An intelligent management and control method and system for an enterprise service platform based on function analysis

Abstract

The application discloses an enterprise service platform intelligent management and control method and system based on function analysis, relates to the technical field of platform intelligent management and control, and comprises the following steps: collecting the calling frequency, resource occupancy rate, response time delay, user behavior track and abnormal event data of a function module, performing standardization and time sequence verification, and constructing a multi-dimensional running data flow. A function behavior model is constructed based on a graph neural network, module running characteristics and structural dependencies are extracted, a function heat score graph is dynamically generated, and visual display of module load and abnormal hotspots is realized. A function twin copy is constructed for a key module, a management and control strategy is simulated in an isolated simulation environment, and feedback is obtained. A multi-level management and control strategy generation matrix is constructed based on the feedback, candidate strategies are generated and optimized, and the candidate strategies are dynamically deployed, continuously monitored and automatically adjusted. The system adopts a hierarchical rollback mechanism, realizes adaptive rollback of the strategy in an abnormal situation, and guarantees stable and reliable operation of the platform.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of platform intelligent management and control, in particular to an enterprise service platform intelligent management and control method and system based on function analysis. BACKGROUND

[0002] With the continuous development of enterprise service platforms, the number and complexity of function modules in enterprise service platforms are increasing. The running data of each function module is diverse, including calling frequency, resource occupancy rate, response time delay, user behavior trajectory and abnormal events. How to accurately collect and effectively utilize these multi-dimensional running data to build a behavior model reflecting the function activity, stability and load characteristics has become a key problem of platform intelligent management and control.

[0003] In addition, the deployment of current management and control strategies lacks sufficient security verification means, making it difficult to avoid system performance degradation and business interruption caused by strategy failure or unreasonable adjustment. At the same time, the lack of real-time strategy execution effect monitoring and automatic rollback mechanism increases the risk of platform operation and cannot guarantee the continuity and stability of enterprise service platform operation. SUMMARY

[0004] To solve the above technical problems, an enterprise service platform intelligent management and control method and system based on function analysis are provided. The technical solution solves the problem of how to accurately collect and effectively utilize these multi-dimensional running data to build a behavior model reflecting the function activity, stability and load characteristics, and the current management and control strategy lacks sufficient security verification means and lacks real-time strategy execution effect monitoring and automatic rollback mechanism.

[0005] To achieve the above purposes, the technical solution adopted by the present application is:

[0006] An enterprise service platform intelligent management and control method based on function analysis, specifically comprising:

[0007] Obtain multi-dimensional running data stream, and build a function behavior model through a graph neural network to comprehensively reflect the running state and mutual relationship of function modules and extract the running characteristics of each function module;

[0008] According to the running characteristics of each function module, a function heat score map is dynamically generated, and the function module load distribution and abnormal hotspots are visualized through the heat score map;

[0009] According to the heat score map, identify key function modules, build function twin copies for the identified key function modules, and simulate the execution of the proposed management and control strategy in a simulation environment isolated from the production environment to obtain simulation feedback results, the management and control strategy refers to the control scheme for key function modules in terms of resource allocation, task scheduling and access permission configuration;

[0010] Based on the functional behavior model, the thermal score map, and the simulation feedback result of the functional twin copy, a multi-level control strategy generation matrix is constructed, a candidate control strategy template is generated, and a self-evolving control strategy set is formed through a strategy optimization and deployment mechanism and is deployed to an enterprise service platform.

[0011] By monitoring the key operation characteristics extracted based on the functional behavior model during the execution of the control strategy of the enterprise service platform in real time, combining the control flow chart and feedback indicators, the execution effect of the strategy is dynamically evaluated, and a closed-loop feedback mechanism is driven to realize strategy self-adaptation.

[0012] When the execution effect of the strategy is abnormal, a hierarchical rollback mechanism is triggered according to the abnormal level and system feedback information, and relevant rollback events and operation data are recorded by the system to update the rollback knowledge base.

[0013] In an optional embodiment, the multi-dimensional operation data stream is obtained, and a functional behavior model is constructed through a graph neural network to comprehensively reflect the operation state and mutual relationship of the functional modules, and the operation characteristics of each functional module are extracted, specifically including:

[0014] Based on the industrial communication protocol, the distributed control system (DCS), programmable logic controller (PLC), edge acquisition terminal, and user interaction log system connected to each functional module in the enterprise service platform are accessed, the calling frequency, resource occupancy rate, response time delay, user behavior trajectory, and abnormal event corresponding operation data generated during the operation of each functional module are collected, and an original operation data set is obtained;

[0015] The collected original operation data is input to the data adaptation middleware deployed in the platform access layer for format standardization conversion to obtain standardized operation data in a unified format, and is uploaded in batches to the data aggregation module of the enterprise service platform based on a preset time interval;

[0016] The standardized operation data is processed by the data aggregation module for time synchronization, field regularization, and preliminary verification to obtain a multi-dimensional operation data stream;

[0017] Based on the multi-dimensional operation data stream, a functional behavior modeling method using a graph neural network is used to construct a functional behavior model, the interaction topological structure between the functional modules and the dynamic index sequence related to resource use and behavior response are fused, and the structural dependency relationship and operation characteristics of each functional module are extracted.

[0018] In an optional embodiment, the functional thermal score map is dynamically generated according to the operation characteristics of each functional module, and the functional thermal score map is used to realize visual display of the load distribution and abnormal hotspots of the functional modules, specifically including:

[0019] Based on the calling frequency, resource occupancy and response time delay, a local performance tensor is constructed to describe the performance state of the module , wherein is the spatial coordinates of the module, is the time, corresponding to the calling frequency, resource occupancy and response time delay respectively;

[0020] Combined with user behavior trajectory data, a deep clustering algorithm is used to model the access frequency and path pattern of users between different functional modules, construct a spatial activity distribution map, and extract the corresponding access heat intensity function , which represents the user activity mapping on the space;

[0021] In the constructed multi-dimensional running data stream, the feature dimensions related to system anomalies are extracted, including system error codes, retry behaviors and interface response timeout logs, and an abnormal event feature sequence is constructed;

[0022] Based on the abnormal event feature sequence, a time series feature extraction method is used, and a deep autoencoder is used to unsupervisedly learn the propagation path, disturbance intensity and action range information of the abnormal event, to generate an abnormal disturbance factor function describing the abnormal evolution state ;

[0023] Through the time series modeling method, the time series changes of the local performance tensor, user activity and abnormal disturbance factor are fused to generate a spatio-temporal heat score function of the functional module: ,

[0024] wherein, represents the spatial position the input feature vector at the time, which is defined as:

[0025]

[0026] In the formula, respectively represent the calling frequency, resource occupancy and response time delay of the module at the time, represents the user access activity at the time at this position, represents the abnormal disturbance factor at the time;

[0027] Based on the heat value output by the heat score function, a three-dimensional functional heat score map is constructed, and the module load and abnormal distribution dynamics are visualized in the form of a heat map.

