Cooperative management system for higher vocational education based on multi-source knowledge graph and agent

Abstract

This invention relates to the field of collaborative management technology in education, and discloses a collaborative management system for higher vocational education based on multi-source knowledge graphs and intelligent agents. The system includes: an ontology graph module for acquiring data from various business systems and generating versioned knowledge graph snapshots; a rule compilation module for compiling regulations and standards to form discrete constraint sets; an event impact module for determining impact subgraphs and filtering sub-constraint sets; a task decomposition and optimization module for decomposing tasks and selecting winning execution plans; a scheduling solution module for solving scheduling results based on constraint models; a job matching module for comparing student abilities with job requirements to obtain a set of job qualifications; and an evidence write-back module for constructing a directed acyclic graph of evidence and generating fingerprint hash values ​​to form a new versioned knowledge graph snapshot. This invention achieves unified data modeling, structured rule expression, automatic task execution, and full-process traceability in higher vocational education management.

Description

Technical Field

[0001] This invention belongs to the field of educational collaborative management technology, specifically involving a higher vocational education collaborative management system based on multi-source knowledge graphs and intelligent agents. Background Technology

[0002] With the continuous development of the higher vocational education system, the curriculum, job requirements, teaching resources, and management rules are characterized by multi-source heterogeneity and dynamic changes. Traditional management models face numerous difficulties in information integration, scheduling, job matching, and rule enforcement. On the one hand, higher vocational colleges have multiple business systems, such as academic affairs systems, practical training systems, and resource management systems. These systems have inconsistent data standards and formats, making it difficult to achieve cross-system collaborative management and consistent scheduling. On the other hand, laws, regulations, industry standards, and internal rules constitute an important institutional foundation for higher vocational education management, but their content is mostly unstructured text, making it difficult for information systems to directly recognize and automatically execute them. This results in the rule implementation process relying on manual operation, which is inefficient and prone to errors.

[0003] Furthermore, course scheduling and job allocation are influenced by multiple factors, including teaching resources, time arrangements, teacher scheduling, student abilities, and regulations. Traditional methods relying on static rules and manual adjustments are ill-suited to the demands of teaching management scenarios characterized by frequent unforeseen events, complex constraints, and rapid changes. They also struggle to ensure the overall coordination and accuracy of course scheduling and job matching. Against the backdrop of increasing demands for industry-education integration and vocational skills development, the rationality of job matching and the flexibility of course scheduling have become crucial factors restricting teaching quality and management efficiency. Summary of the Invention

[0004] This invention provides a collaborative management system for higher vocational education based on multi-source knowledge graphs and intelligent agents, which solves the technical problems in the management of higher vocational education in related technologies, such as the difficulty in unified modeling of multi-source heterogeneous data, the difficulty in structurally expressing management rules, the reliance on manual scheduling and job matching processes, low collaborative efficiency, and lack of process traceability.

[0005] This invention provides a collaborative management system for higher vocational education based on multi-source knowledge graphs and intelligent agents, comprising:

[0006] The ontology graph module is used to acquire data tables from various business systems, form a unified ontology and mapping table according to preset mapping rules, generate course coverage capability data, job requirement capability data, course prerequisite relationship data and course resource adaptation data, and obtain a versioned knowledge graph snapshot.

[0007] The rule compilation module is used to compile regulations, industry standards and school rules to obtain discrete constraint sets and minimum requirements for course competence.

[0008] The event impact module is used to access events that include event type and impact range, determine the minimum impact radius based on the action node to obtain the impact subgraph, and filter the sub-constraint set from the discrete constraint set;

[0009] The task decomposition and optimization module is used to decompose tasks into sub-tasks. Each agent submits proposals with feasibility evidence and impact metrics, and the winning execution plan is selected in a fixed priority order.

[0010] The scheduling solution module is used to construct a scheduling constraint model for scheduling and resource adjustment tasks based on sub-constraint sets, course prerequisite relationship data, course resource adaptation data, available teachers and teaching time slots, and solve the scheduling results according to preset hard constraints.

[0011] The job matching module is used to generate a set of student abilities based on the student's course set and course coverage ability data for job matching tasks, and compare it with the job requirement ability data to obtain a set of student job qualifications.

[0012] The evidence write-back module is used to construct a directed acyclic graph of evidence based on events, rule identifiers, and data fragments, generate fingerprint hash values, and write back the evidence directed acyclic graph and fingerprint hash values ​​as a snapshot of the new version of the knowledge graph based on the course scheduling results, the set of student job qualifications, and the evidence directed acyclic graph and fingerprint hash values.

[0013] Furthermore, the process of generating versioned knowledge graph snapshots includes:

[0014] Step 11: Form a unified ontology and mapping table according to the preset mapping rules, and map the source fields in the data table to course entities, ability entities, job entities and resource entities and their corresponding relationships. The preset mapping rules include field name mapping rules, field type constraint rules and entity relationship matching rules.

[0015] Step 12: Perform encoding unification, enumeration value normalization, and primary-foreign key consistency verification on the mapped data. Generate course coverage capability data, job requirement capability data, course prerequisite relationship data, and course resource adaptation data based on the mapping relationship. When there is a coverage relationship between courses and capabilities, a requirement relationship between jobs and capabilities, a prerequisite relationship between courses, or an adaptation relationship between courses and resources, the corresponding position is assigned a value of 1; otherwise, it is assigned a value of 0.

[0016] Step 13: Write the course coverage capability data, job requirement capability data, course prerequisite relationship data, and course resource adaptation data into the knowledge graph node set and relationship set, establish a two-way index structure of matrix index and graph node, so that the matrix coordinates correspond one-to-one with the node identifier, and generate a version identifier with timestamp and serial number, and solidify the above data and index structure into a versioned knowledge graph snapshot of the corresponding version.

[0017] Furthermore, the versioned knowledge graph snapshot undergoes a multi-source data consistency maintenance process before generation, including:

[0018] Step 21: Detect conflicts in entity field values ​​from different data sources. When multiple candidate values ​​exist for the same field of the same entity, determine the field value to retain according to the preset field priority table, remove other field values, and form a consistent set of field values.

