An intelligent service handling process management and control platform

Abstract

The application discloses a kind of based on intelligent service handling process management and control platform, specifically related to computer technology field, management and control platform is built-in dynamic process engine module, dynamic process engine module is used to parse third-party platform service process structure, support visual configuration process node and rule, realize the automatic execution and scheduling of process, and monitor feedback is carried out to process execution process.The application is adapted to the dynamic change of third-party platform by automatic analysis, strengthens learning dynamic scheduling resource, whole-link monitoring and intelligent diagnosis, systematically solves the problem of high adaptation cost, inefficient resource scheduling, insufficient monitoring capacity in traditional process control, and comprehensively improves the automation rate, execution efficiency and stability of service handling process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, more particularly, the present application relates to a kind of based on intelligent business handling process control platform. BACKGROUND

[0002] With the deepening of enterprise digital transformation, business handling across third-party platforms (such as electronic tax bureau, supplier system, financial institution platform, etc.) is becoming increasingly frequent, covering core scenarios such as tax filing, order processing, and fund settlement. However, existing business process control systems cannot adapt to the dynamic characteristics of third-party platforms: on the one hand, the interface structure and business rules of third-party platforms are frequently updated, requiring manual reanalysis of process nodes and flow logic, resulting in long business interruption time and high adaptation cost; on the other hand, there is a lack of automatic identification and response capability for abnormal interception components (such as CAPTCHA pop-up windows and risk prompt interfaces) in the process, which relies heavily on manual intervention, seriously affecting the automation rate of the process.

[0003] At the same time, the existing system has obvious shortcomings in resource scheduling and monitoring: when multiple processes are processed in parallel, resource allocation relies on fixed rules and cannot be dynamically adjusted according to real-time load, resulting in delayed response of high-priority tasks and low resource utilization; process monitoring can only display basic status and lacks deep traceability and intelligent diagnosis capability for node execution details, making it difficult to quickly locate bottlenecks and abnormal root causes, thereby restricting the improvement of business handling efficiency and stability. SUMMARY

[0004] To solve the above technical problems, the present application provides a kind of based on intelligent business handling process control platform, the control platform is built-in dynamic process engine module, the dynamic process engine module is used to analyze third-party platform business process structure, support visual configuration process node and rule, realize the automatic execution and scheduling of process, and monitor and feedback the process execution process.

[0005] To achieve the above purpose, the present application provides the following technical solutions:

[0006] A kind of based on intelligent business handling process control platform, the control platform is built-in dynamic process engine module, characterized in that, the dynamic process engine module is used to analyze third-party platform business process structure, support visual configuration process node and rule, realize the automatic execution and scheduling of process, and monitor and feedback the process execution process;

[0007] The dynamic process engine module includes a process analysis unit, a configuration management unit, an execution scheduling unit and a monitoring feedback unit, wherein:

[0008] The process analysis unit is used to analyze the node relationship and flow rule of third-party platform business process;

[0009] The configuration management unit is configured to support users to customize the process nodes and execution logic through a visual interface;

[0010] The execution scheduling unit is configured to be responsible for parallel processing and resource allocation of multiple processes;

[0011] The monitoring feedback unit is configured to collect process execution data in real time and generate performance reports.

[0012] As a preferred technical solution proposed in the present application, the process analysis unit analyzes the node relationship and flow transfer rule of the third-party platform business process, which includes:

[0013] Obtain the business process information of the target third-party platform; identify the platform type through the URL, page metadata or specific identification elements of the target platform, establish a connection using the corresponding communication protocol, collect the current version information of the target platform and compare it with the historical process model cached locally to determine whether the process analysis result needs to be updated; if it is determined that the process analysis result needs to be updated, the subsequent analysis steps are re-executed and a new process model is generated to replace the historical cache; if it is determined that the process analysis result does not need to be updated, the historical process model cached locally is directly called;

[0014] Then, the complete DOM structure of the target platform business process page is obtained through a browser automation tool, a node relationship tree is constructed, and key elements in the page are located and extracted; the key elements include form input components, operation buttons and process navigation components, and the visual features and layout relationships of non-standard UI components are identified by combining computer vision technology.

[0015] As a preferred technical solution proposed in the present application, when the process analysis unit locates and extracts key elements in the page, it also includes identifying abnormal interception components in the page and recording their trigger conditions and release methods; the abnormal interception components include CAPTCHA pop-up windows, permission verification windows and risk prompt interfaces.

