponding to each node; wherein the influence index is negatively correlated with the target sampling interval;

[0203] The sampling strategy determination submodule is used to determine the data sampling strategy of each node in each cluster based on the to-be-sampled data of each node in each cluster and the target sampling interval to obtain a data sampling strategy package.

[0204] In one embodiment of the present invention, the model training module 204 includes:

[0205] The data sampling submodule is used to collect multiple historical performance-related data of any node based on the data sampling strategy corresponding to the node; the historical performance-related data includes historical management strategies and historical performance data;

[0206] The label determination submodule is used to send a labeling request for the historical performance-related data to the node for any historical performance-related data, and obtain an optimization strategy label based on the optimization strategy labeling information fed back by the node, and obtain a performance label based on the historical performance data;

[0207] The sample data generation submodule is used to generate sample data corresponding to the historical performance association data based on the management strategy, performance label and optimization strategy label.

[0208] In one embodiment of the present invention, the management strategy optimization module 206 includes:

[0209] An output submodule, used to input the initial management strategy into the target performance optimization model, and output the predicted performance and the target optimization strategy for at least one item to be optimized in the initial management strategy;

[0210] The optimization submodule is used to optimize at least one item to be optimized in the initial management strategy based on the target optimization strategy to obtain the target management strategy corresponding to the cluster.

[0211] It should be noted that the specific implementation of the management system 200 for improving enterprise efficiency based on artificial intelligence in the embodiment of the present invention refers to the specific implementation of the management method for improving enterprise efficiency based on artificial intelligence proposed in the first aspect of the aforementioned embodiment of the present invention, and will not be repeated here.

[0212] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When a computer instruction or computer program is loaded or executed on a computer, a process or function according to an embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable system. The computer program or instruction can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer program or instruction can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD) or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0213] In the present invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0214] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0215] Those of ordinary skill in the art will appreciate that the submodules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0216] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and sub-modules can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0217] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic, for example, the division of submodules is only a logical function division, and there may be other division methods in actual implementation, such as multiple submodules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of systems or submodules, which can be electrical, mechanical or other forms.

[0218] The submodules described as separate components may or may not be physically separated, and the components shown as submodules may or may not be physical submodules, that is, they may be located in one place, or they may be distributed on multiple network submodules. Some or all of the submodules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0219] In addition, each functional submodule in each embodiment of the present invention may be integrated into one processing submodule, or each submodule may exist physically separately, or two or more submodules may be integrated into one submodule.

[0220] If the function is implemented in the form of a software function submodule and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0221] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A management method for improving enterprise efficiency based on artificial intelligence, characterized in that: Applied to the central server, the method comprises: Acquire business data of multiple nodes of the target enterprise, and cluster the multiple nodes based on the business data of the multiple nodes to obtain multiple clusters; different clusters correspond to different business attribute characteristics; For any cluster, based on the transaction data of each node in the cluster and a preset indicator recognition model, the influence index of each node output by the indicator recognition model is obtained, and the node with the highest influence index is determined as the central node of the cluster; Based on the business attribute characteristics of each cluster and the influence index corresponding to each node in each cluster, a data sampling strategy package for each cluster is generated; the data sampling strategy package includes a data sampling strategy for each node in the cluster; For any cluster, the data sampling strategy package and the initial performance optimization model of the cluster are sent to the central node of the cluster, so that the central node collects the historical performance-related data of each node in the cluster according to the data sampling strategy package to obtain a sample data set; and the initial performance optimization model is trained based on the sample data set to obtain a target performance optimization model corresponding to the cluster; wherein different clusters correspond to different target performance optimization models; Inputting the business demand information and resource information uploaded by each node of each cluster into a strategy generation model, and outputting an initial management strategy for each cluster; For any cluster, the initial management strategy of the cluster is sent to the central node of the cluster, so that the central node optimizes the initial management strategy based on the target performance optimization model, obtains the target management strategy corresponding to the cluster, and manages each of the nodes in the cluster based on the target management strategy.

2. According to claim 1, a management method for improving enterprise efficiency based on artificial intelligence is characterized in that: Based on the business data of the multiple nodes, the multiple nodes are clustered to obtain multiple clusters, including: Preprocessing the business data of the plurality of nodes to obtain characteristic data corresponding to each of the plurality of nodes; A clustering operation is performed on the feature data corresponding to each of the plurality of nodes, and the plurality of nodes are divided into a plurality of clusters.

3. A management method for improving enterprise efficiency based on artificial intelligence according to claim 2, characterized in that: Preprocessing the service data of the plurality of nodes to obtain feature data corresponding to each of the plurality of nodes includes: Performing data cleaning and standardization processing on the business data of the plurality of nodes to obtain initial feature data of the plurality of nodes; For any node, extract the associated feature data in the initial feature data to obtain at least one associated data group, and perform feature fusion on the associated feature data in each associated data group to obtain fused feature data corresponding to each associated data group; Based on the fused feature data and the independent feature data, feature data corresponding to the node is obtained; wherein the independent feature data is other feature data in the initial feature data except the associated feature data.