[0019] Step 22: Perform conflict detection on the relationships between course entities, ability entities, job entities and resource entities according to the start node identifier, relationship type and end node identifier. When duplicate relationships exist, retain the unique relationship entry according to the relationship arbitration rules.

[0020] Step 23: Write the consistent set of field values ​​and the unique relation entries into the node set and relation set of the knowledge graph, and keep the node identifier and relation identifier unique to form a consistent entity and relation structure.

[0021] Furthermore, regulations, industry standards, and school rules are compiled to obtain a discrete constraint set and a minimum requirement for course competency, including:

[0022] Step 31: Based on the preset clause parsing rule base, perform sentence-level rule matching on the texts of regulations, industry standards and internal rules, and parse the matching results into a set of triplets consisting of preconditions, constraints and threshold parameters, and use the clause number as a unique identifier.

[0023] Step 32: Based on the type of constraint, divide the set of triples into course attainment constraints, ability coverage constraints, job suitability constraints, and resource suitability constraints, and perform numerical normalization and unit standardization on the threshold parameters to obtain a standardized constraint set.

[0024] Step 33: Map the standardized constraint set to the matrix positions corresponding to the course coverage capability data, job requirement capability data, course prerequisite relationship data, and course resource adaptation data through the index table. When the constraint is valid, assign a value of 1 to the corresponding position; otherwise, assign a value of 0 to form a discrete constraint set and a course capability compliance lower limit matrix.

[0025] Step 34: Generate version identifiers using timestamps and serial numbers. Establish bidirectional indexes between the discrete constraint set and the course competency lower limit matrix and the courses, competencies, positions, and resource nodes in the knowledge graph according to indexing rules, forming versioned constraint objects that correspond one-to-one with the versioned knowledge graph snapshots.

[0026] Furthermore, the determination of subgraphs and sub-constraint sets includes:

[0027] Step 41: Access events that include event type, event source, event occurrence time and scope of impact. Based on the keyword matching and entity identifier correspondence between node attribute fields and event scope of impact, determine matching nodes through a preset priority strategy, and project the matching results to the knowledge graph node space to obtain the set of starting nodes corresponding to the event.

[0028] Step 42: Calculate the shortest path distance from the starting node to other nodes based on the topology of the knowledge graph, and compare it with the preset radius threshold corresponding to the event type. Take the radius value that meets the minimum coverage condition as the minimum influence radius, and use the shortest path search algorithm to retain the nodes and their associated edges whose distance is not greater than the minimum influence radius to form an influence subgraph.

[0029] Step 43: Using the unique identifier of the node in the influence subgraph as an index, retrieve the constraint terms corresponding to the node in the discrete constraint set, remove duplicate constraints, and filter to obtain the constraints corresponding to the influence subgraph. Figure 1 A corresponding set of sub-constraints.

[0030] Furthermore, the process of determining the influence subgraph includes dynamic update and inheritance steps. When multiple events occur consecutively, the set of starting nodes of the new event is merged with the set of nodes in the stored event influence subgraph to form a candidate node set. The candidate node set is then subjected to range calculation based on shortest path search. Nodes exceeding the preset influence radius threshold are removed to obtain the updated influence node set. The updated influence node set is then mapped to a discrete constraint set to replace the original sub-constraint set, forming the updated sub-constraint set.

[0031] Furthermore, the process for determining the winning bid implementation plan includes:

[0032] Step 51: Decompose the overall task in the order of time dimension first, resource dimension second, and object dimension last. When there is a conflict in the dimensions, the time dimension division result takes precedence to form a unique set of subtasks. For each subtask, register the task type, required resources, execution time period and affected objects to generate a subtask attribute table.

[0033] Step 52: Each agent generates a proposal for the subtask, which includes feasibility evidence, impact measurement and solution summary. The proposals are organized into a standardized triplet structure, and a unique index table is constructed using the subtask identifier and agent identifier to form a standardized proposal set.

[0034] Step 53: Set priority order, compare each item in the standardized proposal set, and compare the next priority item in turn when there are multiple proposals in the same order until the winning proposal is uniquely determined. Generate the winning execution plan, which includes sub-task identifier, execution time period, responsible entity, allocated resources and execution method, and register it in the scheduling table. The priority order is feasibility, scope of impact, resource consumption and evidence chain length.

[0035] Furthermore, the process of generating the course scheduling results includes:

[0036] Step 61: Extract the sub-constraint set, course prerequisite relationship data, course resource adaptation data, and available teacher and teaching time slot data from the versioned knowledge graph snapshot. Define the scheduling variable set for courses, teachers, and time windows. The scheduling variable is indexed by the triple of course identifier, teacher identifier, and time window identifier. The value is 1 when the corresponding course is scheduled by a specified teacher in the specified time window, and 0 otherwise.

[0037] Step 62: Establish a scheduling constraint model based on the sub-constraint set and course relationship data. The preset hard constraints include: class hour satisfaction constraint, resource mutual exclusion constraint, teacher mutual exclusion constraint, prerequisite order constraint, and resource adaptation constraint. Construct an objective function by weighted summation of unsatisfied constraint items. Specifically, by traversing all combinations of courses, teachers, and time windows, determine whether the corresponding scheduling variable satisfies the constraints. When the scheduling variable takes a value of 0, the corresponding weight is accumulated; when the variable takes a value of 1, no accumulation is performed. Summate the weights of all combinations to obtain the weighted value of unsatisfied constraints, and minimize this weighted value as the solution objective function.

[0038] Step 63: Solve the course scheduling constraint model. When a feasible solution exists, extract the triplet of course, teacher and time window with a value of 1 and map it to the course scheduling result table.

[0039] Furthermore, the process of generating the student job qualification set includes:

[0040] Step 71: Extract the student course set and course coverage ability data from the versioned knowledge graph snapshot. According to the one-to-many mapping relationship between courses and abilities, take the union of the ability items corresponding to the student courses and remove duplicates to form the student ability set.