[0016] As a preferred technical solution proposed in the present application, when the process analysis unit extracts process nodes, it extracts static nodes based on DOM structure and interactive elements, predicts dynamic nodes by monitoring page jump events, AJAX requests and URL change patterns, and labels metadata for each node;

[0017] The actual number of times that the dynamic node actually appears in the business scenario corresponding to the current process in the historical data is recorded as the actual number of times, the total number of occurrences of the scenario in the historical data, the number of associated events that have been triggered in the current process, and the average total number of associated events triggered in the scenario in the historical data;

[0018] The prediction probability of the appearance of the dynamic node is calculated by the actual number of times, the total number of occurrences, the number of associated events and the average total number.

[0019] The metadata includes a node type, a required field, a pre-operation, a time limit, and a data interaction format, and is associated with a platform knowledge graph to support cross-process node data reuse.

[0020] As a preferred technical solution of the present application, the configuration management unit supports users to customize process nodes and execution logic through a visual interface, and the specific process includes:

[0021] An interactive visual canvas is loaded, and a basic process template is preset in the canvas and a node component library that can be dragged is displayed;

[0022] An attribute configuration panel is triggered after the user drags a node to the canvas, and the panel supports setting node basic information, business parameters, and execution conditions;

[0023] A connection tool is used to define a node flow direction and flow rule including node relationship and priority association attributes, and a loop logic configuration is provided;

[0024] A visual script editor is used to define node execution logic;

[0025] The configured process is saved as a specific version and a valid time is set, and after publishing, it is synchronized to an execution scheduling unit.

[0026] As a preferred technical solution of the present application, when the configuration management unit defines a node flow direction and flow rule through a connection tool, the node relationship configuration of the configuration management unit supports priority setting; when multiple parallel nodes meet execution conditions, the user can define a node execution order, which includes sorting by creation time, manually specifying a priority value, and intelligently analyzing the priority of parallel nodes, and the execution order is linked with the resource allocation mechanism of the execution scheduling unit, so that nodes with high priority can obtain computing resources and network bandwidth preferentially.

[0027] As a preferred technical solution of the present application, the process of the execution scheduling unit responsible for parallel processing and resource allocation of multiple processes includes:

[0028] A process instance is obtained from a task queue, a process definition is parsed, and an execution task graph is constructed;

[0029] System resource status and external dependent system availability are monitored in real time;

[0030] Based on process priority, node dependency relationship, and resource status, a reinforcement learning algorithm is applied to generate a scheduling strategy;

[0031] An execution thread pool is created, and an independent thread is allocated for each node that can be executed in parallel, and node data synchronization and state interaction are realized through a message queue.

[0032] When the resource bottleneck or timeout condition is detected, the resource reallocation or task migration is automatically triggered; the execution results of all nodes are collected and fed back to the monitoring feedback unit.

[0033] As a preferred technical solution proposed in the present application, the reinforcement learning algorithm adopted by the execution scheduling unit comprises a multi-dimensional reward function:

[0034] The node information is acquired, including task completion time, resource utilization rate, execution success rate and abnormal handling efficiency; the node information is analyzed to obtain node analysis results, including task completion time score, resource utilization rate score, execution success rate score and abnormal handling efficiency score; a multi-dimensional reward function is constructed through the node analysis results, and a comprehensive reward value is calculated; the strategy is optimized through continuous learning of historical scheduling data.

[0035] As a preferred technical solution proposed in the present application, the process in which the monitoring feedback unit collects process execution data in real time and generates a performance report comprises:

[0036] Data collection probes are implanted in process key nodes and system resources, and collection parameters are configured;

[0037] A distributed stream processing framework is adopted to clean, aggregate and correlate the collected data, and generate standardized data streams;

[0038] The processed data is classified and stored in a time series database and a relational database, and a cross-database index is established;

[0039] Based on the abnormal detection model, the relative deviation rate between the index and the baseline is calculated in real time; a deviation threshold is set, and when the relative deviation rate is greater than the preset deviation threshold, a multi-level early warning mechanism is triggered;

[0040] Based on the user role, a visual dashboard is customized, and a machine learning algorithm is applied to generate optimization suggestions and predictive reports.

