AI-driven second classroom intelligent management cloud platform

Abstract

The application discloses an AI-driven second-classroom intelligent management cloud platform and particularly relates to the technical field of college second-classroom management, which comprises receiving experience records, declaration records and matter records uploaded by student terminals, teacher terminals and management terminals, extracting student identifiers, experience identifiers, time periods, participation roles, achievement contents, proof contents, declaration matters, declaration sequences, matter combination relations, matter mutual exclusion relations and matter precedence relations, and outputting a unified record set; the same second-classroom experience is constructed into an experience subgraph and a matter subgraph in a knowledge graph, and is further spread into shared pieces, exclusive pieces and linkage pieces that can be sequentially occupied by different identified matters, and unified calculation and result rewriting are performed on the occupation relations, mutual exclusion relations, precedence relations and remaining available relations among multiple matters.

Description

Technical Field

[0001] This invention relates to the field of extracurricular management technology in higher education institutions, and more specifically, to an AI-driven intelligent management cloud platform for extracurricular activities. Background Technology

[0002] In the technology related to the management of extracurricular activities in colleges and universities, the existing platform construction usually focuses on incorporating activity participation, result submission, credit conversion and evaluation application into a unified management link. The common approach is to record, collect, verify conditions and write results of experiences such as competition awards, volunteer service, club positions, practical projects and training lectures according to activity type, recognition items and review process.

[0003] Taking the scenario where a student's same award experience may be used simultaneously for innovation and entrepreneurship credit conversion, award and honor application, comprehensive quality file archiving, and college professional expansion achievement statistics as an example, the platform not only needs to operate continuously under the condition of multiple colleges and multiple lines of concurrent applications, but also needs to meet the restrictions on the use of different recognition items, mutual exclusion restrictions, sequential restrictions, and revocation and rollback restrictions. At the same time, the recognition results are required to be consistent and available for students to query, administrators to review, and for direct access to subsequent statistics.

[0004] However, in this scenario, most existing processing methods still treat the same extracurricular experience as a static record that can be read separately by each assessment item. This usually only allows for single-condition comparison, frequency control, or column restrictions, and it is difficult to handle the linkage effect on the availability of other items after the experience is occupied by a certain item. Therefore, in practice, the following situation often occurs:

[0005] The student's end shows that multiple items corresponding to the same experience meet the application requirements. However, during the final review, summary or graduation verification stage, it is found that the experience has been occupied by a previous item, cannot coexist with another item, the calling order is incorrect, or the remaining available part is insufficient. In the end, the only way to complete the processing is to rely on manual rollback, re-judgment or repeated explanation.

[0006] The technical problem this application aims to solve is: how to uniformly constrain the occupancy, mutual exclusion, sequence, and remaining availability relationships of the same extracurricular experience across multiple recognition items in the extracurricular intelligent management cloud platform, so that the experience can form only a unique, consistent, and traceable valid recognition result in the process of recognizing multiple items. Summary of the Invention

[0007] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an AI-driven intelligent management cloud platform for the second classroom. This platform constructs the same second classroom experience into an experience subgraph and an event subgraph in a knowledge graph, and further expands it into shared pieces, exclusive pieces, and linked pieces that can be occupied in sequence by different identified events. It performs unified calculation and result writing back on the occupancy relationship, mutual exclusion relationship, sequential relationship, and remaining availability relationship between multiple events, thereby solving the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: an AI-driven intelligent management cloud platform for the second classroom, comprising:

[0009] The data acquisition terminal module is used to receive experience records, application records, and event records uploaded by students, teachers, and administrators. It extracts student identifiers, experience identifiers, time periods, participating roles, achievement content, proof content, application items, application order, concurrent application relationships, mutually exclusive application relationships, and sequential application relationships, and outputs a unified record set.

[0010] The knowledge graph construction module is used to read a unified record set, using student identifiers, experience identifiers, occurrence time periods, participating roles, output content, proof content, and application items as nodes, and using attribution relationships, pointing relationships, limiting relationships, usage relationships, mutual exclusion relationships, and sequential relationships as edges to construct a second classroom knowledge graph and output experience subgraphs and item subgraphs.

[0011] The credential expansion module is used to read the experience subgraph and, based on the limiting edges, belonging edges, and pointing edges between the occurrence time node, participating role node, result content node, and proof content node, expand the same extracurricular experience into an experience credential surface composed of shared parts, exclusive parts, and linked parts.

[0012] The slice generation module is used to read the experience voucher surface and the event sub-graph, slice the experience voucher surface into shared slices, exclusive slices and linked slices, and output the occupied slice graph;

[0013] The occupancy calculation module is used to read the occupancy map and declaration records, push each declaration item node into the corresponding slice according to the declaration order, perform occupancy locking on exclusive slices, perform remaining share deduction on shared slices, perform subsequent available status rewriting on linked slices, cut off the occupancy path connected by mutual exclusion edges, and output the item occupancy result map and remaining available map.

[0014] The results generation module is used to read the occupancy result map and the remaining available map of the event, perform conflict cancellation, sequential review and map writing back, and output the unique corresponding valid identification result and its occupancy trajectory.

[0015] In a preferred embodiment, the data acquisition terminal module includes:

[0016] The experience records, application records, and event records uploaded by students, teachers, and administrators are read separately by source. The experience identifier, student identifier, time period, participating role, result content, proof content, application items, application order, concurrent application relationship, mutually exclusive application relationship, and sequential application relationship are extracted according to the type of uploading terminal. Records from different uploading terminals but with the same experience identifier are merged into the same collection group, and the terminal-separated record group is output.

[0017] For each terminal's separate record group, perform a group-based comparison of the occurrence time, participating roles, results content, and proof content. When the occurrence time is consistent and the results content and proof content are mutually referential, generate an experience master piece. When there is a correspondence between the declared items and the items used together, the items are mutually exclusive, or the items are sequential, generate an item constraint piece. Then, associate the experience master piece and the item constraint piece with the experience identifier and write them into a unified record set.

[0018] The unified record set performs missing item filling and conflict annotation on each experience main piece and each matter constraint piece. When there are multiple participating roles or multiple proof contents under the same experience identifier, the original correspondence is preserved and written into the position identifier in the piece respectively. When the same declaration matter corresponds to multiple sequential positions, the results are generated in order of upload time and output as a unified record set for the knowledge graph construction module to read.