[0041] Step 72: Extract job requirement capability data from the versioned knowledge graph snapshot, and uniformly encode the student capability set and the job requirement capability set according to the preset capability dimension standardization rules, mapping capability items from different sources to unique capability identifiers;

[0042] Step 73: Traverse the set of job requirement capabilities. For each job, determine whether the set of job requirement capabilities is completely included in the set of student capabilities. When the determination is true, add the corresponding job identifier to the set of student job qualifications to form the set of student job qualifications.

[0043] Furthermore, the process of generating the new version of the knowledge graph snapshot includes:

[0044] Step 81: Map events, rule identifiers, and data fragments to evidence nodes respectively; establish directed edges according to the event triggering order and rule dependency direction; and detect loops through topological sorting to generate a directed acyclic graph of evidence.

[0045] Step 82: Read the node identifiers, corresponding outgoing edge identifiers, and timestamps of the directed acyclic graph of evidence in order of topological sorting, concatenate them into a hash input sequence using a fixed format string, and calculate the fingerprint hash value using a hash function;

[0046] Step 83: The scheduling results, student job qualification set, directed acyclic graph of evidence, and fingerprint hash value are combined to form an update object set, which is then bound to a unique version identifier and written into the knowledge graph to generate a new versioned knowledge graph snapshot.

[0047] The beneficial effects of this invention are as follows: By introducing a multi-source knowledge graph and an intelligent agent scheduling mechanism, this invention achieves unified modeling, rule-structured expression, and collaborative scheduling of multi-source heterogeneous data in higher vocational education management. By constructing a knowledge graph of courses, skills, positions, and resources, it enables data association and unified management across business systems; through rule compilation and constraint solving, it automates and refines the scheduling and job matching process; through event-driven and task decomposition mechanisms, it improves the flexibility and response speed of the scheduling process; and through versioned snapshots using evidence chains and hash fingerprints, it ensures the traceability and consistency of task execution and data update processes. Overall, this invention reduces manual intervention, improves the efficiency and accuracy of scheduling and job matching, and enhances the synergy and stability of education management. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the modules of the higher vocational education collaborative management system based on multi-source knowledge graphs and intelligent agents of the present invention. Detailed Implementation

[0049] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0050] like Figure 1 As shown, the vocational education collaborative management system based on multi-source knowledge graphs and intelligent agents includes:

[0051] Ontology Graph Module 1 is used to obtain data tables from various business systems, form a unified ontology and mapping table according to preset mapping rules, generate course coverage capability data, job requirement capability data, course prerequisite relationship data and course resource adaptation data, and obtain a versioned knowledge graph snapshot.

[0052] Rule compilation module 2 is used to compile rules from regulations, industry standards and school rules to obtain discrete constraint sets and lower limits for course competency.

[0053] Event Impact Module 3 is used to access events that include event type and impact range, determine the minimum impact radius based on the action node to obtain the impact subgraph, and filter the sub-constraint set from the discrete constraint set;

[0054] Task decomposition and optimization module 4 is used to decompose tasks into sub-tasks. Each agent submits proposals with feasibility evidence and impact metrics, and selects the winning execution plan in a fixed priority order.

[0055] The scheduling solution module 5 is used to construct a scheduling constraint model based on sub-constraint sets, course prerequisite relationship data, course resource adaptation data, available teachers and teaching time slots for scheduling and resource adjustment tasks, and solves the scheduling results according to the preset hard constraints.

[0056] The job matching module 6 is used to form a student ability set based on the student's course set and course coverage ability data for job matching tasks, and compare it with the job requirement ability data to obtain a student job qualification set.

[0057] The evidence write-back module 7 is used to construct a directed acyclic graph of evidence based on events, rule identifiers and data fragments and generate fingerprint hash values. Based on the scheduling results and the set of student job qualifications, the directed acyclic graph of evidence and fingerprint hash values ​​are written back as a snapshot of the new version of the knowledge graph.

[0058] In one embodiment of the present invention, the process of generating a versioned knowledge graph snapshot includes:

[0059] Step 11: Form a unified ontology and mapping table according to the preset mapping rules, and map the source fields in the data table to course entities, ability entities, job entities and resource entities and their corresponding relationships. The preset mapping rules include field name mapping rules, field type constraint rules and entity relationship matching rules.

[0060] Specifically, field name mapping rules are used to map fields with different names but the same meaning across different systems to a one-to-one correspondence. For example, course code and course number are uniformly mapped to the course identifier field. Field type constraint rules are used to ensure the consistency of the data types of each field after mapping, such as converting character, integer, and enumeration types to standardized data types. Entity relationship matching rules are used to determine the type of entity relationship between fields. For example, the association between courses and skills is mapped to a coverage relationship, the association between positions and skills is mapped to a requirement relationship, the association between courses is mapped to a prerequisite relationship, and the association between courses and resources is mapped to an adaptation relationship.

[0061] Step 12: Perform encoding unification, enumeration value normalization, and primary-foreign key consistency verification on the mapped data. Generate course coverage capability data, job requirement capability data, course prerequisite relationship data, and course resource adaptation data based on the mapping relationship. When there is a coverage relationship between courses and capabilities, a requirement relationship between jobs and capabilities, a prerequisite relationship between courses, or an adaptation relationship between courses and resources, the corresponding position is assigned a value of 1; otherwise, it is assigned a value of 0.

[0062] Specifically, the encoding unification refers to converting field values ​​from different sources into a unified encoding system, such as unifying course codes into internal standard codes; enumeration value normalization refers to mapping data enumeration items from different sources to a unified set of standard enumeration values, avoiding multiple expressions for the same enumeration item; primary and foreign key consistency verification is used to ensure the correctness of reference relationships between data, ensuring that the connection between entities and relationships is legal and traceable; the course coverage capability data is used to describe the one-to-many mapping relationship between courses and capability items, reflecting the capability dimensions and capability item identifiers that each course can cover; the job requirement capability data is used to describe the one-to-many mapping relationship between jobs and capability items, reflecting the capability requirement structure of the job for applicants; the course prerequisite relationship data is used to describe the prerequisite dependency relationship between courses, defining the teaching sequence constraints of courses; and the course resource adaptation data is used to describe the matching relationship between courses and teaching resources, used for scheduling resource allocation and conflict detection.