[0041] As a preferred technical solution proposed in the present application, the visual dashboard of the monitoring feedback unit supports multi-dimensional drilling function, and the user can drill down from the global view to the specific node layer by layer to view the execution log, input and output data and context information;

[0042] The dashboard further comprises an intelligent diagnosis module; the intelligent diagnosis module is further used to identify whether there is a bottleneck in the process node, specifically:

[0043] The actual execution time consumption of the current node and the historical average execution time consumption are acquired;

[0044] Identify the actual execution time of the node at any time in the set time region before the current time, and calculate the time consumption fluctuation of the actual execution time in the set time region by using the variance formula;

[0045] The actual execution time of the current node, the historical average execution time and the time consumption fluctuation are used to obtain a bottleneck score;

[0046] A preset bottleneck threshold is set, and if the bottleneck score is greater than the preset bottleneck threshold, an optimization scheme is pushed; the optimization scheme includes node merging suggestion, resource allocation adjustment strategy and rule engine parameter optimization direction.

[0047] The technical effects and advantages of the present application are as follows:

[0048] 1. The present application is aimed at the problem of adapting to the dynamic characteristics of the third-party platform. By automatically analyzing the process nodes, identifying abnormal interception components and recording the triggering and release methods, the manual intervention is reduced, the problems of business interruption and high adaptation cost caused by interface update are solved, and the process automation rate and continuity are improved.

[0049] 2. For the resource scheduling short board, a dynamic scheduling strategy is generated by combining the reinforcement learning algorithm, the resource allocation is adjusted according to the real-time load, the problems of high priority task delay and low resource utilization under fixed rules are solved, and the efficient execution of multiple processes in parallel is ensured.

[0050] 3. The present application is aimed at the problem of insufficient monitoring capability. Through full-link data tracking, intelligent diagnosis and multi-dimensional dashboard, accurate early warning and bottleneck positioning are realized, the problems of lack of deep traceability and diagnosis in traditional monitoring are solved, and the business handling efficiency and stability are improved.

[0051] In summary, the present application automatically analyzes and adapts to the dynamic changes of the third-party platform, dynamically schedules resources by reinforcement learning, and monitors and diagnoses the full link. The problems of high adaptation cost, low resource scheduling efficiency and insufficient monitoring capability in traditional process control are systematically solved, and the automation rate, execution efficiency and stability of the business handling process are comprehensively improved. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 A module connection relationship diagram based on the intelligent business handling process control platform is provided. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present application will be described in detail below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0054] As Figure 1 shown, an intelligent business handling process control platform is provided, which is internally provided with a dynamic process engine module, the dynamic process engine module is used for analyzing third-party platform business process structure, supporting visual configuration of process nodes and rules, realizing automatic execution and scheduling of the process, and monitoring and feeding back the process execution process;

[0055] The dynamic process engine module includes a process analysis unit, a configuration management unit, an execution scheduling unit and a monitoring feedback unit, wherein:

[0056] The process analysis unit is used for analyzing the node relationship and flow rule of the third-party platform business process;

[0057] The configuration management unit is used for supporting users to customize process nodes and execution logic through a visual interface;

[0058] The execution scheduling unit is used for parallel processing and resource allocation of multiple processes;

[0059] The monitoring feedback unit is used for real-time collection of process execution data and generation of performance reports.

[0060] In the present application, the process of analyzing the node relationship and flow rule of the third-party platform business process by the process analysis unit includes:

[0061] Obtain the business process information of the target third-party platform; identify the platform type through the URL, page metadata or specific identification element of the target platform, establish a connection using the corresponding communication protocol, collect the current version information of the target platform and compare it with the historical process model cached locally to determine whether the process analysis result needs to be updated; if it is determined that the process analysis result needs to be updated (i.e., the current version is inconsistent with the historical model), the subsequent analysis steps are re-executed and a new process model is generated to replace the historical cache; if it is determined that the process analysis result does not need to be updated (i.e., the current version is consistent with the historical model), the historical process model cached locally is directly called; it should be noted that the comparison method of the current version information and the historical process model cached locally uses existing technology (such as conventional version checking methods based on version number field matching, key feature hash value comparison, etc.), and the present application does not improve the comparison method itself, but only uses it to realize the judgment of version consistency to trigger the selection execution of the subsequent analysis logic;

[0062] The complete DOM structure of the target platform business process page is acquired through a browser automation tool, a node relationship tree is constructed, and key elements in the page are located and extracted; the key elements include form input components, operation buttons, and process navigation components, and computer vision technology is used to identify the visual features and layout relationships of non-standard UI components, wherein the visual features include shape, color, size, and icon style, and the layout relationships include relative position, arrangement order, and nesting level of other components; the key elements are completely extracted and their functional roles and associated logic with other components are accurately restored to support the complete construction of the node relationship tree.