[0028] In an optional embodiment, the key function module is identified according to the thermal score map, a function twin copy is constructed for the identified key function module, and a proposed control strategy is simulated to obtain simulation feedback results in a simulation environment isolated from the production environment, specifically including:

[0029] According to the spatial distribution characteristics of the thermal value in the three-dimensional function thermal score map and the time evolution trend thereof, and in combination with the structural dependency relationship between the modules, a key function module with an abnormal sudden increase in thermal value or continuously in a high-thermal state is identified;

[0030] For the identified key function module, a corresponding function twin copy is constructed in a simulation environment isolated from the production environment of the enterprise service platform, based on the structural dependency relationship and running characteristics extracted from the function behavior model thereof, and in combination with the historical evolution track thereof during running;

[0031] In the function twin copy, in combination with resource allocation, task scheduling and access permission configuration, a plurality of candidate control strategies are proposed, and the candidate control strategies are pre-screened based on historical control strategy execution feedback information of the system to obtain a plurality of primary screening control strategy schemes;

[0032] An interactive simulation sandbox with behavior track playback and intervention response recording capabilities is deployed to simulate the execution process of the primary screening control strategy scheme in the function twin copy, collect the performance of each candidate control strategy in the simulation environment and the mitigation effect on abnormal propagation, and obtain simulation feedback results of the candidate control strategy.

[0033] In an optional embodiment, the function behavior model, the thermal score map and the function twin copy simulation feedback results are used to construct a multi-level control strategy generation matrix, generate a candidate control strategy template, and form a self-evolving control strategy set through a strategy optimization and deployment mechanism, and the control strategy set is deployed to the enterprise service platform, specifically including:

[0034] Based on the running characteristics of the key function module in the function behavior model, the thermal distribution state and evolution trend thereof in the thermal score map, and the multi-dimensional simulation feedback results thereof in the twin simulation environment, a multi-level control strategy generation matrix is constructed, the matrix is indexed by the function module identifier, the resource intervention parameter, the scheduling sequence parameter and the permission configuration parameter are used as the strategy dimension, and a plurality of candidate control strategy templates are generated through multi-dimensional cross combination;

[0035] The candidate control strategy template is input into a strategy optimization module, the performance index response curve, the abnormal propagation change track and the task execution offset recorded in the simulation feedback results are used as the feedback-driven strategy optimization mechanism, and the most optimal control strategy set is dynamically screened out;

[0036] Performing a structured expression on the control policy set to generate a policy deployment package, which includes a policy parameter mapping table, policy trigger conditions, policy scope, and policy effectiveness priority, and is mapped and bound to the system configuration interface of the enterprise service platform;

[0037] Based on the structural dependencies between functional modules and their spatial location coordinates, the policy deployment package is injected into the running unit where the target module is located, and the policy is dynamically loaded through the platform's built-in policy mounting mechanism. The dynamic loading allows the policy version to be updated while the system is running;

[0038] By constructing a policy execution control flow chart, marking the policy trigger node, policy intervention path and policy effectiveness module, the control flow chart is maintained synchronously with the functional behavior model to reflect the policy response path and the interaction relationship between modules;

[0039] Based on the operating environment after policy deployment, the platform system continuously monitors the status of functional modules within the policy-effective area, and automatically adjusts the policy triggering frequency, action granularity and parameter weights based on the interaction between modules and resource competition dynamics.

[0040] In an optional embodiment, when the policy execution effect is abnormal, a hierarchical rollback mechanism is triggered according to the abnormality level and system feedback information, and the system records the relevant rollback events and operation data and updates the rollback knowledge base, specifically including:

[0041] By real-time monitoring of key operational features extracted based on functional behavior models during the execution of enterprise service platform control strategies, and automatically evaluating the execution effect of control strategies using comprehensive evaluation indicators, it is determined whether a rollback should be triggered.

[0042] Based on the assessment results, the abnormalities are graded and classified into mild abnormalities, moderate abnormalities, and severe abnormalities;

[0043] When a minor anomaly is identified, key parameters in the control strategy are automatically adjusted to attempt to restore system stability without the need for an immediate rollback.

[0044] When parameter fine-tuning fails to achieve the expected results and the degree of anomaly is determined to be moderate, the impact of the policy on some key functional modules is gradually revoked, and a partial rollback is implemented to control the spread of the anomaly;

[0045] When the abnormality is severe and a partial rollback fails to effectively alleviate the system abnormality, a full system rollback is performed to restore the enterprise service platform to its stable state before the policy adjustment.

[0046] Based on the rollback process, the intelligent anomaly diagnosis module continuously analyzes the root cause of the anomaly and dynamically adjusts the execution strength and scope of the rollback strategy based on the anomaly level and system feedback.

[0047] The system records all rollback events and related key operation data in real time, and builds a rollback knowledge base.

[0048] Further, an enterprise service platform intelligent management and control system based on function analysis is proposed, which is used to implement any of the above intelligent management and control methods, and specifically includes:

[0049] An original data acquisition module is configured to acquire operation data of each function module in the enterprise service platform, including call frequency, resource occupancy rate, response time delay, user behavior trajectory, and abnormal events.

[0050] A data format standardization module is configured to standardize the acquired original operation data, and generate data in a unified format for uploading to a data aggregation module.

[0051] A multi-dimensional operation data flow construction module is configured to perform time sequence alignment and field normalization on the standardized data, and construct a multi-dimensional operation data flow.

[0052] A function behavior modeling module is configured to establish a function behavior model based on a graph neural network, and extract operation characteristics of each function module.

[0053] A heat score generation module is configured to fuse a local performance tensor, user activity, and an abnormal disturbance factor, construct a heat score function, and generate a three-dimensional heat score map.

[0054] A key module identification and simulation module is configured to identify a key function module based on the heat score map, and construct a corresponding function twin copy to test a management and control strategy in a simulation environment.

[0055] A management and control strategy generation and deployment module is configured to construct a multi-level strategy generation matrix, form a candidate management and control strategy template, and deploy the template to a target module.

[0056] An execution effect evaluation and rollback module is configured to monitor the execution effect of the strategy in real time, trigger hierarchical rollback and update the rollback knowledge base when an abnormality occurs.

[0057] In an optional embodiment, the heat score generation module specifically includes:

[0058] A performance tensor construction unit is configured to extract call frequency, resource occupancy rate, and response time delay, and construct a local performance tensor according to spatial coordinates and time sequence.