[0063] Step 13 involves writing course coverage capability data, job requirement capability data, course prerequisite relationship data, and course resource adaptation data into the knowledge graph node set and relationship set. A bidirectional index structure of matrix index and graph nodes is established, ensuring a one-to-one correspondence between matrix coordinates and node identifiers. This achieves precise mapping between matrix data and graph entities. Version identifiers are generated using timestamps and sequence numbers, solidifying the above data and index structure into a versioned knowledge graph snapshot. This enables time-series management of multi-source knowledge graphs. The versioned knowledge graph snapshot refers to the fixed sealing of all node sets, relationship sets, and index structures in the knowledge graph at a specific point in time, forming a static copy with a unique version identifier. In the knowledge graph, nodes represent courses, capabilities, jobs, or resource entities, and edges represent the coverage, adaptation, or dependency relationships between them.

[0064] This embodiment constructs a unified ontology structure based on a multi-source knowledge graph, realizing the structured integration and version management of data from multiple heterogeneous business systems in the higher vocational education scenario. This enables key information such as course capabilities, job requirements, resource adaptation, and course prerequisites to be accurately expressed and indexed in a unified graph framework, improving the data consistency and traceability of core business processes such as course scheduling and job matching.

[0065] In one embodiment of the present invention, in order to ensure the structural integrity and logical consistency of the versioned knowledge graph snapshot under multi-source data fusion conditions, the versioned knowledge graph snapshot performs a multi-source data consistency maintenance process before generation, including:

[0066] Step 21: Conflict detection is performed on entity field values ​​from different data sources. When multiple candidate values ​​exist for the same field of the same entity, it indicates a conflict in the definition of that field across multiple data sources. The retained field values ​​are determined according to a preset field priority table, while other field values ​​are eliminated to form a consistent set of field values. Specifically, entity field values ​​refer to the attribute values ​​corresponding to nodes such as course entities, ability entities, job entities, and resource entities in the knowledge graph, such as course codes, ability identifiers, job identifiers, and resource names. The preset field priority table determines the retention priority of field values ​​based on the credibility, timeliness, and importance of the data source.

[0067] Step 22: Conflict detection is performed on the relationships between course entities, capability entities, job entities, and resource entities according to the starting node identifier, relationship type, and ending node identifier. When duplicate relationships exist, a unique relationship entry is retained according to the relationship arbitration rules, ensuring that each pair of entities has only one unique record in the knowledge graph for a specific relationship type. The relationship arbitration rules include a determination mechanism based on data source priority, timestamp, and relationship weight, i.e., retaining a unique record according to the order of data source priority, timestamp, and relationship weight.

[0068] Step 23 involves writing the consistent set of field values ​​and unique relation entries into the node set and relation set of the knowledge graph, ensuring the uniqueness of node and relation identifiers to form a consistent entity and relation structure. The node set refers to the set of all entity nodes, and the relation set refers to the set of directed associations between nodes. During the writing process, the uniqueness of all node and relation identifiers is checked to ensure that there are no entities or relations with duplicate identifiers within the same knowledge graph.

[0069] This embodiment effectively solves the field conflict and relationship duplication problems that may occur after the fusion of multi-source heterogeneous data in the context of higher vocational education management by introducing a multi-source data consistency maintenance process before the generation of versioned knowledge graph snapshots, and ensures the uniqueness and consistency of core educational elements such as courses, abilities, positions and resources in the knowledge graph.

[0070] In one embodiment of the present invention, regulations, industry standards, and school rules are compiled into discrete constraint sets and minimum requirements for course competency, including:

[0071] Step 31: Based on the preset clause parsing rule base, perform sentence-level rule matching on the texts of regulations, industry standards, and school rules. Parse the matching results into a set of triplets consisting of preconditions, constraints, and threshold parameters, and use the clause number as a unique identifier. Specifically, the preset clause parsing rule base is a semantic parsing set set set for the field of higher vocational education management, used to identify preconditions, constraints, and corresponding threshold parameters in the text. The threshold parameters refer to the quantitative regulations on course hours, ability coverage ratio, job ability requirements, and resource utilization limits.

[0072] Step 32: Based on the type of constraint, the set of triples is divided into course achievement constraints, competency coverage constraints, job suitability constraints, and resource suitability constraints. The threshold parameters are then normalized and standardized in terms of both numerical and unit, resulting in a standardized constraint set. Specifically, rules related to course achievement are categorized into course achievement constraints, rules related to competency coverage constraints, rules related to job requirements constraints, and rules related to teaching resource allocation constraints.

[0073] Step 33: Map the standardized constraint set to the matrix positions corresponding to course coverage capability data, job requirement capability data, course prerequisite relationship data, and course resource adaptation data using an index table. When a constraint is valid, assign a value of 1 to the corresponding position; otherwise, assign a value of 0, forming a discrete constraint set and a course capability compliance lower limit matrix. The index table is a mapping structure between constraint objects and entity nodes and relationship nodes in the knowledge graph, used to determine the correspondence between each constraint and the corresponding course, capability, job, and resource in the knowledge graph. The discrete constraint set refers to a computable constraint matrix composed of multiple 0 or 1 values, used as constraint input in subsequent course scheduling and matching solutions. The course capability compliance lower limit describes the minimum standard of course capability coverage under the conditions of meeting teaching and management specifications.

[0074] Step 34: Generate version identifiers using timestamps and sequence numbers. Establish bidirectional indexes between the discrete constraint set and the course competency lower limit matrix and the courses, competencies, positions, and resource nodes in the knowledge graph according to indexing rules. This forms versioned constraint objects that correspond one-to-one with the versioned knowledge graph snapshots, ensuring the traceability and verifiability of the scheduling and matching process.