[0063] In the present application, when locating and extracting key elements in the page, the process analysis unit also includes identifying abnormal interception components in the page and recording their trigger conditions and release methods; the abnormal interception components include CAPTCHA pop-up windows, permission verification windows, and risk prompt interfaces.

[0064] It should be noted that by setting abnormal interception components and recording their trigger conditions and release methods, the trigger conditions (such as high-frequency operation triggering CAPTCHA, insufficient permissions triggering verification windows) and release methods (such as correctly entering the CAPTCHA, supplementing permission credentials) can be recorded in advance to predict possible obstacles encountered during process execution; during subsequent process automation execution, the interception can be automatically responded and released to avoid process interruption and ensure the continuity and stability of the business handling process, improve the ability of the intelligent management and control platform to cope with complex scenarios, and make up for the shortcomings of traditional process analysis that only focuses on regular nodes and ignores abnormal interception interference.

[0065] In the present application, when extracting process nodes, the process analysis unit extracts static nodes based on DOM structure and interactive element analysis, and predicts dynamic nodes by monitoring page jump events, AJAX requests, and URL change patterns, and labels metadata for each node;

[0066] The actual number of times of occurrence of the dynamic node in the current process under the business scenario in the historical data is recorded as the actual number C, the total number of occurrences T of the scenario in the historical data, the number of associated events K triggered in the current process, and the average total number of associated events in the scenario in the historical data is recorded as ;

[0067] The prediction probability P of the occurrence of the dynamic node is calculated by the actual number, the total number of occurrences, the number of associated events, and the average total number, and the formula is , wherein is the weight coefficient of the historical data dimension, represents dynamic node prediction based on historical data, which can reflect long-term experience rules, represents dynamic node prediction based on real-time events, which can reflect the real-time trend of the current process; Specifically, the data / information can be dynamically or statically configured according to business process stability (such as version update, node change frequency), historical data reliability (such as integrity, error rate), real-time event correlation (such as the correlation between events and node triggering), and the like according to scene requirements.

[0068] The metadata includes node type, mandatory field, pre-operation, time limit, and data interaction format, and is associated with the platform knowledge graph to support cross-process node data reuse.

[0069] It should be noted that the present application extracts static nodes through DOM structure, predicts dynamic nodes by combining page interaction monitoring, and calculates the probability of dynamic node occurrence based on historical and real-time data. It not only accurately captures the process infrastructure through DOM, but also adapts to process changes through dynamic prediction. The annotated metadata is associated with the knowledge graph, enabling cross-process data reuse, improving parsing efficiency and data utilization value, making process parsing more comprehensive and intelligent, providing accurate, flexible, and reusable node information support for subsequent process configuration and execution, and solving the problem of traditional parsing difficulty in adapting to dynamic changes and data isolation.

[0070] In the present application, the configuration management unit supports users to customize process nodes and execution logic through a visual interface, and the specific process includes:

[0071] Load an interactive visual canvas, which predefines a basic process template and displays a dragable node component library;

[0072] Trigger the attribute configuration panel after the user drags the node to the canvas, which supports setting node basic information, business parameters, and execution conditions;

[0073] Define the node flow direction and flow rule containing node relationship and priority association attribute through the connection tool, and provide loop logic configuration;

[0074] Define node execution logic through a visual script editor;

[0075] Save the configured process as a specific version and set the effective time, and after publishing, synchronize to the execution scheduling unit.

[0076] It should be noted that the present application provides visual canvas, dragable components, attribute panels, and script editors, etc. Users can intuitively customize process nodes and execution logic, configure node relationship, priority, and loop logic through the connection tool, and support version management and effective time setting. This not only reduces the technical threshold of process configuration and meets the operation needs of non-professional users, but also ensures the accuracy and traceability of process logic through structured configuration. The mechanism of synchronizing to the execution scheduling unit realizes seamless connection between configuration and execution, and improves the flexibility and efficiency of process control.