[0059] A user behavior modeling unit extracts a user behavior trajectory and forms a user activity function;

[0060] An abnormal disturbance learning unit learns abnormal propagation and disturbance based on an abnormal event feature sequence by using a deep auto-encoder for unsupervised learning, and forms an abnormal disturbance factor;

[0061] A time series score modeling unit takes the performance tensor, activity and abnormal disturbance as input, and performs joint time series modeling by using an LSTM to output a thermal score function.

[0062] In optional embodiments, the key module identification and simulation module specifically comprises:

[0063] A key module identification unit automatically identifies key functional modules with abnormal sudden increase in thermal value and high thermal state by using the spatial distribution and time evolution trend of the thermal score map and combining the functional behavior model;

[0064] A functional twin construction unit constructs a corresponding functional twin copy based on the structural dependence and running characteristics of the key module, and realizes a simulation model isolated from the production environment;

[0065] A policy simulation test unit deploys the control policy in the simulation environment, executes the policy behavior trajectory playback, and collects the policy performance and abnormality alleviation effect feedback data.

[0066] In optional embodiments, the execution effect evaluation and rollback module specifically comprises:

[0067] An execution effect evaluation unit monitors the policy execution effect in real time, and evaluates the execution state in combination with the feedback index;

[0068] An abnormality level determination unit divides the abnormality level according to the comprehensive evaluation index;

[0069] A rollback policy scheduling unit adjusts the parameters, performs local rollback or full system rollback according to the abnormality level;

[0070] A rollback knowledge base construction unit records the rollback events and key running data, and constructs a rollback knowledge base.

[0071] Compared with the prior art, the present application has the following beneficial effects:

[0072] The scheme provides an intelligent management and control method and system of an enterprise service platform based on function analysis, multi-dimensional operation data of each function module is collected and standardized, multi-dimensional operation data flow is constructed, a function behavior model is established based on a graph neural network, the running state and correlation characteristics of the module are accurately reflected, and the problems of insufficient use of multi-dimensional data and inaccurate behavior modeling are effectively solved. Through dynamically generating a function heat score map and identifying key modules, combined with function twin copy simulation feedback, multi-level generation and self-evolution optimization of management and control strategies are realized, and the problem of insufficient safety verification of management and control strategies in the prior art is solved. At the same time, the system supports dynamic strategy effect evaluation based on real-time operation characteristics and feedback indexes, cooperates with a hierarchical rollback mechanism and intelligent abnormal diagnosis, guarantees stable operation and business continuity of the enterprise service platform, effectively reduces the risk caused by strategy failure, and improves the intelligent management and control capability and operation reliability of the platform. BRIEF DESCRIPTION OF DRAWINGS

[0073] Figure 1 A flowchart of an intelligent management and control method of an enterprise service platform based on function analysis is provided for the present application.

[0074] Figure 2 A function behavior model construction and key module identification flowchart is provided for the present application.

[0075] Figure 3 A management and control strategy generation, deployment and rollback mechanism flowchart is provided for the present application.

[0076] Figure 4 A system framework diagram of an intelligent management and control system of an enterprise service platform based on function analysis is provided for the present application. DETAILED DESCRIPTION

[0077] The following description is used to disclose the present application so that those skilled in the art can implement the present application. The preferred embodiments in the following description are only as examples, and other obvious modifications can be thought of by those skilled in the art.

[0078] REFERENCE Figure 1 - Figure 4 As shown in the figure, an intelligent management and control method of an enterprise service platform based on function analysis, specifically includes:

[0079] Obtain multi-dimensional operation data flow, and construct a function behavior model through a graph neural network to comprehensively reflect the running state and mutual relationship of the function module, and extract the running characteristics of each function module;

[0080] According to the running characteristics of each function module, a function heat score map is dynamically generated, and the function module load distribution and abnormal hotspots are visually displayed through the heat score map;

[0081] According to the thermal score map, a key function module is identified, a function twin copy is constructed for the identified key function module, a proposed control strategy is simulated and executed in a simulation environment isolated from the production environment, a simulation feedback result is obtained, and the control strategy refers to a regulation scheme for the key function module in terms of resource allocation, task scheduling and access permission configuration;

[0082] Based on the function behavior model, the thermal score map and the function twin copy simulation feedback result, a multi-level control strategy generation matrix is constructed, a candidate control strategy template is generated, and a self-evolving control strategy set is formed through a strategy optimization and deployment mechanism, and is deployed to the enterprise service platform;

[0083] By monitoring the key operating characteristics extracted based on the function behavior model during the execution of the control strategy of the enterprise service platform in real time, combining the control flow chart and the feedback index, the effect of the strategy execution is dynamically evaluated, and the closed-loop feedback mechanism is driven to realize the self-adaptation of the strategy.

[0084] When the effect of the strategy execution is abnormal, a hierarchical rollback mechanism is triggered according to the abnormal level and system feedback information, and the related rollback events and operating data are recorded by the system to update the rollback knowledge base.

[0085] Further, a multi-dimensional operating data stream is obtained, and a function behavior model is constructed through a graph neural network, which comprehensively reflects the operating state and mutual relationship of the function modules, and extracts the operating characteristics of each function module, specifically including:

[0086] Based on the industrial communication protocol, the distributed control system DCS, programmable logic controller PLC, edge acquisition terminal and user interaction log system connected to each function module in the enterprise service platform are accessed, the operating data corresponding to the calling frequency, resource occupancy rate, response time delay, user behavior trajectory and abnormal events generated during the operation of each function module are collected, and an original operating data set is obtained;

[0087] The collected original operating data is input into the data adaptation middleware deployed in the platform access layer for format standardization conversion to obtain standardized operating data in a unified format, and is uploaded in batches to the data aggregation module of the enterprise service platform based on a preset time interval;

[0088] The standardized operating data is processed by the data aggregation module for time synchronization, field regularization and preliminary verification to obtain a multi-dimensional operating data stream;

[0089] Based on the multi-dimensional operating data stream, a function behavior modeling method using a graph neural network is used to construct a function behavior model, which integrates the interaction topology between the function modules and the dynamic index sequence related to resource usage and behavior response, and extracts the structural dependency relationship and operating characteristics of each function module.

[0090] Specifically, in order to accurately model the running behavior of each functional module in the enterprise service platform and the interdependence relationship therebetween, the embodiment adopts a functional behavior modeling method based on a graph neural network (GNN) to construct a functional behavior model, and clearly extracts the structural dependence relationship and dynamic running characteristics of each functional module.