[0075] This embodiment compiles regulations, industry standards, and school rules, transforming unstructured text information into a discrete set of constraints with clear semantics and numerical standards, thus establishing a connection between management rules and intelligent agent task decision-making. In the scenario of collaborative management in higher vocational education, it improves the automation level of rule execution and avoids ambiguity caused by manual interpretation of regulations.

[0076] In one embodiment of the present invention, events including event type and impact range are accessed, an impact subgraph is obtained by determining the minimum impact radius based on the action node, and a sub-constraint set is filtered from the discrete constraint set, including:

[0077] Step 41: Access events that include event type, event source, event occurrence time and scope of impact. Based on the keyword matching and entity identifier correspondence between node attribute fields and event scope of impact, determine matching nodes through a preset priority strategy, and project the matching results to the knowledge graph node space to obtain the set of starting nodes corresponding to the event.

[0078] Specifically, the event refers to a change in status that affects course scheduling, job matching, or allocation of teaching resources during the management of higher vocational education, such as teacher leave, course adjustments, changes in the use of teaching resources, or updates to job requirements; the event types include resource-related events, course-related events, and job-related events; the event source is used to identify the system or entity that triggered the event; the event occurrence time is used to record the time when the event takes effect; the event impact scope is used to describe the courses, skills, positions, or resource objects that the event may affect;

[0079] String matching and synonym expansion matching are performed on the attribute fields of knowledge graph nodes. For objects containing unique identifiers, nodes are directly matched to form a preliminary candidate node set. Unique identifiers include course codes, job IDs, and resource codes. Field matching scores are calculated by weighting three categories of coefficients: field matching, event type relevance, and timeliness. The nodes are sorted by field matching scores, and the nodes with the highest scores are selected first to obtain the starting node set corresponding to the event. The preset priority strategy is to select the nodes with the highest scores first. When there are multiple nodes with the same score, the field matching degree, event type relevance, and timeliness are compared in turn.

[0080] Step 42: Calculate the shortest path distance from the starting node to other nodes based on the topology of the knowledge graph, and compare it with the preset radius threshold corresponding to the event type. Take the radius value that meets the minimum coverage condition as the minimum influence radius, and use the shortest path search algorithm to retain the nodes and their associated edges whose distance is not greater than the minimum influence radius to form an influence subgraph; the influence subgraph is a subset of nodes and relationships within the influence range of the current event.

[0081] Step 43: Using the unique identifier of the node in the influence subgraph as an index, retrieve the constraint terms corresponding to the node in the discrete constraint set, remove duplicate constraints, and filter to obtain the constraints corresponding to the influence subgraph. Figure 1 A corresponding set of sub-constraints. Specifically, each constraint item is linked to a specific course, ability, position, or resource entity in the knowledge graph through a bidirectional index. This index relationship allows for the rapid location of constraints related to the influencing nodes.

[0082] This embodiment introduces an automatic filtering mechanism for event-affected subgraphs and sub-constraint sets, enabling dynamic response to events in a multi-source knowledge graph environment. It can accurately locate the starting node of an event in the knowledge graph, automatically calculate the scope of influence based on the topology, and quickly filter the constraint set related to the affected object, thereby reducing the computational scale and the complexity of rule parsing, and enhancing the flexibility and scalability of the higher vocational education collaborative management system.

[0083] In one embodiment of the present invention, to adapt to the actual situation of continuous occurrence and dynamic changes in the scope of influence of events in the collaborative management scenario of higher vocational education, a dynamic update and inheritance mechanism is introduced in the process of determining the event influence subgraph. When multiple events occur consecutively, the set of starting nodes of the new event is merged with the set of nodes in the stored event influence subgraph to form a candidate node set; and a range calculation based on shortest path search is performed on the candidate node set, with a preset influence radius threshold as the boundary, and nodes exceeding the threshold are eliminated to obtain an updated influence node set; the updated influence node set is mapped to a discrete constraint set, replacing the original sub-constraint set to form an updated sub-constraint set.

[0084] Among them, the event influence subgraph refers to a subset of nodes and their associated edges determined by the influence radius calculation of the events that have occurred, representing the comprehensive range of the current events on the knowledge graph structure; the preset influence radius threshold is set according to the event type.

[0085] This embodiment realizes the dynamic adjustment of the scope of event impact and the automatic inheritance and update of constraint sets based on multi-source knowledge graphs; in the scenario of collaborative management of higher vocational education, the system can quickly respond to the impact changes brought about by the superposition of multiple events, and enhance the dynamic adaptability and decision-making efficiency of the system.

[0086] In one embodiment of the present invention, the process of determining the winning bid execution plan includes:

[0087] Step 51: Decompose the overall task in the order of time dimension first, resource dimension second, and object dimension last. When there is a conflict in dimensions, the time dimension division result takes precedence, forming a unique set of subtasks. For each subtask, register the task type, required resources, execution time period, and affected objects to generate a subtask attribute table. The overall task refers to the scheduling requirement generated by the combination of event triggering, rule constraints, and scheduling objectives, such as an adjustment operation involving multiple courses, multiple resources, or positions. The object dimension refers to the course, ability, or position entity involved in the task.

[0088] Step 52: Each agent generates a proposal for each subtask, containing feasibility evidence, impact metric, and solution summary. The proposals are organized into a standardized triplet structure, and a unique index table is constructed using the subtask identifier and agent identifier to form a standardized proposal set. The agent refers to a computational unit with rule recognition and constraint solving capabilities, capable of generating candidate solutions for specific tasks based on knowledge graphs and constraint sets. Feasibility evidence is the basis for an agent to determine whether it possesses the conditions to execute the task, such as resource availability and time conflict verification results. Impact metric quantifies the degree of impact of task execution on the overall system, such as the number of students involved in course adjustments or the amount of resource changes. Solution summary refers to the execution method and main parameters of the proposal.

[0089] Step 53: Set priority order, compare each standardized proposal in the set, and when multiple proposals are tied, compare the next priority item in turn until a winning proposal is uniquely determined. Generate a winning execution plan that includes sub-task identifiers, execution time periods, responsible parties, allocated resources, and execution methods, and register it in the scheduling table. The priority order is feasibility, scope of impact, resource consumption, and evidence chain length. The scheduling table is used to drive actual execution operations such as scheduling, resource adjustments, or job matching.