[0077] In the present application, when the node flow direction and flow rule are defined by the connection tool in the configuration management unit, the node relationship configuration of the configuration management unit supports priority setting; when multiple parallel nodes meet the execution condition, the user can define the node execution order, which includes sorting by creation time, manually specifying priority value, and intelligently analyzing the priority of parallel nodes, and the execution order is linked with the resource allocation mechanism of the execution scheduling unit, so that the nodes with high priority can obtain computing resources and network bandwidth preferentially.

[0078] Among them, the priority of the parallel nodes is intelligently analyzed, which is specifically realized by the following way:

[0079] The business value weight V, the execution risk level R and the process criticality K associated with the node are obtained, the business influence factor I is calculated, and the calculation formula is , wherein is the business value weight coefficient, is the preset maximum risk level; wherein the business value weight can be obtained by the value score of the corresponding business of the node, which is specifically marked by a business expert, and the range is [0, 1], such as a tax declaration node V=0.9, and a general information query node V=0.3; the execution risk level can be preset, and the process criticality can be determined by whether the node is a “must-pass node” or a “data submission node”. If it is a must-pass node, K=1, and if it is a non-must-pass node, K=0.2;

[0080] The number of upstream dependent nodes N, the data interaction amount Q and the data criticality coefficient C of the node are obtained, the data dependency D is calculated, and the calculation formula is ; wherein is the average number of upstream dependencies of the same type of process node, is the average data interaction amount of the same type of process node, and the data criticality coefficient is a preset coefficient, specifically, if it is a critical business data C=1, and if it is a general log data C=0.3, is the preset dependency number weight coefficient;

[0081] The remaining executable time R, the historical timeout rate O and the timeliness sensitivity coefficient S of the current node are obtained, the execution timeliness requirement E is calculated, and the calculation formula is ; wherein is the total execution time of the node, is the preset remaining time weight coefficient; the historical timeout rate can be obtained by dividing the number of historical timeout times by the total execution times, and the timeliness sensitivity coefficient is specifically set by the business category, such as S=1 for a timeliness sensitive business and S=0.4 for a non-sensitive business.

[0082] The business impact factor, data dependency, and execution timeliness requirement are weighted to obtain a priority judgment value of the node; and the nodes are automatically assigned priorities according to the size order of the priority judgment values of the nodes.

[0083] It should be noted that the application defines the parallel node execution order in three ways: according to the creation time, manual specification, and intelligent analysis (multi-dimensional calculation of business value, data dependency, and execution timeliness), and links the resource allocation mechanism, which not only meets the user's flexible configuration needs, but also realizes the scientific quantification of priorities through intelligent analysis, ensures that high-value, high-dependence, and high-timeliness nodes obtain resources first, avoids resource waste and process blocking, and improves the efficiency and rationality of process execution in the multi-node parallel scene.

[0084] In the application, the process of the execution scheduling unit responsible for the parallel processing of multiple processes and the allocation of resources includes:

[0085] Obtain a process instance from a task queue, parse a process definition, and construct an execution task graph;

[0086] Real-time monitoring of system resource status and external dependency system availability;

[0087] Based on the process priority, node dependency relationship, and resource status, a reinforcement learning algorithm is applied to generate a scheduling strategy;

[0088] Create an execution thread pool and assign an independent thread to each node that can be executed in parallel, and realize node data synchronization and state interaction through a message queue;

[0089] When a resource bottleneck or timeout condition is detected, automatically trigger resource reallocation or task migration; collect the execution results of all nodes and feed them back to the monitoring feedback unit.

[0090] It should be noted that the application constructs an execution task graph, real-time monitors resource and dependency system status, generates a scheduling strategy in combination with a reinforcement learning algorithm, allocates independent threads to parallel nodes and synchronizes data with the help of a message queue, and automatically reallocates resources or migrates tasks when a resource bottleneck or timeout occurs, which not only realizes the orderly scheduling of multi-process parallel processing, but also dynamically adapts to system load changes, avoids resource idling and task delay, ensures efficient and stable process execution, and improves the platform's control ability over complex multi-task scenarios.

[0091] In the application, the reinforcement learning algorithm used by the execution scheduling unit includes a multi-dimensional reward function:

[0092] Obtain node information, including task completion time, resource utilization, execution success rate, and abnormal handling efficiency; analyze the node information to obtain node analysis results, including task completion time score , resource utilization score , the success rate score of execution , and the efficiency score of exception handling ; a multi-dimensional reward function is constructed based on the node analysis result, and the expression is , and the comprehensive reward value F is calculated , respectively, the weight coefficients of the scores of each dimension in the node analysis result; the strategy is optimized by continuously learning the historical scheduling data, specifically: the comprehensive reward value F under different strategies in the historical scheduling data is taken as a learning sample, and the weight coefficients and the scheduling strategy are continuously adjusted by a strategy gradient algorithm (such as A2C, PPO, and other existing reinforcement learning optimization methods), so that the algorithm optimizes the process scheduling performance in iteration.