[0019] In a preferred embodiment, the knowledge graph construction module includes:

[0020] Read the student identifier, experience identifier, occurrence time, participating role, result content, proof content, and application item of each record in the unified record set. Based on the rules of consistent student identifier, overlapping occurrence time, consistent item name in the result content and item name in the proof content, participating role not being a conflicting role limited by the mutual exclusion relationship of items, and the sequential position of the same application item uniquely corresponding to one record, establish record corresponding edges, and group records with record corresponding edges into the same candidate node group, and output candidate node group and candidate relationship group;

[0021] Based on candidate node groups and candidate relationship groups, each candidate node group is statistically analyzed for time period breaks, role conflicts, result proof mismatches, item sequence reversals, and item co-occurrence conflicts. The candidate node groups are then reconstructed in a fixed order of first splitting and then merging. The splitting rule is to separate records containing any conflict from the original candidate node group, and the merging rule is to merge candidate node groups with consistent student identifiers and no conflict. After each round of reconstruction, candidate relationship groups are regenerated based on experience identifier, occurrence time, participating role, result content, proof content, and application item, until the members of the candidate node groups obtained from two consecutive rounds of reconstruction are completely consistent and the edge items of the candidate relationship groups are completely consistent. A stable node set and a stable edge set are then output.

[0022] In a preferred embodiment, the knowledge graph construction module further includes:

[0023] Based on stable node sets and stable edge sets, consistency checks are performed on each edge from the perspectives of time period, role, result proof, and matter. Specifically, the time period perspective checks whether the occurrence time corresponds to the limiting relationship; the role perspective checks whether the participating roles correspond to the attribution relationship; the result proof perspective checks whether the result content corresponds to the pointing relationship; and the matter perspective checks whether the declared matter corresponds to the concurrent, mutually exclusive, and sequential relationships. Each edge is processed according to the following rules: if the number of supporting items is greater than the number of counter-evidence items, the edge is retained; if the number of supporting items is equal to the number of counter-evidence items, the edge is retained and a conflict flag is added; and if the number of supporting items is less than the number of counter-evidence items, the edge is deleted. Then, experience subgraphs containing attribution, pointing, and limiting relationships are extracted based on experience identifiers, and matter subgraphs containing concurrent, mutually exclusive, and sequential relationships are extracted based on declared matters. For each perspective check, a supported item is recorded if the check is successful, and a counter-evidence item is recorded if the check is unsuccessful.

[0024] In a preferred embodiment, the credential spreading module includes:

[0025] Read the occurrence time node, participating role node, result content node, and proof content node from the experience subgraph. Generate a shared part according to the rule that the same proof content node is connected to two or more result content nodes through pointing edges and falls into the same occurrence time node through limiting edges. Generate an exclusive part according to the rule that the same proof content node is connected to only one result content node through pointing edges and that the result content node corresponds to only one participating role node. Generate a linked part according to the rule that the result content node connected to the same proof content node through pointing edges corresponds to multiple participating role nodes and that any change in any participating role node will cause a change in the affiliation relationship of the remaining result content nodes. Output the initial voucher surface.

[0026] Based on the initial credential surface, record the corresponding experience identifier, occurrence time node, participating role node, result content node, and proof content node for each shared part, each exclusive part, and each linked part. Then, perform a rearrangement according to the rules that the same proof content node cannot be written into both the shared part and the exclusive part at the same time, the same result content node can only be assigned to one part under the same occurrence time node, and the same participating role node can only retain one ownership chain under the same proof content node, and output the experience credential surface.

[0027] Based on the experience credential surface, the limiting edge, attribution edge, and pointing edge of each part are checked item by item. The limiting edge is used to check the correspondence between the proof content node and the occurrence time node, the attribution edge is used to check the correspondence between the result content node and the participating role node, and the pointing edge is used to check the correspondence between the proof content node and the result content node. The part where the limiting edge, attribution edge, and pointing edge are all completely corresponding is retained in the experience credential surface, and the part with any missing correspondence is deleted from the experience credential surface. The experience credential surface is then output for the slice generation module to read.

[0028] In a preferred embodiment, the slice generation module includes:

[0029] Based on the shared, exclusive, and linked parts in the experience document surface, as well as the shared edges, mutually exclusive edges, and sequential edges in the item subgraph, each part is read item by item, including its corresponding proof content node, result content node, participating role node, and occurrence time node. The part that has a shared edge connection with two or more application item nodes and whose corresponding proof content node, result content node, participating role node, and occurrence time node are consistent is cut into a shared piece. The part that has a connection with only one application item node and whose application item node is separated from the other application item nodes by a mutually exclusive edge is cut into an exclusive piece. The part that has a sequential edge connection with two or more application item nodes and whose available relationship between the preceding application item node and the subsequent application item node is rewritten is cut into a linked piece. The initial occupied piece group is output.

[0030] Based on the initial occupied piece groups, write the experience identifier, piece identifier, corresponding declaration item node, corresponding proof content node, corresponding result content node, corresponding participating role node, and corresponding occurrence time node to each shared piece, each exclusive piece, and each linked piece. Then, perform a rearrangement according to the following rules: the same proof content node is assigned to only one shared piece or one exclusive piece under the same occurrence time node; the same result content node is assigned to only one exclusive piece or one linked piece under the same participating role node; and the linked pieces corresponding to the same declaration item node under the same experience identifier are connected end to end in the order of their edges. Output the occupied piece table.

[0031] Based on the occupancy map, the edge correspondence between each shared map, each exclusive map, and each linked map and the item subgraph is checked item by item. For shared maps, it is checked whether there are shared edges between the corresponding application item nodes. For exclusive maps, it is checked whether there are mutually exclusive edges between the corresponding application item node and other application item nodes. For linked maps, it is checked whether there are sequential edges between the corresponding application item nodes and whether the connection order is consistent with the arrangement order within the map. The complete map corresponding to each edge item and its interconnection relationship are written into the occupancy map, and the occupancy map is output for the pressure occupancy calculation module to read.

[0032] In a preferred embodiment, the occupancy calculation module includes:

[0033] Based on the shared pieces, exclusive pieces, linked pieces, mutually exclusive edges, and sequential edges in the occupied piece map, as well as the application items and application order in the application records, establish a piece matrix, piece continuation matrix, mutually exclusive matrix, and sequential matrix according to the same experience identifier. Multiply the row vector of the piece matrix corresponding to the current application item with the piece continuation matrix to generate a candidate piece sequence. Delete candidate piece sequences with non-zero mutually exclusive matrix mapping values, non-corresponding sequential matrix positions, and duplicate exclusive pieces. Output the candidate path table.

[0034] Based on the candidate path table, a path matrix, shared surplus vector, exclusive occupancy vector, and linkage transmission vector are established for each candidate piece sequence. Shared deduction, exclusive locking, and linkage transmission are sequentially performed on the path matrix. The path matrix is ​​then multiplied by the item piece matrix to obtain the item expansion matrix. Singular value decomposition is performed on the item expansion matrix to extract the column space corresponding to non-zero singular values. Orthogonal projection is performed on the column space to generate a standard path code. One candidate path with the same standard path code and consistent shared surplus vector, exclusive occupancy vector, and linkage transmission vector is retained. After each declared item is written sequentially, the process of shared deduction, exclusive locking, linkage transmission, multiplying the path matrix and item piece matrix to obtain the item expansion matrix, performing singular value decomposition on the item expansion matrix to extract the column space corresponding to non-zero singular values, performing orthogonal projection on the column space to generate a standard path code, and retaining candidate paths is repeated until the standard path code set, shared surplus vector set, exclusive occupancy vector set, and linkage transmission vector set obtained in two adjacent rounds are consistent. A stable path table is then output.