[0090] Through the above steps, this embodiment realizes the determination of a unique winning execution plan by using an intelligent agent proposal competition and priority ranking decision mechanism supported by a multi-source knowledge graph. It can quickly complete scheduling decisions in complex educational management scenarios with multiple tasks running in parallel and limited resources, ensuring the uniqueness of task decomposition, the standardization of proposal comparison, and the certainty of winning the bid; effectively improving the dynamic response capability and intelligent decision-making level of the collaborative management system.

[0091] In one embodiment of the present invention, the process of generating the course scheduling result includes:

[0092] Step 61: Extract the sub-constraint set, course prerequisite relationship data, course resource adaptation data, and available teacher and teaching time slot data from the versioned knowledge graph snapshot. Define a set of scheduling variables for courses, teachers, and time windows. The scheduling variables are indexed by a triplet of course identifier, teacher identifier, and time window identifier. The value is 1 when the corresponding course is scheduled by a specified teacher in a specified time window, and 0 otherwise. The available teacher and teaching time slot data are used to record the available teachers and their available time ranges.

[0093] Step 62: Establish a scheduling constraint model based on the sub-constraint set and course relationship data. The preset hard constraints include: time-satisfaction constraints, resource mutual exclusion constraints, teacher mutual exclusion constraints, prerequisite order constraints, and resource adaptation constraints. An objective function is constructed by weighted summation of unsatisfied constraints. Specifically, by traversing all combinations of courses, teachers, and time windows, it is determined whether the corresponding scheduling variables satisfy the constraints. When a scheduling variable is 0, the corresponding weight is accumulated; when the variable is 1, no accumulation occurs. The weights of all combinations are summed to obtain the weighted value of unsatisfied constraints, and minimizing this weighted value is used as the objective function. Specifically, the time-satisfaction constraint ensures that the total scheduled time of courses is not less than the teaching plan requirements; the resource mutual exclusion constraint ensures that the same resource can only be allocated to one teaching activity at the same time; the teacher mutual exclusion constraint prevents teachers from being assigned to multiple courses in the same time period; the prerequisite order constraint ensures that the scheduling order of courses satisfies prerequisite logic; and the resource adaptation constraint ensures that courses match the required resource types. The formula for calculating the objective function is:

[0094] Where F represents the value of the objective function. Let C represent the course index, t represent the time window index, T represent the time window set, r represent the teacher index, R represent the teacher set, k represent the constraint type index, and K represent the preset number of hard constraints. This represents the constraint weight of the k-th constraint type. Represents the course scheduling variable. This represents the constraint violation indicator function. That is, when the corresponding k-th type constraint is not satisfied, the value of the constraint violation indicator function is 1, and otherwise it is 0.

[0095] Step 63: Solve the scheduling constraint model. When a feasible solution exists, extract the triplet of course, teacher, and time window with a value of 1, and map it to a scheduling result table. This scheduling result table records information such as course scheduling time, responsible teacher, and resources used.

[0096] Through the above steps, this embodiment can quickly generate reasonable scheduling results under multiple constraints of teaching plan, teacher availability and resource allocation, providing deterministic output for the autonomous scheduling and dynamic adjustment of the intelligent agent, and improving the flexibility and efficiency of teaching plan implementation.

[0097] In one embodiment of the present invention, the process of generating the student job qualification set includes:

[0098] Step 71: Extract the student course set and course coverage ability data from the versioned knowledge graph snapshot. According to the one-to-many mapping relationship between courses and abilities, take the union of the ability items corresponding to the student courses and remove duplicates to form the student ability set. This student ability set reflects the ability structure that students acquire during the course learning process.

[0099] Step 72: Extract job requirement capability data from the versioned knowledge graph snapshot, and uniformly encode the student capability set and the job requirement capability set according to the preset capability dimension standardization rules, mapping capability items from different sources to unique capability identifiers; the preset capability dimension standardization rules include synonym merging, encoding mapping and hierarchical alignment of capability items. This step ensures the semantic consistency and comparability of the student and job capability sets.

[0100] Step 73: Iterate through the set of job requirement capabilities. For each job, determine whether the set of job requirement capabilities is completely included in the student's capability set. If the determination is true, it means that the student possesses all the capabilities required for the job. Add the corresponding job identifier to the student's job qualification set to form the student's job qualification set. The job qualification set refers to the set of all jobs that the student can be competent for at the current capability level, and can be used for subsequent business processing such as internship recommendation, job matching, and job scheduling.

[0101] Through the above steps, this embodiment utilizes the association structure between courses, abilities, and positions in a multi-source knowledge graph to achieve automatic derivation of students' abilities and standardized matching of job requirements. Compared with traditional manual screening or static rule matching methods, it has higher accuracy and adaptability. In the scenario of collaborative management of higher vocational education, this mechanism can update the job matching results in real time according to the dynamic changes in students' course learning, providing deterministic input for agent-driven job recommendation, job adjustment, and practical training allocation, thereby improving the intelligence level and scheduling efficiency of job matching.

[0102] In one embodiment of the present invention, the process of generating a new versioned knowledge graph snapshot includes:

[0103] Step 81: Map events, rule identifiers, and data fragments to evidence nodes respectively. Establish directed edges based on the event triggering order and rule dependency direction, and detect loops using topological sorting to generate a directed acyclic graph (DAG) of evidence. The rule identifier refers to a unique rule code corresponding to constraint logic or legal clauses; the data fragment refers to a data unit referenced or updated during task execution. Through the above mapping process, a directed graph is formed where nodes represent evidence sources and edges represent dependencies or causal relationships. The graph structure is checked using a topological sorting algorithm. If a loop exists, a logical conflict is indicated and writing is stopped; if no loop exists, a DAG of evidence is generated. This DAG of evidence completely records the logical relationships between event triggering, rule dependencies, and data changes.