[0093] Among them, the node information is analyzed, and the node analysis result is obtained by normalization processing and index mapping, specifically:

[0094] The actual time consumption t1 is compared with the preset maximum allowed time tmax, and the task evaluation value is calculated by the formula

[0095] The actual resource utilization rate is compared with the preset optimal utilization rate threshold , and the resource utilization evaluation value is calculated by the formula

[0096] The ratio of the number of successful executions to the total number of executions is directly taken as the execution success evaluation value;

[0097] The exception handling time is compared with the preset maximum processing time to obtain the exception handling evaluation value;

[0098] The task evaluation value, the resource utilization evaluation value, the execution success evaluation value, and the exception handling evaluation value are taken as the node evaluation result;

[0099] The preset node monitoring time length is taken as the node monitoring time zone, and the node evaluation result at any collection time of the node monitoring time zone is obtained;

[0100] The statistical indicators of any evaluation value in the node evaluation result are calculated, including the mean value, the difference between the maximum value and the minimum value, and the variance value;

[0101] The corresponding evaluation value in the node evaluation result and its corresponding statistical indicators are weighted to obtain the corresponding score; all the scores in the node evaluation result are taken as the node analysis result.

[0102] ​​It should be noted that the score is calculated by weighting the statistical indicators of the node evaluation results in the monitoring period, such as the mean value, fluctuation range (difference between maximum value and minimum value), dispersion degree (variance), etc. The timeliness of real-time evaluation values is retained, and the stability analysis of historical performance is integrated, so that the node analysis results are more comprehensive. In combination with the multi-dimensional reward function and the existing reinforcement learning algorithm optimization strategy, while preferentially guaranteeing high-priority tasks, through the balanced consideration of indicators such as resource utilization, the over-occupation of resources by a single task is avoided, indirectly realizing the dynamic reconciliation of resource allocation between efficiency and fairness, and improving the rationality and adaptability of the scheduling strategy.

[0103] In the present application, the process in which the monitoring feedback unit collects process execution data in real time and generates a performance report includes:

[0104] Data collection probes are implanted in process key nodes and system resources, and collection parameters are configured;

[0105] A distributed stream processing framework is used to clean, aggregate and correlate the collected data to generate standardized data streams;

[0106] The processed data is classified and stored in a time series database and a relational database, and a cross-database index is established;

[0107] Based on the abnormality detection model, the relative deviation rate D between the index and the baseline is calculated in real time, through the deviation rate formula: is realized, where X is the real-time index value, is the baseline value, which can be obtained by calculating the mean value of historical normal data; a deviation threshold is set, and when the relative deviation rate is greater than the preset deviation threshold, a multi-level early warning mechanism is triggered;

[0108] Based on user roles, visual dashboards are customized, and machine learning algorithms are used to generate optimization suggestions and predictive reports.

[0109] It should be noted that by implanting probes to collect data at key nodes, classifying and storing the data after cleaning and correlation by a stream processing framework and establishing a cross-database index, and by real-time comparison of the index and the baseline through the deviation rate formula to trigger early warning, and by customizing dashboards for different roles and generating intelligent suggestions, the full-link tracking and standardized management of process data are realized, the problem response efficiency is improved through accurate abnormality detection and role-based dashboards, and optimization directions are provided through machine learning, forming a closed loop of collection, processing, storage, monitoring, feedback and optimization, enhancing the real-time, accuracy and decision support capability of process monitoring.