4. A management method for improving enterprise efficiency based on artificial intelligence according to claim 2 or 3, characterized in that: The business data includes business model data and customer attribute data; Performing a clustering operation on the feature data corresponding to each of the plurality of nodes to divide the plurality of nodes into a plurality of clusters includes: Based on the business category in the business model data and the customer category in the customer attribute data, a plurality of business attribute features are obtained, and the number of the plurality of business attribute features is determined as the number of clusters of the target clustering algorithm; The target clustering algorithm is used to perform a clustering operation on the feature data according to the number of clusters, so as to divide the plurality of nodes into a plurality of clusters whose number is the same as the number of clusters.

5. The management method for improving enterprise efficiency based on artificial intelligence according to claim 1, characterized in that: The indicator recognition model is constructed based on a graph convolutional neural network; the indicator recognition model is used to perform the following steps: For any cluster, each node in the cluster is mapped to a graph vertex, and based on the transaction data of each node, the edges between the graph vertices are constructed to obtain the graph structure data corresponding to the cluster; Determine the centrality index, clustering coefficient and importance index of each graph vertex in the graph structure data; For any graph vertex, determine the influence index of the graph vertex based on the centrality index, the clustering coefficient and the importance index of the graph vertex; Based on the influence index of each of the graph vertices and the mapping relationship between each of the graph vertices and each of the nodes, the influence index of each of the nodes is determined.

6. A management method for improving enterprise efficiency based on artificial intelligence according to claim 5, characterized in that: For any graph vertex, based on the centrality index, the clustering coefficient and the importance index of the graph vertex, determining the influence index of the graph vertex includes: For any graph vertex, obtaining a first weight of the graph vertex for the centrality index, a second weight for the clustering coefficient, and a third weight for the importance index; The influence index of the graph vertex is determined according to the sum of a first product of the centrality index and the first weight, a second product of the clustering coefficient and the second weight, and a third product of the importance index and the third weight.

7. The management method for improving enterprise efficiency based on artificial intelligence according to claim 1, characterized in that: Based on the business attribute characteristics of each cluster and the influence index corresponding to each node in each cluster, a data sampling strategy package for each cluster is generated, including: Based on the business attribute characteristics of each of the clusters, determining the data to be sampled and the initial sampling interval of each of the nodes in each of the clusters; For any cluster, based on the influence index corresponding to each of the nodes in the cluster, the initial sampling interval corresponding to the cluster is corrected to obtain the target sampling interval corresponding to each of the nodes; wherein the influence index is negatively correlated with the target sampling interval; Based on the to-be-sampled data and the target sampling interval of each of the nodes in each of the clusters, a data sampling strategy for each of the nodes in each of the clusters is determined to obtain the data sampling strategy package.

8. The management method for improving enterprise efficiency based on artificial intelligence according to claim 1, characterized in that: According to the data sampling strategy package, the historical performance association data of each node in the cluster is collected to obtain a sample data set, including: For any node, based on the data sampling strategy corresponding to the node, multiple historical performance-related data of the node are collected; the historical performance-related data includes historical management strategies and historical performance data; For any historical performance-related data, a labeling request for the historical performance-related data is sent to the node, and an optimization strategy label is obtained based on the optimization strategy labeling information fed back by the node, and a performance label is obtained based on the historical performance data; Based on the management strategy, the performance label and the optimization strategy label, sample data corresponding to the historical performance association data is generated.

9. A management method for improving enterprise efficiency based on artificial intelligence according to claim 8, characterized in that: The initial management strategy is optimized based on the target effectiveness optimization model to obtain the target management strategy corresponding to the cluster, including: Inputting the initial management strategy into the target performance optimization model, and outputting the predicted performance and the target optimization strategy for at least one item to be optimized in the initial management strategy; Based on the target optimization strategy, at least one item to be optimized in the initial management strategy is optimized to obtain the target management strategy corresponding to the cluster.

10. A management system based on artificial intelligence to improve enterprise efficiency, characterized in that: The system comprises: A node clustering module is used to obtain business data of multiple nodes of a target enterprise, and cluster the multiple nodes based on the business data of the multiple nodes to obtain multiple clusters; different clusters correspond to different business attribute characteristics; An indicator determination module is used to obtain, for any cluster, the influence indicators of each node output by the indicator identification model based on the transaction data of each node in the cluster and a preset indicator identification model, and determine the node with the highest influence indicator as the central node of the cluster; A sampling strategy generation module, used to generate a data sampling strategy package for each cluster based on the business attribute characteristics of each cluster and the influence index corresponding to each node in each cluster; the data sampling strategy package includes a data sampling strategy for each node in the cluster; A model training module is used to send the data sampling strategy package and the initial performance optimization model of any cluster cluster to the central node of the cluster cluster, so that the central node collects the historical performance-related data of each node in the cluster cluster according to the data sampling strategy package to obtain a sample data set; and train the initial performance optimization model based on the sample data set to obtain a target performance optimization model corresponding to the cluster cluster; A management strategy generation module, used to input the business demand information and resource information uploaded by each node of each cluster into a strategy generation model, and output an initial management strategy for each cluster; The management strategy optimization module is used to send the initial management strategy of any cluster cluster to the central node of the cluster cluster, so that the central node optimizes the initial management strategy based on the target performance optimization model, obtains the target management strategy corresponding to the cluster cluster, and manages each of the nodes in the cluster cluster based on the target management strategy.

Images