[0091] Firstly, based on the five types of core indexes including the calling frequency, resource occupancy rate, response time delay, user behavior trajectory and abnormal event contained in the multi-dimensional running data flow, an interaction topology graph between the functional modules is constructed. In the graph, each node represents a functional module, and the edge embodies the interaction connection and dependence relationship between the modules. The weight of the edge comprehensively considers the collaborative change, resource conflict, abnormal event propagation and other factors between the indexes, is automatically obtained through training based on the historical running data of the platform, and forms a structural dependence weight matrix reflecting the real business dependence;

[0092] Then, the model adopts a dependence-aware feature aggregation mechanism as an information propagation unit of the graph neural network; in this mechanism, the influence weight of the neighbor nodes on the center node is dynamically adjusted according to the dependence strength between the modules, so as to accurately capture the structural dependence relationship between the modules; for example, if the frequent occurrence of the abnormal event of module A is directly affected by the calling of module B, the model will enhance the feature transmission weight from module B, so as to ensure that the structural dependence is effectively reflected.

[0093] Meanwhile, in order to extract the running characteristics of the modules, the time sequence indexes (such as the response time delay fluctuation and resource occupancy change) of each module are encoded into a dynamic feature sequence as the input feature of the node. Through iterative calculation of the multi-layer graph convolution network, the model aggregates the state information of the neighbor nodes, and at the same time combines the time sequence characteristic change of itself, to form a running characteristic vector of each module, which comprehensively reflects the current state and evolution trend of the module.

[0094] Finally, the functional behavior model trained not only clearly depicts the structural dependence relationship between the modules, but also accurately extracts the dynamic running characteristics of each module, thereby providing a solid data foundation for subsequent generation of a function heat score graph, identification of a key module and construction of a function twin copy.

[0095] Further, a function heat score graph is dynamically generated according to the running characteristics of each functional module, and the function heat score graph is used to realize visual display of the load distribution and abnormal hotspots of the functional modules, specifically including:

[0096] Based on the calling frequency, resource occupancy rate and response time delay, a local performance tensor for describing the performance state of the module is constructed wherein, is the spatial coordinate of the module, is the time, , respectively, correspond to the calling frequency, resource occupancy rate and response time delay.

[0097] In combination with user behavior trajectory data, a deep clustering algorithm is used to model the access frequency and path pattern of users between different functional modules, construct a spatial activity distribution map, and extract the corresponding access heat intensity function , representing the user activity mapping on space;

[0098] In the constructed multi-dimensional running data stream, the feature dimensions related to system anomalies are extracted, including system error codes, retry behavior and interface response timeout logs, and the abnormal event feature sequence is constructed;

[0099] Based on the abnormal event feature sequence, a time series feature extraction method is used, and a deep autoencoder is used to unsupervisedly learn the propagation path, disturbance intensity and range information of the abnormal event, generating an abnormal disturbance factor function describing the abnormal evolution state;

[0100] Through a time series modeling method, the time series changes of local performance tensors, user activity and abnormal disturbance factors are fused to generate a spatio-temporal heat score function of the functional module: ,

[0101] where represents the input feature vector at the th time, which is defined as:

[0102]

[0103] In the formula, respectively represent the call frequency, resource occupancy rate and response time delay of the module at the th moment, represents the user access activity at the th moment at this position, represents the abnormal disturbance factor at the th moment;

[0104] Based on the heat value output by the heat score function, a three-dimensional functional heat score map is constructed, and the module load and abnormal distribution dynamics are visualized in the form of a heat map.

[0105] Further, according to the heat score map, key functional modules are identified, functional twin replicas are constructed for the identified key functional modules, and the proposed control strategy is simulated and executed in a simulation environment isolated from the production environment to obtain simulation feedback results, including:

[0106] According to the spatial distribution characteristics of the heat value and its time evolution trend in the three-dimensional functional heat score map, in combination with the structural dependency relationship between modules, key functional modules with abnormal sudden increase of heat value and continuously high heat state are identified;​

[0107] Specifically, after the three-dimensional functional thermal score map is constructed, the system first performs statistical analysis on the thermal values of each functional module in space to identify abnormal hotspot modules with thermal values significantly higher than the overall level. Then, for the change in the thermal value of each module, the system uses a sliding time window statistical method to calculate the mean and variance of the thermal value in each window, and uses a change point detection algorithm to automatically identify the time points at which the thermal value of the module changes significantly. By detecting the case where the thermal value remains at a high level in multiple consecutive windows, the system determines that the module is in a sustained high-heat state, thereby distinguishing between temporary abnormalities and long-term abnormalities.

[0108] Finally, combined with the structural dependency relationships between modules in the functional behavior model, the system performs weighted propagation analysis on the influence path of the abnormal module to determine the key functional modules that not only have abnormal thermal values but also have a significant impact on other modules. This comprehensive analysis method of spatial distribution, dynamic change, and network structure effectively improves the accuracy of key module identification.

[0109] For the identified key functional modules, in a simulation environment isolated from the enterprise service platform production environment, based on the structural dependency relationships and running characteristics extracted from the functional behavior model, and combined with the historical evolution trajectory in the running process, a corresponding functional twin replica is constructed.

[0110] In the functional twin replica, combined with resource allocation, task scheduling, and access permission configuration, multiple candidate control strategies are proposed, and based on the historical control strategy execution feedback information of the system, the candidate control strategies are pre-screened to obtain multiple primary screening control strategy schemes.

[0111] By deploying an interactive simulation sandbox with behavior trajectory playback and intervention response recording capabilities, the execution process of the primary screening control strategy scheme in the functional twin replica is simulated, the performance of each candidate control strategy in the simulation environment and the mitigation effect on abnormal propagation are collected, and the simulation feedback results of the candidate control strategies are obtained.

[0112] Specifically, to improve the efficiency and accuracy of control strategy generation and screening, the platform system first proposes multiple candidate control strategies in the functional twin replica around the three intervention dimensions of resource allocation, task scheduling, and access permission configuration. Each candidate strategy combines the running bottlenecks and dependency structure differences of key functional modules to construct representative control schemes through parameter perturbation and strategy permutation and combination.