[0035] In a preferred embodiment, the occupancy calculation module further includes:

[0036] Based on the stable path table, a pressure check matrix, a surplus check matrix, and a transfer check matrix are constructed for each stable path. The pressure check matrix is ​​multiplied by the item piece matrix to check the item correspondence. The surplus check matrix is ​​checked against the shared surplus vector position by position to check the shared deduction result. The transfer check matrix is ​​checked against the linkage transfer vector position by position to check the linkage transfer result. The stable paths that have been checked are written into the item pressure result graph, and their corresponding shared surplus vector and linkage transfer vector are written into the remaining available graph.

[0037] In a preferred embodiment, the result generation module includes:

[0038] Based on the occupancy result diagram and the remaining available diagram, the occupancy paths, occupancy sequence, shared reserve and linkage transmission results corresponding to each declared item are collected according to the same experience identifier. Conflict cancellation is performed on each occupancy path under the same experience identifier. Occupancy paths with mutually exclusive edge connections, duplicate deductions after writing back the shared reserve, and linkage transmission results that do not correspond to the order of the edges are deleted. The pending result group is output.

[0039] Based on the pending results group, the order of each occupied path retained under the same experience identifier is reviewed. The order of the declared items in each occupied path is matched with the order of the occupied area sequence, and the occupied path corresponding to each position is retained as a valid path. The occupied path with the reversed order is deleted. Then, the valid identification result and the occupied trajectory are generated according to the occupied area sequence, shared reserve and linkage transmission result in the valid path.

[0040] Based on the valid identification results and the occupation trajectory, the valid identification results are written into the result nodes of the corresponding experience identifier and the declared items, the occupation trajectory is written into the connection edges between the corresponding occupied pieces, and the shared surplus and linkage transmission results are written back to the corresponding nodes and corresponding edges, outputting a unique valid identification result and its occupation trajectory.

[0041] The technical effects and advantages of this invention are as follows:

[0042] 1. This solution constructs an experience voucher, an occupancy map, and an occupancy result map, which unifies the occupancy relationship, mutual exclusion relationship, sequential relationship, and remaining availability relationship of the same experience among multiple assessment items into the same solution chain, thereby relatively reducing manual rollback and reversal caused by prior occupancy and subsequent conflicts in the final review stage;

[0043] 2. Constructing a unified record set into a knowledge graph containing experience subgraphs and event subgraphs, and performing reconstruction and consistency verification on the corresponding edges of the records, can transform scattered records into experience nodes and event nodes with clear relationships, thereby relatively improving the problems of unstable cross-terminal record grouping and misaligned event attachments;

[0044] 3. Expand the same experience into shared parts, exclusive parts, and linked parts, and then divide it into shared slices, exclusive slices, and linked slices. This can transform the original identification method of reading the entire experience into an identification method of reading the callable structure, which can relatively improve the clarity of the call boundaries when multiple matters are submitted in parallel.

[0045] 4. According to the order of application, the occupied area is subject to shared deduction, exclusive locking and linkage transmission. Combined with mutual exclusion matrix, priority matrix and path retention rules, invalid paths are screened out. This can relatively suppress the repeated deduction, order reversal and incompatible coexistence of the same experience in multiple calls.

[0046] 5. For the execution items of the stable path, corresponding verification, surplus verification and transfer verification are performed, and then the item occupancy result map and remaining availability map are generated. This allows the identification results and remaining availability status to be output synchronously, which relatively improves the reusability of status during subsequent review, statistics and re-application. Attached Figure Description

[0047] Figure 1 This is a schematic diagram of the system module structure of the present invention. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] Refer to the instruction manual appendix Figure 1 The AI-driven intelligent management cloud platform for the second classroom of this invention includes:

[0050] The data acquisition terminal module is used to receive experience records, application records, and event records uploaded by students, teachers, and administrators. It extracts student identifiers, experience identifiers, time periods, participating roles, achievement content, proof content, application items, application order, concurrent application relationships, mutually exclusive application relationships, and sequential application relationships, and outputs a unified record set.

[0051] In this implementation, the purpose of the data acquisition terminal module is to first compress the heterogeneous records uploaded by students, teachers, and administrators into the same field definition, and then organize the experience information and event constraint information corresponding to the same extracurricular experience into a unified record set that can be directly read by the subsequent knowledge graph construction module. This avoids problems such as mixed record sources, unstable experience grouping, and misaligned event constraints during subsequent node graph construction. To this end, the original uploaded records are first split into columns and fields are extracted. Then, the experience information within the same acquisition group is compared and the experience master piece and event constraint piece are generated. Finally, the generated results are filled with missing items, conflict annotations are made, and the order is sorted, so that the output results have a unified field structure, stable attribution relationships, and clear order relationships. This implementation process includes the following steps:

[0052] First, the original records from different upload terminals are converted into terminal-separated record groups with consistent field structures that can be grouped by experience. The input includes experience records uploaded by students, experience records uploaded by teachers, experience records uploaded by administrators, application records uploaded by students, application records uploaded by teachers, application records uploaded by administrators, and item records. The processing actions include: for each upload record, first write the upload terminal category and platform receiving time, then extract the experience identifier, student identifier, occurrence time, participating role, result content, proof content, application item, application order, item usage relationship, item mutual exclusion relationship, and item sequence relationship according to a fixed field order. Among them, experience records retain at least the experience identifier, student identifier, occurrence time, participating role, result content, and proof content; application records retain at least the experience identifier, student identifier, application item, and application order; and item records retain at least the item identifier, item name, item usage relationship, item mutual exclusion relationship, and item sequence relationship.

[0053] When the original record has an experience identifier, it is read directly. When the original record does not have an experience identifier, a joint experience identifier is generated by concatenating the student identifier, the date of occurrence, the event name, and the result content in that order. Then, records from different upload terminals are merged into the same collection group according to the rule of consistent experience identifiers. Within the same collection group, records are stored separately according to the upload terminal category. The output is the terminal-separated record group, which is written to the collection buffer for the same group alignment and comparison step. The handling of anomalies or missing records is as follows: when a time period is missing, the record is retained but a time period missing mark is written. When the student identifier is missing or the experience identifier cannot be generated, the record is written to the supplementary record table and does not enter the collection group.