[0110] In the present application, the visual dashboard of the monitoring feedback unit supports multi-dimensional drilling functions, and users can drill down from a global view to specific nodes to view execution logs, input and output data, and context information;

[0111] The look board further comprises an intelligent diagnosis module; the intelligent diagnosis module is further used for identifying whether a bottleneck exists in the process node, and specifically:

[0112] The actual execution time consumption of the current node is denoted as , the historical average execution time consumption is ;

[0113] The actual execution time consumption of the node at any time within a set time range before the current time is identified, and the actual execution time consumption within the set time range is calculated by using a variance formula to obtain time consumption fluctuation ;

[0114] The actual execution time consumption, the historical average execution time consumption and the time consumption fluctuation of the current node are weighted and calculated to obtain a bottleneck score B, and the formula is ; wherein d represents a preset dependency coefficient of the node, and d>1, the more downstream nodes the node depends on, the larger d is, 、 、 respectively represent the weights corresponding to the actual execution time consumption, the historical average execution time consumption and the time consumption fluctuation;

[0115] A preset bottleneck threshold is set, and if the bottleneck score is greater than the preset bottleneck threshold, an optimization scheme is pushed; the optimization scheme comprises a node merging suggestion, a resource allocation adjustment strategy and a rule engine parameter optimization direction.

[0116] The above formula is dimensionless after being calculated by standardization or the like, is generated based on a large amount of data simulation, and the preset parameters are set by a person skilled in the art according to actual conditions.

[0117] The embodiment can be realized by software, hardware, firmware or a combination thereof. If realized by software, it can be embodied as a computer program product, containing computer instructions or programs, and realizes corresponding processes or functions after loading and execution. The computer can be a general-purpose computer, a special-purpose computer, a network or other programmable devices, and the instructions can be stored on a readable medium or transmitted between media (such as wired or wireless mode). The readable medium includes computer accessible magnetic media, optical media, semiconductor media and the like.

[0118] It should be noted that the process numbers do not represent the execution order, and the order is determined by the function and logic. A person skilled in the art can select hardware or software to realize each example unit and algorithm step according to the application requirements, and the related implementation does not exceed the scope of the present application.

[0119] The system, device and method of the embodiment of the present application can be realized in other ways, and the unit division is only a logical function division. The actual division can be another division. Each functional unit can be integrated, exist alone or integrated with two or more.

[0120] If the functions are implemented in the form of software units and sold independently, they can be stored in readable media. The contribution part of the technical solutions of the present application can be embodied as a software product containing instructions to cause a computer device to execute corresponding method steps, and the storage medium includes a U disk, a mobile hard disk, a ROM, a RAM, a magnetic disk, an optical disk, etc.

[0121] The above merely provides a specific implementation of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An intelligent service handling process management and control platform, the management and control platform is built-in with a dynamic process engine module, characterized in that, The dynamic process engine module is used for analyzing a third-party platform business process structure, supporting visual configuration of a process node and a rule, realizing automatic execution and scheduling of the process, and monitoring and feeding back a process execution process; The dynamic process engine module comprises a process analysis unit, a configuration management unit, an execution scheduling unit and a monitoring and feeding back unit, wherein: The process analysis unit is used for analyzing a node relationship and a flow rule of a third-party platform business process, and specifically comprises: Obtaining business process information of a target third-party platform; identifying a platform type through a URL, page metadata or a specific identification element of the target platform, establishing a connection by using a corresponding communication protocol, collecting current version information of the target platform and comparing the current version information with historical process models cached locally to determine whether the process analysis result needs to be updated; if it is determined that the process analysis result needs to be updated, the subsequent analysis steps are re-executed and a new process model is generated to replace the historical cached process model; if it is determined that the process analysis result does not need to be updated, the historical cached process model is directly called; Then, a complete DOM structure of a target platform business process page is obtained through a browser automation tool, a node relationship tree is constructed, and key elements in the page are located and extracted; the key elements include form input components, operation buttons and process navigation components, and the visual features and layout relationships of non-standard UI components are identified by combining computer vision technology; The process analysis unit extracts static nodes based on the DOM structure and interactive elements, predicts dynamic nodes by monitoring page jump events, AJAX requests and URL change modes, and labels metadata for each node; The actual number of times that a dynamic node actually appears in a business scenario corresponding to the current process in historical data is recorded as an actual number, the total number of occurrences of the scenario in the historical data, the number of associated events triggered in the current process, and the average total number of associated events triggered in the scenario in the historical data; The prediction probability of the appearance of the dynamic node is calculated by using the actual number, the total number of occurrences, the number of associated events and the average total number; The metadata includes node types, mandatory fields, pre-operation, time limits and data interaction formats, and the metadata is associated with a platform knowledge graph to support cross-process node data reuse; The configuration management unit is used for supporting a user to customize a process node and execution logic through a visual interface; The execution scheduling unit is responsible for parallel processing and resource allocation of multiple processes, and the process comprises: Obtaining a process instance from a task queue, analyzing a process definition and constructing an execution task graph; Real-time monitoring of system resource states and availability of external dependent systems; Based on process priorities, node dependency relationships and resource states, a reinforcement learning algorithm is applied to generate a scheduling strategy; Creating an execution thread pool and allocating an independent thread to each node that can be executed in parallel, realizing node data synchronization and state interaction through a message queue; When a resource bottleneck or a timeout condition is detected, resource reallocation or task migration is automatically triggered; execution results of all nodes are collected and fed back to the monitoring and feeding back unit; The monitoring and feeding back unit is used for collecting process execution data in real time and generating a performance report.