[0113] After the candidate management strategy is formulated, the candidate management strategy is preliminarily screened based on system historical management strategy execution feedback information, and candidate management strategies that are obviously not matched with the existing running scene or have poor historical feedback effects are removed, to obtain multiple preliminary screening management strategy schemes; the historical management strategy is a strategy instance actually deployed and executed by the enterprise service platform in different running stages, usually including intervention elements such as resource allocation parameters, task scheduling logic and access permission configuration, and is bound with corresponding function modules, trigger conditions and target states, and each historical strategy has a clear applicable background, strategy structure and expected target; the feedback information is derived from the strategy execution log, module performance evaluation result and abnormal response process data recorded by the enterprise service platform in the actual running process, and the system forms a historical strategy feedback database through structured archiving and dynamic updating, supporting data-driven evaluation of the expected performance of the candidate strategy in a similar scene;

[0114] Subsequently, the system deploys the preliminary screening management strategy schemes one by one to an interactive simulation sandbox environment with behavior trajectory playback and intervention response recording capabilities; the simulation sandbox relies on the function twin copy of the platform to accurately reproduce the current system running state and module interaction logic, ensuring that the performance of the strategy in the simulation process can fully reflect its actual feasibility;

[0115] During the simulation process, the system tracks the intervention behavior of each strategy throughout the process, collects changes in key performance indicators (such as response time delay changes, resource utilization rate fluctuations, task delay offsets, etc.) and abnormal propagation paths, and evaluates its performance improvement effect on the system and its inhibition ability on abnormal diffusion; all simulation feedback data will be used as the basis for decision-making by the subsequent strategy optimization module, ensuring that the selected strategy has good controllability and adaptability.

[0116] Further, based on the function behavior model, the thermal score map and the simulation feedback results of the function twin copy, a multi-level management strategy generation matrix is constructed, a candidate management strategy template is generated, and a self-evolving management strategy set is formed through a strategy optimization and deployment mechanism and deployed to the enterprise service platform, specifically including:

[0117] Based on the running characteristics of the key function modules in the function behavior model, the thermal distribution state and its evolution trend in the thermal score map, and the multi-dimensional simulation feedback results in the twin simulation environment, a multi-level management strategy generation matrix is constructed, the generation matrix takes the function module identifier as the index dimension, takes the resource intervention parameters, scheduling sequence parameters and permission configuration parameters as the strategy dimensions, and generates multiple candidate management strategy templates through multi-dimensional cross combination;

[0118] It can be understood that the strategy generation matrix takes the unique identifier of the functional module as the index dimension to define the corresponding strategy action unit; in the strategy dimension, resource intervention parameters (such as CPU, memory limit threshold), task scheduling sequence parameters (such as scheduling priority, execution time) and access permission configuration parameters (such as call permission level, data isolation strategy) are set respectively; the system automatically generates a candidate management and control strategy template set covering different strategy combination scenarios by cross-combining the above multiple dimension parameters, providing input for subsequent strategy optimization and deployment;

[0119] The candidate management and control strategy template is input into the strategy optimization module, and based on the performance index response curve, abnormal propagation change trajectory and task execution offset recorded in the simulation feedback result, a feedback-driven strategy optimization mechanism is adopted to dynamically select the management and control strategy set with the best matching;

[0120] The management and control strategy set is structured and expressed to generate a strategy deployment package, which includes a strategy parameter mapping table, a strategy trigger condition, a strategy action range and a strategy effective priority, and is mapped and bound with the system configuration interface of the enterprise service platform;

[0121] Specifically, the generated candidate management and control strategy template will be input into the strategy optimization module built-in the platform. According to the feedback result of the candidate strategy in the functional twin simulation environment, the module extracts multi-dimensional information such as key performance index response curve (such as system response delay change, resource utilization fluctuation), abnormal propagation change trajectory (such as fault node influence range, propagation path shortening degree) and task execution offset (such as task completion time delay or advance), adopts a feedback-driven strategy optimization mechanism to dynamically select the candidate strategy template; finally, a number of management and control strategies with the highest matching degree, the most controllable risk and the most obvious performance improvement effect are selected to form a management and control strategy set;

[0122] Subsequently, the system structures and expresses the management and control strategy set to generate a strategy deployment package for deployment; the deployment package includes the parameter mapping table of each strategy (indicating the correspondence between the strategy parameters and the module running parameters), the strategy trigger condition (such as reaching the threshold of thermal value, resource occupation exceeding the limit), the strategy action range (specifying the affected functional module set) and the strategy effective priority (used for execution order when there is a conflict); the deployment package also completes the binding with the system configuration interface of the enterprise service platform, ensuring that the strategy parameters can be injected into the target module as needed, and having the ability to automatically load by module granularity;

[0123] Based on the structural dependency relationship between the functional modules and their spatial position coordinates, the strategy deployment package is injected into the running unit where the target module is located, and the dynamic loading of the strategy is completed through the strategy mounting mechanism built-in the platform, which allows updating the strategy version in the system running state.

[0124] By constructing a policy execution control flow chart, marking the policy trigger nodes, policy intervention paths and policy effectiveness modules, the control flow chart and the functional behavior model are maintained synchronously to reflect the policy response paths and the interaction between modules;

[0125] Based on the operational environment after policy deployment, the platform system continuously monitors the status of functional modules within the policy's effective area, and automatically adjusts the policy's trigger frequency, action granularity, and parameter weights based on the interaction between modules and resource competition dynamics.

[0126] On the one hand, after completing policy optimization and structured expression, the system locates the target module and its corresponding operating unit for each policy based on the structural dependencies between functional modules and their spatial location coordinates in the enterprise service platform. The system automatically injects the policy deployment package into the corresponding operating unit and dynamically loads the policy through the platform's built-in policy mounting mechanism. This dynamic loading process supports online hot deployment, which allows policy updates and version switching without stopping system operation, ensuring the continuity of policy effectiveness and minimal disruption to deployment.

[0127] On the other hand, to ensure visual tracking and full-process control of the policy execution process, the system simultaneously constructs a policy execution control flow chart. This flow chart uses the policy trigger node as the starting point, clearly defines the policy intervention path (such as the resource adjustment path and the permission change link) and the policy action module, forming a complete policy response chain. The control flow chart is kept updated with the functional behavior model to depict the interaction evolution and dependency adjustment between platform functional modules after policy intervention, supporting the system to track and analyze operational behavior evolution at the policy level.

[0128] After the policy is formally deployed and enters the running state, the system will continuously monitor the operating status of functional modules in the policy's effective area, and collect key indicators such as call frequency, response delay and resource usage changes; at the same time, the platform combines the dependencies between modules and the dynamics of resource competition to analyze in real time the impact of policy execution on the overall system stability; based on the monitoring results, the system can automatically adjust the policy triggering frequency (such as reducing the frequency of policy activation to alleviate system load), the granularity of action (such as refining module-level intervention to the interface level), and the weight configuration of policy parameters, so as to achieve adaptive dynamic adjustment and effect enhancement during the policy operation process.