[0054] Secondly, experience master pieces that can represent the facts of the experience and event constraint pieces that can represent the constraints of the event are extracted from the same collection group, so that the experience information and event information have been structurally separated before entering the unified record set; the input is the terminal split record group; the processing actions include: performing same-group alignment comparison on the occurrence time, participating roles, result content and proof content in the same collection group, where the judgment rule for the occurrence time to be consistent is that the start time and end time are the same, and records with overlapping but not completely identical occurrence times are only marked with an overlap mark and are not directly regarded as consistent; the judgment rule for the result content and proof content to be mutually referential is that the result content contains the proof material number, the proof content contains the result name or experience identifier, and at least two fields of the two are consistent;

[0055] When the occurrence time is consistent and the results and proofs are mutually referential, the experience identifier, student identifier, occurrence time, participating role, results, and proofs in the set of records are written into the experience main piece; when there is a corresponding relationship between the application items and the items, the mutual exclusion relationship between the items, or the sequential relationship between the items, the experience identifier, student identifier, application item, application order, and corresponding item relationship in the set of records are written into the item constraint piece, where the corresponding judgment rule is that the application item identifier is consistent with the item identifier in the item record; then the experience main piece and the item constraint piece are associated with the experience identifier and written into a unified record set, the output being a unified record set containing associated item items and written into the item association table for the missing item completion step to read; the abnormal or missing handling is as follows: when the results and proofs are not mutually referential, no experience main piece is generated and a missing reference mark is written; when the application item has no corresponding item identifier in the item record, no item constraint piece is generated and an item missing mark is written;

[0056] Secondly, the unified record set is used to complete fields, reveal conflicts, and fix the order of the segments, so that the records read by the subsequent knowledge graph construction module have complete fields, clear conflicts, and a unique order. The input is the unified record set and the segment association table. The processing actions include: completing missing items for each experience main segment and matter constraint segment in a fixed completion order, with priority given to completing the participating roles and proof content for teacher-side fields, priority given to completing the application items and relationship between items for management-side fields, and priority given to completing the result content for student-side fields; and then performing conflict annotation on the completed segments. Conflict types include time period conflict, role conflict, result proof mismatch conflict, and matter order conflict. The time period conflict is determined by the occurrence of two different start and end times under the same experience identifier. The role conflict is determined by the occurrence of two different participating roles for the same result content. The result proof mismatch conflict is determined by the occurrence of the result content and proof content no longer satisfying the mutual pointing condition. The matter order conflict is determined by the occurrence of two different application orders for the same application item.

[0057] When multiple participating roles or multiple proof contents exist under the same experience identifier, the original correspondence between each participating role and each proof content is retained, and the position identifiers within the slice are written in the order of platform reception time. When the same declaration item corresponds to multiple sequential positions, the order of the results is generated according to the platform reception time and written back to the item constraint slice. The output is a unified record set for the knowledge graph construction module to read and is written to a unified record storage table. The handling of anomalies or missing items is as follows: missing marks are retained for fields that cannot be filled and written along with the slice item. Slices with conflict marks are not deleted, and only the conflict codes are retained for subsequent disassembly and verification by the knowledge graph construction module.

[0058] In practical applications: For example, when a student participates in a school-level competition, the student uploads the award certificate and application details, the teacher uploads the guidance record and participation role, and the administrator uploads the corresponding credit information, evaluation information, and the relationship between the two. The data collection module first merges the records from the three ends into the same collection group according to the experience identifier. Then, it generates the main experience piece based on the correspondence between the certificate number, competition name, and experience identifier, and generates the event constraint piece based on the correspondence between the application details and the event identifiers of the event records. Subsequently, it fills in the missing proof content on the teacher's end, fills in the missing application order on the administrator's end, and writes the conflict code for the order conflict caused by repeated applications for the same competition. Finally, it outputs a unified record set. Through this implementation process, the subsequent knowledge graph construction module can directly build experience subgraphs and event subgraphs according to unified fields and stable grouping relationships, thereby reducing the graph construction deviation caused by inconsistent record sources, missing fields, and chaotic application order.

[0059] The knowledge graph construction module is used to read a unified record set, using student identifiers, experience identifiers, occurrence time periods, participating roles, output content, proof content, and application items as nodes, and using attribution relationships, pointing relationships, limiting relationships, usage relationships, mutual exclusion relationships, and sequential relationships as edges to construct a second classroom knowledge graph and output experience subgraphs and item subgraphs.

[0060] In this implementation, the purpose of the knowledge graph construction module is to organize the discrete records in the unified record set into a second-classroom knowledge graph that can stably express the factual relationships and event constraints, so that the experience subgraphs and event subgraphs read by the subsequent credential expansion module have clear attribution, clear relationships, and clear conflict states. This implementation process first establishes corresponding edges for records and forms candidate node groups and candidate relationship groups, then reconstructs the candidate node groups based on conflict items, and finally performs consistency checks on the stable edge set and extracts the experience subgraphs and event subgraphs. This implementation process includes the following steps:

[0061] First, candidate node groups and candidate relationship groups are formed based on the unified record set. The input quantities are student identifier, experience identifier, occurrence time, participating role, result content, proof content, and application item in the unified record set. The processing actions are as follows: read the unified record set one by one, and establish the corresponding edge of the record based on the following rules: consistent student identifier, overlapping occurrence time, the item name code corresponding to the result content is consistent with the item name code corresponding to the proof content, the participating role does not belong to the prohibited role set corresponding to the application item, and the same application order of the same application item corresponds to only one record. The judgment rule for overlapping occurrence time is that the start time of one record is not later than the end time of another record and the end time is not earlier than the start time of another record. The item name code is generated by normalization from the item name code table, and the prohibited role set is generated from the applicable role field of the item record.

[0062] When two records simultaneously meet the above rules, a record-corresponding edge is established between the two records, and the records connected by the record-corresponding edge are grouped into the same candidate node group. All record-corresponding edges in the candidate node group are summarized into a candidate relationship group. The output is the candidate node group and the candidate relationship group, which are written to the candidate node group table and the candidate relationship group table respectively for subsequent reconstruction reading. The handling of anomalies or missing items is as follows: records with missing items in the declaration only participate in the experience side grouping, and records with item names that cannot be unified are written to the name pending verification table and no record-corresponding edge is established.

[0063] Secondly, conflicts within candidate node groups are resolved to form stable node sets and stable edge sets. The inputs are candidate node groups and candidate relationship groups. The processing actions are as follows: For each candidate node group, time period break items, role conflict items, result proof mismatch items, item sequence reversal items, and item co-occurrence conflict items are counted. The rule for determining time period break items is that the occurrence time periods of two records are neither intersecting nor continuous. The rule for determining continuity is that the end time of the previous record is the same as the start time of the next record. The rule for determining role conflict items is that the same result content corresponds to two different participating roles, and at least one participating role belongs to the prohibited role set of the corresponding declared item. The rule for determining result proof mismatch items is that the item name code corresponding to the result content is inconsistent with the item name code corresponding to the proof content, or the proof content does not contain any two of the experience identifier, result name, and proof material number at the same time.