[0104] Step 82: Read the node identifiers, corresponding outgoing edge identifiers, and timestamps of the directed acyclic graph of evidence in order of topological sorting. Concatenate them into a hash input sequence using a fixed format string and calculate the fingerprint hash value using a hash function. This fingerprint hash value can perform consistency verification and anti-tampering verification on the knowledge graph update chain.

[0105] Step 83 involves constructing an update object set from the course scheduling results, student job qualification set, directed acyclic graph of evidence, and fingerprint hash value, binding it with a unique version identifier, and writing it into the knowledge graph to generate a new versioned knowledge graph snapshot. The unique version identifier is a version number generated by combining a timestamp and a sequence number; this step ensures that the knowledge graph version is verifiable, traceable, and comparable.

[0106] Through the above steps, this invention realizes a knowledge graph version solidification mechanism based on evidence chains and fingerprint hashes. Compared with the traditional snapshot storage method, it not only records the result data, but also performs structured expression and unique identifier binding of the process logic. In the scenario of collaborative management of higher vocational education, the system can automatically generate the corresponding directed acyclic graph of evidence and hash value, so that different versions of teaching and management status have verifiability and traceability, thereby improving the credibility and controllability of the system.

[0107] It should be noted that the interval and threshold sizes are set for ease of comparison. The size of the threshold depends on the amount of sample data and the base number set by those skilled in the art for each set of sample data, as long as it does not affect the proportional relationship between the parameter and the quantized value. Furthermore, the above formulas are all dimensionless calculations, and the formulas are derived from software simulations using a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0108] The embodiments of the present invention have been described above, but the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of the present embodiments, all of which are within the protection scope of the present embodiments.

Claims

1. A collaborative management system for higher vocational education based on multi-source knowledge graphs and intelligent agents, characterized in that: include: The ontology graph module is used to acquire data tables from various business systems, form a unified ontology and mapping table according to preset mapping rules, generate course coverage capability data, job requirement capability data, course prerequisite relationship data and course resource adaptation data, and obtain a versioned knowledge graph snapshot. The rule compilation module is used to compile regulations, industry standards and school rules to obtain discrete constraint sets and minimum requirements for course competence. The event impact module is used to input events containing event type and impact range, determine the minimum impact radius based on the affected nodes to obtain an impact subgraph, and filter the sub-constraint set from the discrete constraint set, including: Step 41: Access events that include event type, event source, event occurrence time and scope of impact. Based on the keyword matching and entity identifier correspondence between node attribute fields and event scope of impact, determine matching nodes through a preset priority strategy, and project the matching results to the knowledge graph node space to obtain the set of starting nodes corresponding to the event. Step 42: Calculate the shortest path distance from the starting node to other nodes based on the topology of the knowledge graph, and compare it with the preset radius threshold corresponding to the event type. Take the radius value that meets the minimum coverage condition as the minimum influence radius, and use the shortest path search algorithm to retain the nodes and their associated edges whose distance is not greater than the minimum influence radius to form an influence subgraph. Step 43: Using the unique identifier of the node in the influence subgraph as an index, retrieve the constraint item corresponding to the node in the discrete constraint set, remove duplicate constraints, and filter to obtain a sub-constraint set that corresponds one-to-one with the influence subgraph. The process of determining the influence subgraph includes dynamic updating and inheritance steps. When multiple events occur consecutively, the set of starting nodes of the new event is merged with the set of nodes in the stored influence subgraph to form a candidate node set. Then, a range calculation based on shortest path search is performed on the candidate node set, and nodes exceeding the preset influence radius threshold are removed to obtain the updated influence node set. The updated influence node set is then mapped to a discrete constraint set, replacing the original sub-constraint set to form the updated sub-constraint set. The task decomposition and optimization module is used to decompose tasks into sub-tasks. Each agent submits proposals with feasibility evidence and impact metrics, and the winning execution plan is selected in a fixed priority order. The scheduling solution module is used to construct a scheduling constraint model based on sub-constraint sets, course prerequisite relationship data, course resource adaptation data, available teachers and teaching time slots for scheduling and resource adjustment tasks, and solve the scheduling results according to the preset hard constraints. The job matching module is used to generate a set of student abilities based on the student's course set and course coverage ability data for job matching tasks, and compare it with the job requirement ability data to obtain a set of student job qualifications. The evidence write-back module is used to construct a directed acyclic graph of evidence based on events, rule identifiers, and data fragments, generate fingerprint hash values, and write back a new version of the knowledge graph snapshot based on the scheduling results, student job qualification sets, the directed acyclic graph of evidence, and fingerprint hash values. This includes: Step 81: Map events, rule identifiers, and data fragments to evidence nodes respectively; establish directed edges according to the event triggering order and rule dependency direction; and detect loops through topological sorting to generate a directed acyclic graph of evidence. Step 82: Read the node identifiers, corresponding outgoing edge identifiers, and timestamps of the directed acyclic graph of evidence in order of topological sorting, concatenate them into a hash input sequence using a fixed format string, and calculate the fingerprint hash value using a hash function; Step 83: The scheduling results, student job qualification set, directed acyclic graph of evidence, and fingerprint hash value are combined to form an update object set, which is then bound to a unique version identifier and written into the knowledge graph to generate a new versioned knowledge graph snapshot.

2. The higher vocational education collaborative management system based on multi-source knowledge graphs and intelligent agents according to claim 1, characterized in that, The process of generating a versioned knowledge graph snapshot includes: Step 11: Form a unified ontology and mapping table according to the preset mapping rules, and map the source fields in the data table to course entities, ability entities, job entities and resource entities and their corresponding relationships. The preset mapping rules include field name mapping rules, field type constraint rules and entity relationship matching rules. Step 12: Perform encoding unification, enumeration value normalization, and primary-foreign key consistency verification on the mapped data. Generate course coverage capability data, job requirement capability data, course prerequisite relationship data, and course resource adaptation data based on the mapping relationship. When there is a coverage relationship between courses and capabilities, a requirement relationship between jobs and capabilities, a prerequisite relationship between courses, or an adaptation relationship between courses and resources, the corresponding position is assigned a value of 1; otherwise, it is assigned a value of 0. Step 13: Write the course coverage capability data, job requirement capability data, course prerequisite relationship data, and course resource adaptation data into the knowledge graph node set and relationship set, establish a two-way index structure of matrix index and graph node, so that the matrix coordinates correspond one-to-one with the node identifier, and generate a version identifier with timestamp and serial number, and solidify the above data and index structure into a versioned knowledge graph snapshot of the corresponding version.