[0129] Furthermore, when the policy execution effect is abnormal, a hierarchical rollback mechanism is triggered based on the abnormality level and system feedback information. The system records the relevant rollback events and operation data and updates the rollback knowledge base, including:

[0130] By monitoring the key operating characteristics extracted based on the functional behavior model during the execution of the enterprise service platform management and control strategy in real time, and using comprehensive evaluation indexes to automatically evaluate the execution effect of the management and control strategy, it is determined whether to trigger rollback;

[0131] According to the evaluation result, the abnormal situation is classified and determined, and is divided into slight abnormality, moderate abnormality and serious abnormality;

[0132] When it is determined to be a slight abnormality, the key parameters in the management and control strategy are automatically adjusted to try to restore the system operation stability, without immediately rolling back;

[0133] When the parameter fine-tuning fails to achieve the expected effect, and the abnormality degree is determined to be moderate abnormality, the influence of the strategy on part of the key function modules is gradually cancelled out, and local rollback is implemented to control the spread of abnormality;

[0134] When the abnormality degree is serious abnormality, and the local rollback fails to effectively alleviate the system abnormality, the whole system rollback is executed to restore the enterprise service platform to the stable state before the strategy adjustment;

[0135] Based on the rollback process, the intelligent abnormality diagnosis module continuously analyzes the abnormality root cause, combines the abnormality level and system feedback, and dynamically adjusts the execution strength and range of the rollback strategy;

[0136] The system records all rollback events and related key operating data in real time, and constructs a rollback knowledge base.

[0137] Specifically, after the deployment of the management and control strategy, the enterprise service platform system continuously monitors the operating performance in the execution process, constructs an operating characteristic index set based on the functional behavior model, such as the response time delay of the function module, the resource utilization rate change amplitude, the abnormal event triggering frequency, etc., and extracts and records the key characteristic values at fixed time intervals; the system constructs a multi-dimensional execution effect scoring model by combining the above characteristic parameters through the built-in comprehensive evaluation algorithm, forming a complete management and control strategy execution performance evaluation system;

[0138] In actual implementation process, the platform sets multiple evaluation thresholds; for example, the full score of comprehensive score is 100 points, less than 85 points is considered to be a slight abnormality, less than 70 points is a moderate abnormality, and less than 50 points is a serious abnormality; when the system score is lower than the slight abnormality threshold (such as 85 points), the platform will automatically trigger the abnormality determination process;

[0139] When a slight anomaly is detected, the system first optimizes the strategy through a parameter fine-tuning mechanism, such as adjusting key parameters such as resource allocation ratio, task scheduling time interval, or access permission granularity, to attempt to restore the stability of platform operation without withdrawing the strategy. If the execution effect is still not as expected after parameter adjustment, and the score falls to the moderate anomaly interval (e.g. below 70 points), the system will trigger a local rollback mechanism; the local rollback is based on the current functional behavior model and module dependency graph to locate the key modules within the affected range, revoke the intervention configuration of the strategy to them, and limit the further propagation of anomalies;

[0140] When the anomaly score further decreases to the severe level (e.g. below 50 points), and the local rollback fails to significantly alleviate the anomaly, the platform will start a full-system rollback process, withdraw all deployed strategies, and restore the platform to a stable state before the strategy deployment according to the system running state snapshot;

[0141] During the entire strategy anomaly handling process, the platform also schedules an intelligent anomaly diagnosis module to analyze real-time multi-dimensional factors such as anomaly trigger causes, propagation paths, and intervention responses, and dynamically adjusts the rollback range and execution pace to ensure that the rollback operation is targeted and controllable;

[0142] The system synchronously records strategy version information related to all rollback events, anomaly level determination results, trigger time, key parameter changes, and running feature indicators, and writes the above data into a rollback knowledge base as a decision support and strategy optimization basis for subsequent similar scenario anomaly handling. Through the continuous expansion and update of the knowledge base, the platform system has a continuously evolving rollback response capability.

[0143] Further, an enterprise service platform intelligent management and control system based on functional analysis is proposed, which is used to implement any of the above intelligent management and control methods, and specifically includes:

[0144] An original data acquisition module, the original data acquisition module is used to acquire running data of each functional module in the enterprise service platform, including call frequency, resource occupancy rate, response time delay, user behavior trajectory, and abnormal events;

[0145] A data format standardization module, the data format standardization module is used to standardize the collected original running data, and generate data in a unified format uploaded to a data aggregation module;

[0146] A multi-dimensional running data flow construction module, the multi-dimensional running data flow construction module is used to time-align and field-normalize the standardized data, and construct a multi-dimensional running data flow;

[0147] A functional behavior modeling module, the functional behavior modeling module is used to establish a functional behavior model based on a graph neural network, and extract running features of each functional module;

[0148] A thermal score generation module, the thermal score generation module is used for fusing a local performance tensor, user activity and an abnormal disturbance factor, constructing a thermal score function and generating a three-dimensional thermal score map;

[0149] A key module identification and simulation module, the key module identification and simulation module is used for identifying a key function module based on the thermal score map and constructing a corresponding function twin copy to test a control strategy in a simulation environment;

[0150] A control strategy generation and deployment module, the control strategy generation and deployment module is used for constructing a multi-level strategy generation matrix, forming a candidate control strategy template and deploying the candidate control strategy template to a target module;

[0151] An execution effect evaluation and rollback module, the execution effect evaluation and rollback module is used for monitoring a strategy execution effect in real time, triggering a hierarchical rollback and updating a rollback knowledge base when an abnormality occurs.

[0152] Further, the thermal score generation module specifically comprises:

[0153] A performance tensor construction unit, the performance tensor construction unit extracts a calling frequency, a resource occupancy rate and a response time delay, and constructs a local performance tensor according to a spatial coordinate and a time sequence;

[0154] A user behavior modeling unit, the user behavior modeling unit extracts a user behavior trajectory to form a user activity function;

[0155] An abnormal disturbance learning unit, the abnormal disturbance learning unit learns abnormal propagation and disturbance based on an abnormal event feature sequence by using a deep autoencoder for unsupervised learning to form an abnormal disturbance factor;

[0156] A time series score modeling unit, the time series score modeling unit takes the performance tensor, the activity and the abnormal disturbance as inputs, and performs joint time series modeling by using an LSTM to output a thermal score function.

[0157] Further, the key module identification and simulation module specifically comprises:

[0158] A key module identification unit, the key module identification unit automatically identifies a key function module with an abnormal sudden increase in a thermal value and a key function module that is continuously in a high-heat state by using a spatial distribution and a time evolution trend of the thermal score map and combining a function behavior model;

[0159] A function twin construction unit, the function twin construction unit constructs a corresponding function twin copy based on a structural dependence and a running feature of the key module to realize a simulation model isolated from a production environment;

[0160] The policy simulation test unit deploys the management and control policy in the simulation environment, performs policy behavior trajectory playback, and collects policy performance and abnormality mitigation effect feedback data.