[0064] The rules for determining items with reversed order are: the two application items corresponding to the two records have a sequential relationship in the item order relationship table, and the application order in the records is the opposite of this sequential relationship; the rules for determining items with concurrent conflict are: the two application items corresponding to the two records do not appear in the item concurrent relationship table; then, candidate node groups are reconstructed in a fixed order of splitting and merging groups. When splitting groups, records containing any conflict items are separated from the original candidate node groups. When merging groups, candidate node groups with consistent student identifiers and no conflict items of the above five types are merged; after each round of reconstruction, candidate relationship groups are regenerated according to experience identifier, occurrence time, participating role, result content, proof content, and application item, until the members of the candidate node groups obtained from two consecutive rounds of reconstruction are completely consistent and the edge items of the candidate relationship groups are completely consistent; the output is a stable node set and a stable edge set, which are written to the stable node table and the stable edge table respectively for consistency verification; the handling of anomalies or missing items is: candidate node groups consisting of a single record are directly retained, and if multiple groups do not meet the merging conditions, the original group is retained;

[0065] Next, consistency checks are performed on the stable edge set, and experience subgraphs and event subgraphs are extracted. The inputs are the stable node set and the stable edge set. The processing actions are as follows: consistency checks are performed on each edge from the perspectives of time period, role, result proof, and event. The time period perspective checks whether the occurrence time corresponds to the limiting relationship. The rule for the establishment of the limiting relationship is that the occurrence date or start and end time corresponding to the proof content is consistent with or completely contained in the occurrence time in the record. The role perspective checks whether the participating role corresponds to the attribution relationship. The rule for the establishment of the attribution relationship is that the executing entity corresponding to the result content is consistent with the participating role. The result proof perspective checks whether the result content corresponds to the pointing relationship. The rule for the establishment of the pointing relationship is that any two of the result name, result number, and experience identifier appear in the proof content and are consistent with the result content. The event perspective checks whether the declared event corresponds to the combined relationship, the mutually exclusive relationship, and the sequential relationship. The rule for the establishment of the combined relationship is that the event pair appears in the event combined relationship table. The rule for the establishment of the mutually exclusive relationship is that the event pair does not appear in the event mutually exclusive relationship table. The rule for the establishment of the sequential relationship is that the declaration order of the event pair is consistent with the order in the event sequential relationship table.

[0066] For each perspective verification, a supporting item is recorded if the verification is true, and a counter-verification item is recorded if the verification is false. Edges are processed according to the following rules: if the number of supporting items is greater than the number of counter-verification items, the edge is retained; if the number of supporting items is equal to the number of counter-verification items, the edge is retained and a conflict flag is written; if the number of supporting items is less than the number of counter-verification items, the edge is deleted. Subsequently, experience subgraphs containing attribution, pointing, and limiting relationships are extracted based on experience identifiers. Item subgraphs containing combined, mutually exclusive, and sequential relationships are extracted based on declared items, and these are written to the experience subgraph and item subgraph for the voucher expansion module to read. Abnormal or missing data is handled as follows: if an input field for a certain perspective is missing, no supporting or counter-verification items are counted for that perspective; only a missing field flag is written. Isolated nodes uniquely connected by the deleted edge are not included in the subgraph extraction results.

[0067] In practical applications: For example, if the same student submits award certificates, guidance records, and credit application records for the same competition experience, the knowledge graph construction module first establishes corresponding edges for the records based on student identification, intersection of occurrence time, consistency of event name codes, and role adaptation. Then, based on the inversion of event sequence, mismatch of achievement certificates, and events, and using conflict items, it performs splitting and merging of candidate node groups to obtain a stable node set and a stable edge set. Finally, it performs a four-view consistency check on the stable edge set and extracts the experience subgraph and event subgraph. Through this implementation process, the unified record set is transformed into a second classroom knowledge graph with a stable structure, clear relationships, and traceable conflicts, providing direct input for the subsequent generation of experience evidence surfaces.

[0068] The credential expansion module is used to read the experience subgraph and, based on the limiting edges, belonging edges, and pointing edges between the occurrence time node, participating role node, result content node, and proof content node, expand the same extracurricular experience into an experience credential surface composed of shared parts, exclusive parts, and linked parts.

[0069] In this implementation, the purpose of the credential expansion module is to organize the node relationships in the experience subgraph into experience credential surfaces that can directly participate in the generation of subsequent slices. This allows content that can be called in parallel, can only be called individually, and will cause changes in subsequent usable relationships within the same second-classroom experience to be first layered and merged, and then its position is rearranged and its relationship is checked. This avoids problems such as duplicate inclusion of the same proof content, confusion in the attribution of the same result content, or distortion of linkage relationships when generating subsequent slices. The implementation process first generates an initial credential surface composed of shared parts, exclusive parts, and linked parts based on the experience subgraph, then performs field write-back and position rearrangement on the initial credential surface, and finally performs relationship integrity check on the rearrangement result and outputs the experience credential surface. The implementation process includes the following steps:

[0070] First, identify the shareable, exclusive, and interconnected credential structures within the same extracurricular experience from the experience subgraph. The input consists of occurrence time nodes, participating role nodes, outcome content nodes, proof content nodes, as well as limiting edges, belonging edges, and pointing edges in the experience subgraph. The processing steps are as follows: read each proof content node, retrieve the set of outcome content nodes connected to the proof content node via pointing edges, then retrieve the occurrence time nodes connected to the proof content node via limiting edges, and retrieve the participating role nodes connected to each outcome content node via belonging edges. When the same proof content node is simultaneously connected to two or more outcome content nodes via pointing edges, and the proof relationships corresponding to the two or more outcome content nodes all fall into the same occurrence time node via limiting edges, the proof content node, the corresponding outcome content node, the corresponding participating role node, and the corresponding occurrence time node are written as the shared part. When the same proof content node is only connected to one outcome content node via pointing edges, and the outcome content node is only connected to one participating role node via belonging edges, the proof content node, the outcome content node, the participating role node, and the corresponding occurrence time node are written as the exclusive part.