3. The higher vocational education collaborative management system based on multi-source knowledge graphs and intelligent agents according to claim 2, characterized in that, Versioned knowledge graph snapshots undergo a multi-source data consistency maintenance process before generation, including: Step 21: Detect conflicts in entity field values ​​from different data sources. When multiple candidate values ​​exist for the same field of the same entity, determine the field value to retain according to the preset field priority table, remove other field values, and form a consistent set of field values. Step 22: Perform conflict detection on the relationships between course entities, ability entities, job entities and resource entities according to the start node identifier, relationship type and end node identifier. When duplicate relationships exist, retain the unique relationship entry according to the relationship arbitration rules. Step 23: Write the consistent set of field values ​​and the unique relation entries into the node set and relation set of the knowledge graph, and keep the node identifier and relation identifier unique to form a consistent entity and relation structure.

4. The higher vocational education collaborative management system based on multi-source knowledge graphs and intelligent agents according to claim 1, characterized in that, By compiling regulations, industry standards, and school rules, a discrete constraint set and a minimum requirement for course competency are obtained, including: Step 31: Based on the preset clause parsing rule base, perform sentence-level rule matching on the texts of regulations, industry standards and internal rules, and parse the matching results into a set of triplets consisting of preconditions, constraints and threshold parameters, and use the clause number as a unique identifier. Step 32: Based on the type of constraint, divide the set of triples into course attainment constraints, ability coverage constraints, job suitability constraints, and resource suitability constraints, and perform numerical normalization and unit standardization on the threshold parameters to obtain a standardized constraint set. Step 33: Map the standardized constraint set to the matrix positions corresponding to the course coverage capability data, job requirement capability data, course prerequisite relationship data, and course resource adaptation data through the index table. When the constraint is valid, assign a value of 1 to the corresponding position; otherwise, assign a value of 0 to form a discrete constraint set and a course capability compliance lower limit matrix. Step 34: Generate version identifiers using timestamps and serial numbers. Establish bidirectional indexes between the discrete constraint set and the course competency lower limit matrix and the courses, competencies, positions, and resource nodes in the knowledge graph according to indexing rules, forming versioned constraint objects that correspond one-to-one with the versioned knowledge graph snapshots.

5. The higher vocational education collaborative management system based on multi-source knowledge graphs and intelligent agents according to claim 1, characterized in that, The process of determining the winning bid execution plan includes: Step 51: Decompose the overall task in the order of time dimension first, resource dimension second, and object dimension last. When there is a conflict in the dimensions, the time dimension division result takes precedence to form a unique set of subtasks. For each subtask, register the task type, required resources, execution time period and affected objects to generate a subtask attribute table. Step 52: Each agent generates a proposal for the subtask, which includes feasibility evidence, impact measurement and solution summary. The proposals are organized into a standardized triplet structure, and a unique index table is constructed using the subtask identifier and agent identifier to form a standardized proposal set. Step 53: Set priority order, compare each item in the standardized proposal set, and compare the next priority item in turn when there are multiple proposals in the same order until the winning proposal is uniquely determined. Generate the winning execution plan, which includes sub-task identifier, execution time period, responsible entity, allocated resources and execution method, and register it in the scheduling table. The priority order is feasibility, scope of impact, resource consumption and evidence chain length.

6. The vocational education collaborative management system based on multi-source knowledge graphs and intelligent agents according to claim 1, characterized in that, The process of generating the course scheduling results includes: Step 61: Extract the sub-constraint set, course prerequisite relationship data, course resource adaptation data, and available teacher and teaching time slot data from the versioned knowledge graph snapshot. Define the scheduling variable set for courses, teachers, and time windows. The scheduling variable is indexed by the triple of course identifier, teacher identifier, and time window identifier. The value is 1 when the corresponding course is scheduled by a specified teacher in the specified time window, and 0 otherwise. Step 62: Establish a scheduling constraint model based on the sub-constraint set and course relationship data. The preset hard constraints include: class hour satisfaction constraint, resource mutual exclusion constraint, teacher mutual exclusion constraint, prerequisite order constraint, and resource adaptation constraint. Construct an objective function by weighted summation of unsatisfied constraint items. Specifically, by traversing all combinations of courses, teachers, and time windows, determine whether the corresponding scheduling variable satisfies the constraints. When the scheduling variable takes a value of 0, the corresponding weight is accumulated; when the variable takes a value of 1, no accumulation is performed. Summate the weights of all combinations to obtain the weighted value of unsatisfied constraints, and minimize this weighted value as the solution objective function. Step 63: Solve the course scheduling constraint model. When a feasible solution exists, extract the triplet of course, teacher and time window with a value of 1 and map it to the course scheduling result table.

7. The higher vocational education collaborative management system based on multi-source knowledge graphs and intelligent agents according to claim 1, characterized in that, The process of generating the set of student job qualifications includes: Step 71: Extract the student course set and course coverage ability data from the versioned knowledge graph snapshot. According to the one-to-many mapping relationship between courses and abilities, take the union of the ability items corresponding to the student courses and remove duplicates to form the student ability set. Step 72: Extract job requirement capability data from the versioned knowledge graph snapshot, and uniformly encode the student capability set and the job requirement capability set according to the preset capability dimension standardization rules, mapping capability items from different sources to unique capability identifiers; Step 73: Traverse the set of job requirement capabilities. For each job, determine whether the set of job requirement capabilities is completely included in the set of student capabilities. When the determination is true, add the corresponding job identifier to the set of student job qualifications to form the set of student job qualifications.

Images