[0161] Further, the execution effect evaluation and rollback module specifically includes:

[0162] The execution effect evaluation unit monitors the policy execution effect in real time, and evaluates the execution state in combination with the feedback index;

[0163] The abnormality level determination unit divides the abnormality level according to the comprehensive evaluation index;

[0164] The rollback policy scheduling unit takes parameter adjustment, local rollback or full system rollback according to the abnormality level;

[0165] The rollback knowledge base construction unit records the rollback events and key operation data, and constructs the rollback knowledge base.

[0166] In summary, the advantages of the present application are as follows: through the standardized collection and fusion of multi-dimensional operation data, the functional behavior model is constructed in combination with the graph neural network, the precise state perception and dynamic feature extraction of the functional module of the enterprise service platform are realized; based on the functional heat score map and the functional twin simulation copy, the key module is effectively identified and the management and control policy is optimized, the multi-level generation and adaptive evolution of the policy are supported; in addition, the real-time execution effect evaluation and hierarchical rollback mechanism are integrated, in cooperation with the intelligent abnormality diagnosis, the abnormal situation can be quickly responded, the system stability and reliability are guaranteed, and the intelligent management and control level and operation efficiency of the platform are comprehensively improved.

[0167] The basic principles, main features and advantages of the present application are shown and described above. Those skilled in the art should understand that the present application is not limited by the above examples, and the above examples and descriptions in the specification are only the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection claimed by the present application is defined by the appended claims and their equivalents.

Claims

1. A function analysis-based intelligent management and control method for an enterprise service platform, characterized in that, Comprise: Obtain multi-dimensional running data stream, and build function behavior model through graph neural network, comprehensively reflect the running state and mutual relationship of function modules, and extract the running characteristics of each function module; According to the running characteristics of each function module, dynamically generate function heat score map, and realize the visual display of function module load distribution and abnormal hot spot through the heat score map; According to the heat score map, identify the key function modules, build function twin copies for the identified key function modules, simulate the proposed control strategy in the simulation environment isolated from the production environment, obtain simulation feedback results, the key function modules are function modules with abnormal sudden increase and continuous high heat state, and the control strategy refers to the regulation scheme for key function modules in terms of resource allocation, task scheduling and access permission configuration; Based on the function behavior model, the heat score map and the simulation feedback results of the function twin copies, a multi-level control strategy generation matrix is constructed, a candidate control strategy template is generated, and a self-evolving control strategy set is formed through strategy optimization and deployment mechanism, and is deployed to the enterprise service platform; Through real-time monitoring of the key running characteristics extracted based on the function behavior model during the execution of the control strategy of the enterprise service platform, combining the control flow chart and feedback indicators, the strategy execution effect is dynamically evaluated, and the closed-loop feedback mechanism is driven to realize strategy self-adaptation; When the strategy execution effect is abnormal, trigger the hierarchical rollback mechanism according to the abnormal level and system feedback information, and record the related rollback events and running data, and update the rollback knowledge base; Based on the industrial communication protocol, access the distributed control system (DCS), programmable logic controller (PLC), edge acquisition terminal and user interaction log system connected to each function module in the enterprise service platform, collect the running data corresponding to the calling frequency, resource occupancy rate, response time delay, user behavior trajectory and abnormal events generated during the running of each function module, and obtain the original running data set; The collected original running data is input into the data adaptation middleware deployed in the platform access layer for format standardization conversion to obtain standardized running data in a unified format, and is uploaded in batches to the data aggregation module of the enterprise service platform based on a preset time interval; Through time synchronization, field regularization and preliminary checking processing of the standardized running data by the data aggregation module, a multi-dimensional running data stream is obtained; Based on the multi-dimensional running data stream, a function behavior model is built by using the function behavior modeling method of graph neural network, the interaction topology between function modules and the dynamic index sequence related to resource use and behavior response are fused, and the structural dependency relationship and running characteristics of each function module are extracted.

2. The intelligent management and control method of an enterprise service platform based on functional analysis according to claim 1, characterized in that, The function heat score map is dynamically generated according to the running characteristics of each function module, and the visual display of the function module load distribution and abnormal hot spot is realized through the heat score map, specifically including: A local performance tensor is constructed to describe the performance state of the module based on the calling frequency, resource occupancy and response time delay wherein, is the spatial coordinate of the module, is the time, corresponding to the calling frequency, resource occupancy and response time delay respectively; Combined with user behavior trajectory data, a deep clustering algorithm is used to model the access frequency and path pattern of users between different functional modules, construct a spatial activity distribution map, and extract the corresponding access heat intensity function , representing the user activity mapping on space; In the constructed multi-dimensional running data stream, extract the feature dimensions related to system abnormalities, including system error code, retry behavior and interface response timeout log, and build abnormal event feature sequence; Based on the abnormal event feature sequence, a time sequence feature extraction method is adopted, and a deep auto-encoder is used for unsupervised learning of the abnormal event propagation path, disturbance intensity and action range information, to generate an abnormal disturbance factor function describing the abnormal evolution state ; Through the time series modeling method, the local performance tensor, user activity and the time series change of abnormal disturbance factor are fused to generate the space-time thermal score function of the function module: , wherein, represents a spatial position the first input feature vector at the t-th time, which is defined as: wherein, respectively represent the calling frequency, resource occupancy and response time of the module at the time point; represent the user access activity at the time point of the location, represent the abnormal disturbance factor at the time point; Based on the heat value output by the heat force scoring function, a three-dimensional functional heat force scoring graph is constructed, and the module load and abnormal distribution dynamics are visually presented in the form of a heat map. 3.The intelligent management and control method of enterprise service platform based on function analysis of claim 1, characterized in that, The key functional modules are identified according to the heat force scoring graph, a functional twin copy is constructed for the identified key functional modules, and the proposed control strategy is simulated and executed in a simulation environment isolated from the production environment to obtain simulation feedback results, specifically including: According to the spatial distribution characteristics and time evolution trend of the heat value in the three-dimensional functional heat force scoring graph, combined with the structural dependency relationship between modules, the key functional modules with abnormal sudden increase and continuous high heat state of the heat value are identified; For the identified key functional modules, a corresponding functional twin copy is constructed in a simulation environment isolated from the production environment of the enterprise service platform based on the structural dependency relationship and running characteristics extracted from the functional behavior model, and combined with the historical evolution trajectory in the running process; In the functional twin copy, combined with resource allocation, task scheduling and access permission configuration, multiple candidate control strategies are proposed, and the candidate control strategies are pre-screened based on system historical control strategy execution feedback information to obtain multiple primary screening control strategy schemes; Through the deployment of an interactive simulation sandbox with behavior trajectory playback and intervention response recording capability, the execution process of the primary screening control strategy scheme in the functional twin copy is simulated, the performance of each candidate control strategy in the simulation environment and the mitigation effect on abnormal propagation are collected, and the simulation feedback results of the candidate control strategy are obtained.