[0071] When the same proof content node is connected to two or more result content nodes via pointing edges, and the result content node corresponds to two or more participating role nodes, and deleting the belonging edge corresponding to any of the participating role nodes would cause the remaining result content nodes to lose their original belonging relationship or be transferred to other participating role nodes, the proof content node, the corresponding result content node, the corresponding participating role node, and the corresponding occurrence time node are written as the linked part; the output is the initial credential surface and written to the initial credential table for the rearrangement step to read; the abnormal or missing handling is as follows: when the proof content node is not connected to any result content node, no part is generated and a missing pointing mark is written; when the proof content node is connected to multiple occurrence time nodes and the occurrence time nodes are inconsistent, the proof content node is retained but a time conflict mark is written, and it is not directly classified into the shared part, exclusive part, or linked part;

[0072] Secondly, the positions and affiliations of each part in the initial credential surface are fixed and organized to ensure that the credential structure of the same extracurricular experience has a unique part affiliation and a stable node arrangement before entering the subsequent slice generation. The input is the initial credential surface. The processing actions are as follows: for each shared part, each exclusive part, and each linked part, the corresponding experience identifier, occurrence time node, participating role node, result content node, and proof content node are recorded respectively, and the part list is reconstructed with the experience identifier as the merging primary key. Then, the reordering is performed in a fixed order, which is to first handle the conflict between the shared part and the exclusive part, then handle the conflict of the result content node affiliation, and finally handle the conflict of the participating role node affiliation chain. Among them, the processing rule that the same proof content node cannot be written into the shared part and the exclusive part at the same time is: if the same proof content node appears in both the shared part and the exclusive part in the initial credential table, the number of result content nodes connected by the pointing edge of the proof content node is counted. If the number is one, the exclusive part is retained and the shared part is deleted. If the number is two or more, the shared part is retained and the exclusive part is deleted.

[0073] The processing rule for the same result content node being assigned to only one part under the same occurrence time node is as follows: If the same result content node appears in two parts under the same occurrence time node, the part with more connecting edges to the corresponding proof content node is retained first. If the number of connecting edges is the same, the part with the earlier writing time in the initial voucher table is retained. The processing rule for the same participating role node being retained to have only one attribution chain under the same proof content node is as follows: For multiple participating role nodes corresponding to the same proof content node, an attribution chain sequence is generated according to the order of attribution edge writing. Only the first attribution chain is retained and the rest are deleted. The output is the experience voucher surface and is written to the experience voucher table for the verification step to read. The handling of anomalies or missing items is as follows: When the number of connecting edges is the same and the writing time is the same, and they cannot be directly resolved, the original correspondence is retained and a parallel mark is written, and the verification step continues to process it.

[0074] Next, the integrity of the relationships between the various parts of the experience credential surface is verified, retaining only the credential structure that can be fully supported by the experience subgraph. The input quantities are the limiting edges, belonging edges, and pointing edges in the experience credential surface and experience subgraph. The processing action is as follows: for each shared part, each exclusive part, and each linked part in the experience credential surface, the limiting edges, belonging edges, and pointing edges are verified item by item. The limiting edges are used to verify the correspondence between the proof content node and the occurrence time node. The verification rule is that the proof content node is uniquely connected to the occurrence time node corresponding to that part through the limiting edge. The belonging edges are used to verify the correspondence between the result content node and the participating role node. The verification rule is that the result content node is uniquely connected to the participating role node corresponding to that part through the belonging edge.

[0075] Pointing edges are used to verify the correspondence between the proof content nodes and the result content nodes. The verification rule is that the proof content node is directly connected to the corresponding result content node via the pointing edge. When the limiting edge, belonging edge, and pointing edge of a part are all completely corresponding, the part is retained in the experience credential surface. When a part is missing any correspondence, the part is deleted from the experience credential surface. After verification, the retained parts are re-summarized according to the experience identifier to generate an experience credential surface for the slice generation module to read and write it to the credential output table. The abnormal or missing handling is as follows: when a part is missing only one correspondence, a missing type code is written for subsequent traceability. When all parts under the same experience identifier are deleted, an empty credential mark is output and the transmission of the experience identifier to the slice generation module is stopped.

[0076] In practical applications: For example, if a student submits an award certificate and a team division of labor description in an innovation and entrepreneurship competition, and the award certificate points to both the competition award and the credit recognition achievement, and both occurred on the same competition date, the credential expansion module will generate a shared portion of the content corresponding to the award certificate. If another individual defense certificate only points to the student's personal presentation achievement, and this achievement corresponds to only one participating role node for that student, then it will generate an exclusive portion. If the team division of labor description connects both the project achievement node and the defense achievement node, and after deleting the belonging edge corresponding to the team leader role, the project achievement node will be rewritten to be taken over by the instructor role, then the team division of labor description will generate a linked portion. Subsequently, the module will perform a rearrangement according to the certificate content node, achievement content node, and participating role node, and check the limiting edge, belonging edge, and pointing edge item by item, retaining only the part with complete relationships to write into the experience credential surface. Through this implementation process, the experience subgraph is organized into an experience credential surface with a clear structure, unique belonging, and complete relationships, providing direct input for the subsequent generation of shared, exclusive, and linked pieces.

[0077] The slice generation module is used to read the experience voucher surface and the event sub-graph, slice the experience voucher surface into shared slices, exclusive slices and linked slices, and output the occupied slice graph;

[0078] In this implementation, the purpose of the slice generation module is to further convert the shared, exclusive, and linked parts already layered in the experience voucher into an occupancy slice diagram that can be directly read by the subsequent occupancy calculation module. This allows for the segmentation of the call relationships of items that can be used, mutually exclusive, or passed sequentially within the same second-classroom experience, followed by placement and verification. This avoids problems such as duplicate occupancy of the same proof content, cross-slice mixing of the same result content, or incorrect linkage order during subsequent occupancy calculations. The implementation process first generates an initial occupancy slice group based on the experience voucher and item sub-graphs, then performs field writing and position rearrangement on the initial occupancy slice group, and finally verifies the edge item correspondence between each slice and the item sub-graph and outputs the occupancy slice diagram. This implementation process includes the following steps:

[0079] First, based on the experience credential surface and the item subgraph, identify the shared, exclusive, and linked pieces that can be called by the item. The input quantities are the shared parts, exclusive parts, and linked parts in the experience credential surface, as well as the shared edges, mutually exclusive edges, and sequential edges in the item subgraph. The processing actions are as follows: for each part, read its corresponding proof content node, result content node, participating role node, and occurrence time node item by item, and then read the application item nodes that have a connection relationship with this part. Among them, the connection relationship between the piece and the application item node is established according to the item name code correspondence rule. The item name code correspondence rule is that the item name contained in the proof content node or result content node is consistent with the item identifier of the application item node after being normalized by the item name code table. When the same part has a shared edge connection relationship with two or more application item nodes, and the proof content node, result content node, participating role node, and occurrence time node corresponding to this part are consistent under each application item node, this part is cut into a shared piece.

[0080] When a part is connected to only one application item node, and that application item node is separated from the other application item nodes by a mutual exclusion edge, the part is cut into an exclusive slice. When a part is connected to two or more application item nodes by successive edges, and the calling status of the subsequent application item node changes from callable to restricted or uncallable after the preceding application item node is called, the part is cut into a linked slice. The output is the initial occupied slice group and is written to the initial occupied slice table for the reordering step to read. The exception or missing item handling is as follows: when a part does not match any application item node, it is not sliced ​​and an item missing mark is written. When the same part meets two types of slicing conditions at the same time, it is judged in a fixed order of shared slice, exclusive slice, and linked slice. If the former is true, the latter will not be judged.