2. The intelligent business process management and control platform according to claim 1, characterized in that, The flow parsing unit further comprises identifying abnormal interception components in the page and recording their triggering conditions and release methods when locating and extracting key elements in the page; the abnormal interception components include verification code pop-up windows, permission verification windows, and risk prompt interfaces.

3. The intelligent business process management and control platform according to claim 1, characterized in that, The configuration management unit supports users to customize flow nodes and execution logic through a visual interface, and the specific process comprises: loading an interactive visual canvas, presetting a basic flow template in the canvas, and displaying a node component library that can be dragged; triggering an attribute configuration panel after a user drags a node to the canvas, the panel supporting setting node basic information, business parameters, and execution conditions; defining node flow direction and flow rules including node relationship and priority association attributes through a connection tool, and providing loop logic configuration; defining node execution logic through a visual script editor; saving the configured flow as a specific version and setting an effective time, and synchronizing to the execution scheduling unit after publishing.

4. The intelligent business process management and control platform according to claim 3, characterized in that, When the configuration management unit defines node flow direction and flow rules through the connection tool, the node relationship configuration of the configuration management unit supports priority setting; when multiple parallel nodes meet execution conditions, the user can define node execution order, which includes sorting by creation time, manually specifying priority values, and intelligently analyzing the priority of parallel nodes, and the execution order is linked with the resource allocation mechanism of the execution scheduling unit, so that nodes with high priority can obtain computing resources and network bandwidth preferentially.

5. The intelligent business process management and control platform according to claim 1, characterized in that, The reinforcement learning algorithm adopted by the execution scheduling unit comprises a multi-dimensional reward function: obtaining node information including task completion time, resource utilization rate, execution success rate, and exception handling efficiency; analyzing the node information to obtain node analysis results including task completion time score, resource utilization rate score, execution success rate score, and exception handling efficiency score; constructing a multi-dimensional reward function through the node analysis results to calculate a comprehensive reward value; and optimizing the strategy through continuous learning of historical scheduling data.

6. The intelligent business process management and control platform according to claim 1, characterized in that, The process of the monitoring feedback unit for real-time collection of flow execution data and generation of performance reports comprises: implanting data collection probes in key nodes of the flow and system resources, and configuring collection parameters; adopting a distributed stream processing framework to clean, aggregate, and correlate collected data to generate standardized data streams; storing the processed data into a time series database and a relational database, and establishing cross-database indexes; based on an anomaly detection model, comparing indicators with baselines in real time, calculating the relative deviation rate between the indicators and the baselines; setting a deviation threshold, and when the relative deviation rate is greater than the preset deviation threshold, triggering a multi-level early warning mechanism; customizing visual dashboards based on user roles, and applying machine learning algorithms to generate optimization suggestions and predictive reports.

7. The intelligent business process management and control platform according to claim 6, characterized in that, The visual dashboards of the monitoring feedback unit support multi-dimensional drilling functions, and users can drill down from a global view to specific nodes to view execution logs, input and output data, and context information; the dashboards further comprise an intelligent diagnosis module; the intelligent diagnosis module is further configured to identify whether a bottleneck exists in a flow node, and the specific process comprises: obtaining actual execution time consumption of the current node and historical average execution time consumption; The actual execution time consumption of the node at any time within a set time range before the current time is identified, and the actual execution time consumption within the set time range is calculated using a variance formula to obtain time consumption fluctuation; The actual execution time consumption of the current node, the historical average execution time consumption, and the time consumption fluctuation are weighted and calculated to obtain a bottleneck score; A preset bottleneck threshold is set, and if the bottleneck score is greater than the preset bottleneck threshold, an optimization scheme is pushed; the optimization scheme includes node merging suggestions, resource allocation adjustment strategies, and rule engine parameter optimization directions.

Images