4. The intelligent management and control method of an enterprise service platform based on functional analysis according to claim 1, characterized in that, Based on the functional behavior model, heat force scoring graph and functional twin copy simulation feedback results, a multi-level control strategy generation matrix is constructed, a candidate control strategy template is generated, and a self-evolving control strategy set is formed through strategy optimization and deployment mechanism, which is deployed to the enterprise service platform, specifically including: Based on the running characteristics of the key functional modules in the functional behavior model, the heat distribution state and its evolution trend in the heat force scoring graph, and the multi-dimensional simulation feedback results in the twin simulation environment, a multi-level control strategy generation matrix is constructed, the matrix is indexed with functional module identification as the dimension, and resource intervention parameters, scheduling sequence parameters and permission configuration parameters are used as strategy dimensions, and multiple candidate control strategy templates are generated through multi-dimensional cross combination; The candidate control strategy template is input into the strategy optimization module, based on the performance index response curve, abnormal propagation change trajectory and task execution offset recorded in the simulation feedback results, a feedback-driven strategy optimization mechanism is adopted to dynamically select the optimal control strategy set with the best matching; The control strategy set is structured and expressed to generate a strategy deployment package, which includes a strategy parameter mapping table, a strategy trigger condition, a strategy scope and a strategy priority, and is mapped and bound with the system configuration interface of the enterprise service platform; Based on the structural dependency relationship and spatial position coordinates between functional modules, the strategy deployment package is injected into the running unit where the target module is located, and the dynamic loading of the strategy is completed through the built-in strategy mounting mechanism of the platform, which allows updating the strategy version in the system running state; By constructing a strategy execution control flow chart, a strategy trigger node, a strategy intervention path and a strategy effective module are marked, and the control flow chart is maintained synchronously with the functional behavior model to reflect the interaction relationship between the strategy response path and the modules; Based on the running environment after the strategy is deployed, the platform system continuously monitors the state of the functional modules in the strategy effective area, and automatically adjusts the trigger frequency, action granularity and parameter weight of the strategy according to the interaction relationship between the modules and the resource competition dynamics.

5. The intelligent management and control method of an enterprise service platform based on functional analysis according to claim 1, characterized in that, When the strategy execution effect is abnormal, a hierarchical rollback mechanism is triggered according to the abnormal level and system feedback information, and the system records related rollback events and running data, updates the rollback knowledge base, and specifically includes: By real-time monitoring of the key running characteristics extracted based on the functional behavior model during the execution of the enterprise service platform management and control strategy, and using comprehensive evaluation indexes to automatically evaluate the execution effect of the management and control strategy, it is judged whether to trigger rollback; According to the evaluation result, the abnormal situation is judged by level, and is divided into slight abnormality, moderate abnormality and serious abnormality; When it is determined to be a slight abnormality, automatically adjust the key parameters in the management and control strategy to try to restore the system running stability, without immediately rolling back; When the parameter fine-tuning fails to achieve the expected effect, and the abnormality degree is determined to be moderate abnormality, gradually cancel the influence of the strategy on part of the key functional modules, and implement local rollback to control the spread of abnormality; When the abnormality degree is serious abnormality, and the local rollback fails to effectively alleviate the system abnormality, execute the whole system rollback to restore the enterprise service platform to the stable state before the strategy adjustment; Based on the rollback process, the intelligent abnormality diagnosis module continuously analyzes the abnormal root cause, dynamically adjusts the execution strength and range of the rollback strategy according to the abnormal level and system feedback; The system records all rollback events and related key running data in real time, and builds a rollback knowledge base.

6. An intelligent management and control system of an enterprise service platform based on functional analysis, used to implement the intelligent management and control method of any one of claims 1-5, and specifically comprising: An original data acquisition module, which is used to acquire running data of each functional module in the enterprise service platform, including calling frequency, resource occupancy rate, response time delay, user behavior trajectory and abnormal events; A data format standardization module, which is used to standardize the collected original running data, and generate data in a unified format for uploading to a data aggregation module; A multi-dimensional running data flow construction module, which is used to time-align and field-normalize the standardized data, and construct a multi-dimensional running data flow; A functional behavior modeling module, which is used to establish a functional behavior model based on a graph neural network, and extract running characteristics of each functional module; A thermal score generation module, which is used to fuse local performance tensors, user activity and abnormal disturbance factors, construct a thermal score function and generate a three-dimensional thermal score map; A key module identification and simulation module, which is used to identify key functional modules based on the thermal score map, and construct corresponding functional twin copies to test the management and control strategy in a simulation environment; The control strategy generation and deployment module is used for building a multi-level strategy generation matrix, forming a candidate control strategy template and deploying to a target module; The execution effect evaluation and rollback module is used for monitoring the strategy execution effect in real time, triggering hierarchical rollback when abnormal and updating the rollback knowledge base.

7. The intelligent management and control system of an enterprise service platform based on functional analysis according to claim 6, characterized in that, The thermal score generation module specifically includes: The performance tensor construction unit extracts the call frequency, resource occupancy rate and response delay, and constructs a local performance tensor according to the spatial coordinates and time sequence; The user behavior modeling unit extracts the user behavior trajectory and forms a user activity function; The abnormal disturbance learning unit learns the abnormal disturbance and propagation based on the abnormal event feature sequence by using a deep autoencoder for unsupervised learning, and forms an abnormal disturbance factor; The time series score modeling unit takes the performance tensor, activity and abnormal disturbance as input, and outputs a thermal score function through LSTM joint time series modeling.

8. The intelligent management and control system of an enterprise service platform based on functional analysis according to claim 6, characterized in that, The key module identification and simulation module specifically includes: The key module identification unit automatically identifies the key functional modules with abnormal sudden increase in thermal value and high heat state by using the spatial distribution and time evolution trend of the thermal score map, and combining the functional behavior model; The functional twin construction unit constructs the corresponding functional twin copy based on the structure dependence and running characteristics of the key module, and realizes the simulation model isolated from the production environment; The strategy simulation test unit deploys the control strategy in the simulation environment, executes the strategy behavior trajectory playback, and collects the strategy performance and abnormal relief effect feedback data.

9. The intelligent management and control system of an enterprise service platform based on functional analysis according to claim 6, characterized in that, The execution effect evaluation and rollback module specifically includes: The execution effect evaluation unit monitors the strategy execution effect in real time, and evaluates the execution state in combination with the feedback index; The abnormal level determination unit divides the abnormal level according to the comprehensive evaluation index; The rollback strategy scheduling unit adjusts the parameters, performs local rollback or full system rollback according to the abnormal level; The rollback knowledge base construction unit records the rollback events and key operation data, and constructs the rollback knowledge base.

Images