[0081] Secondly, the execution fields of each piece in the initial occupied piece group are fixed and the positions are organized to ensure that occupied pieces under the same experience identifier have unique piece ownership and stable arrangement; the input is the initial occupied piece group; the processing action is: write the experience identifier, piece identifier, corresponding declaration item node, corresponding proof content node, corresponding result content node, corresponding participating role node and corresponding occurrence time node to each shared piece, each exclusive piece and each linked piece, where the piece identifier is generated by sequentially concatenating the experience identifier and the sequence number in the piece, and the sequence number in the piece is generated by incrementing according to the initial occupied piece writing time;

[0082] The reordering is then performed in a fixed order: first, handling conflicts related to the attribution of proof content nodes; then, handling conflicts related to the attribution of result content nodes; and finally, handling the connection order of linked pieces. Specifically, the rule for assigning the same proof content node to only one shared piece or one exclusive piece under the same occurrence time node is as follows: if the same proof content node appears in both shared and exclusive pieces under the same occurrence time node, the number of application item nodes corresponding to that proof content node is counted; if the number is one, the exclusive piece is retained; if the number is two or more, the shared piece is retained. The rule for assigning the same result content node to only one exclusive piece or one linked piece under the same participating role node is as follows: if the same result content node appears in both exclusive and linked pieces under the same participating role node, the number of preceding and following edges corresponding to that result content node is read first; if preceding and following edges exist, the linked piece is retained; if preceding and following edges do not exist, the exclusive piece is retained. The rule for connecting linked pieces corresponding to the same application item node under the same experience identifier according to the preceding and following edge order is as follows:

[0083] Map the corresponding application item node pairs of each linkage segment to the sequential edge table, generate the linkage segment arrangement sequence according to the order of the preceding item in front and the following item in back, and connect the end of the previous linkage segment to the beginning of the next linkage segment; the output is the occupied segment table and written to the occupied segment storage table for the verification step to read; the abnormal or missing handling is as follows: when the segment identifier is duplicated when it is generated, the sequence number is appended to the sequence number in the original segment to regenerate the segment identifier; when there is no sequential edge that can be connected between linkage segments, the original linkage segment is retained and a connection missing mark is written.

[0084] Next, the edge correspondence between each piece in the occupied piece table and the item subgraph is checked, retaining only occupied pieces with complete relationships and valid connections and their inter-piece connections; the input quantities are shared edges, mutually exclusive edges, and sequential edges in the occupied piece table and the item subgraph; the processing action is: check the edge correspondence between each shared piece, each exclusive piece, and each linked piece and the item subgraph, where the verification rule for shared pieces is that there are shared edges between all the application item nodes corresponding to the shared piece; the verification rule for exclusive pieces is that there are mutually exclusive edges between the application item nodes corresponding to the exclusive piece and all other connected application item nodes.

[0085] The verification rule for linked segments is that there are sequential edges between the application item nodes corresponding to the linked segment, and the order of the sequential edges is consistent with the order of the linked segment in the occupied segment table. When all the corresponding edge items of a segment are true, the segment is retained and written into the occupied segment map. When a segment is missing any corresponding edge item, the segment is deleted from the occupied segment map. After the verification is completed, the retained segments and their inter-segment connections are summarized and written into the occupied segment map according to the experience identifier. The occupied segment map is output for the pressure and occupancy calculation module to read and written into the occupied segment map. The abnormal or missing handling is as follows: when the number of corresponding item nodes of a shared segment is less than two, the number of corresponding item nodes of a single segment is not one, or the number of corresponding item nodes of a linked segment is less than two, the segment is directly deleted and a segment type mismatch mark is written. When all occupied segments under a certain experience identifier are deleted, an empty occupied segment map mark is output and the transmission of the experience identifier to the pressure and occupancy calculation module is stopped.

[0086] In practical applications: For example, a student's competition experience forms a shared part, a unique part, and a linked part in the experience credential surface. The shared part corresponds to both innovation and entrepreneurship credits and comprehensive quality items, the unique part only corresponds to the evaluation item, and the linked part corresponds to two application item nodes that first participate in the college-level assessment and then enter the university-level assessment. The slice generation module first slices the shared part into shared slices, the unique part into unique slices, and the linked part into linked slices according to the item name code correspondence. Then, it performs a rearrangement according to the proof content node, achievement content node, and application item node. Finally, it checks whether there are shared edges between the item nodes corresponding to the shared slice, whether there are mutually exclusive edges between the item nodes corresponding to the unique slice and other item nodes, and whether there are sequential edges between the item nodes corresponding to the linked slice and the order of arrangement is consistent. Through this implementation process, the experience credential surface is converted into an occupied slice graph with clear edge items, unique slice ownership, and stable connection order, providing direct input for the subsequent occupancy calculation module to perform occupancy locking, shared deduction, and linked transmission.

[0087] The occupancy calculation module is used to read the occupancy map and declaration records, push each declaration item node into the corresponding slice according to the declaration order, perform occupancy locking on exclusive slices, perform remaining share deduction on shared slices, perform subsequent available status rewriting on linked slices, cut off the occupancy path connected by mutual exclusion edges, and output the item occupancy result map and remaining available map.

[0088] In this implementation, the purpose of the occupancy calculation module is to calculate a set of stable paths that satisfy mutual exclusion constraints, sequence constraints, and occupancy constraints by combining the shared, exclusive, and linked pieces in the occupancy map with the application items and application order in the application record. This further generates an application occupancy result map and a remaining availability map, ensuring that the calling relationship between multiple application items for the same extracurricular experience first completes path filtering, then status transfer, and finally result verification. This avoids problems such as item mismatch, duplicate deduction of shared pieces, or distortion of linked piece transfer during subsequent result generation. The implementation process first establishes a candidate piece sequence and filters out invalid paths, then performs status expansion and stability retention on the candidate piece sequence, and finally verifies the stable paths and writes them into the application occupancy result map and the remaining availability map respectively. This implementation process includes the following steps:

[0089] First, a candidate piece sequence satisfying basic constraints is generated based on the piece occupancy map and declaration records. The input quantities are shared pieces, exclusive pieces, linked pieces, mutually exclusive edges, and sequential edges in the piece occupancy map, as well as the declaration items and declaration order in the declaration records. The processing actions are as follows: all declaration items and all occupied pieces corresponding to the same experience are aggregated according to the same experience identifier, and a piece matrix, piece continuation matrix, mutual exclusion matrix, and sequential matrix are established. In the piece matrix, rows correspond to declaration items, columns correspond to occupied pieces, and an element value of 1 indicates that the declaration item can call the occupied piece, while an element value of 0 indicates that the declaration item cannot call the occupied piece. In the piece continuation matrix, both rows and columns correspond to occupied pieces, and an element value of 1 indicates that the previous occupied piece can be followed by the next occupied piece, while an element value of 0 indicates that continuation is not allowed. In the mutual exclusion matrix, both rows and columns correspond to declaration items, and an element value of 1 indicates that two declaration items are mutually exclusive, while an element value of 0 indicates that two declaration items are not mutually exclusive.

[0090] The rows and columns of the sequence matrix correspond to the declared items. An element value of 1 indicates that the declared item in the row precedes the declared item in the column, and an element value of 0 indicates that there is no such sequence relationship. Then, the current declared item is read item by item in the declaration order. The row vector of the item slice matrix corresponding to the current declared item is multiplied continuously by the slice continuation matrix. The stopping condition for continuous multiplication is that no new occupied slice position appears in the product result of the new round of multiplication, so as to obtain the candidate slice sequence corresponding to the current declared item. Each candidate slice sequence is deleted. The deletion rules include: when the element value of the item pair corresponding to the candidate slice sequence is 1 after mapping to the mutual exclusion matrix, the candidate slice sequence is deleted.

[0091] If the order of items corresponding to the candidate piece sequence is inconsistent after being mapped to the sequence matrix, the candidate piece sequence is deleted; if the same exclusive piece appears twice or more in the same candidate piece sequence, the candidate piece sequence is deleted; the output is a candidate path table and is written to the candidate path storage table for the state expansion step to read; the abnormal or missing handling is as follows: when the value of all elements in the corresponding row of a certain declaration item in the item piece matrix is ​​zero, a no-call piece mark is directly written for the declaration item and no candidate piece sequence is generated; when all candidate piece sequences under the same experience identifier are deleted, an empty path mark is written and the subsequent pressure calculation of the experience identifier is stopped;

[0092] Secondly, the candidate slice sequences are expanded in terms of occupancy status and converged in terms of path to obtain a stable path table that can be used for result verification. The inputs are the candidate path table, the slice matrix, the slice continuation matrix, and the sequence matrix. The processing actions are as follows: for each candidate slice sequence, a path matrix, a shared surplus vector, an exclusive occupancy vector, and a linkage transmission vector are established. In the path matrix, the rows correspond to the declaration order position, the columns correspond to the occupied slices, the element value of 1 indicates that the occupied slice is called at that order position, and the element value of 0 indicates that the occupied slice is not called at that order position. Each bit of the shared surplus vector corresponds to the remaining available space of each shared slice. The initial value of the call count is taken from the initial callable count field of the shared slice in the occupancy slice diagram. The initial callable count is given by the event relationship constraint rules, that is, the number of event nodes that can be used together corresponding to the shared slice. Each bit of the exclusive occupancy vector corresponds to the occupancy status of each exclusive slice. The initial value of each bit is zero. A value of zero indicates that the slice is not occupied, and a value of one indicates that the slice is occupied. Each bit of the linkage transmission vector corresponds to the status code of each linkage slice. The initial value is taken from the initial status code field of the occupancy slice diagram. A value of zero indicates that the slice has not been transmitted, a value of one indicates that the slice has been transmitted, and a value of two indicates that the subsequent event calls are restricted after the slice has been transmitted.

[0093] Subsequently, the path matrix is ​​processed sequentially using shared deduction, exclusive locking, and linkage propagation. The shared deduction is calculated by subtracting the shared remainder vector bit by bit according to the frequency of each shared piece in the path matrix. The exclusive locking is calculated by rewriting the corresponding bit of the exclusive occupancy vector to one when the value of an element corresponding to an exclusive piece in the path matrix is ​​first written as one. The linkage propagation is calculated by passing the linkage piece status code corresponding to the preceding and following items in the sequence matrix to the linkage piece position corresponding to the following item. If, after propagation, the linkage piece restricts the subsequent item's access, the corresponding piece position of the following item is written as restricted. After the above processing, the path matrix and the item piece matrix are multiplied to obtain the item expansion matrix. Singular value decomposition is performed on the item expansion matrix, and the column space corresponding to non-zero singular values ​​is extracted. All elements of the item expansion matrix are discrete integers, and non-zero singular values ​​are strictly greater than zero singular values.

[0094] Perform orthogonal projection on the column space to generate standardized path codes. The generation rule for standardized path codes is to read the column positions retained after projection in the order of declaration and concatenate the corresponding piece identifiers in sequence. For candidate paths with the same standardized path code and consistent shared surplus vector, exclusive occupancy vector, and linkage transmission vector, only one path is retained, and the rest are deleted. Repeat the above operation after each declaration item is written in sequence until the standardized path code set, shared surplus vector set, exclusive occupancy vector set, and linkage transmission vector set obtained in two adjacent rounds are completely consistent. The output is a stable path table and written to the stable path storage table for the verification step to read. The abnormal or missing handling is as follows: when any bit of the shared surplus vector is less than zero after deduction, the candidate path is deleted. When the same subsequent item is simultaneously in the callable state and the restricted state after linkage transmission, the restricted state is retained and a transmission conflict flag is written. When multiple candidate paths are retained under the above rules, the next round of writing continues without premature deletion in this step.

[0095] Next, a complete verification of the correspondence between items executed on the stable path, the shared deduction results, and the linkage transmission results is performed, and an item occupancy result diagram and a remaining availability diagram are generated. The input quantities are the stable path table, the item slice matrix, the shared surplus vector, and the linkage transmission vector. The processing actions are as follows: for each stable path, an occupancy verification matrix, a surplus verification matrix, and a transmission verification matrix are constructed. In the occupancy verification matrix, the rows correspond to the declared items, the columns correspond to the occupied slices, and the element values ​​are taken from the actual call status of the declared item for the occupied slice in the stable path. In the surplus verification matrix, the rows correspond to the declared items, the columns correspond to the shared slices, and the element values ​​are taken from the actual deduction times of the declared item for the shared slice in the stable path. In the transmission verification matrix, the rows correspond to the preceding declared items, the columns correspond to the following declared items, and the element values ​​are taken from the linkage slice status code transmission results in the stable path.

[0096] Then, the occupancy check matrix and the item slice matrix are multiplied to check the item correspondence. The rule for a successful check is that the element value of each non-zero position in the occupancy check matrix is ​​one in the corresponding position in the item slice matrix. The surplus check matrix and the shared surplus vector are checked bit by bit to verify the shared deduction result. The rule for a successful check is that the result obtained by subtracting the sum of the corresponding column elements of the surplus check matrix from the initial callable number of a shared slice is the same as the corresponding bit of the shared surplus vector. The transfer check matrix and the linkage transfer vector are checked bit by bit to verify the linkage transfer result. The rule for a successful check is that the element value of the item pair with a sequential relationship in the transfer check matrix is ​​consistent with the status code corresponding to the linkage transfer vector.

[0097] When all three types of verifications are true, the stable path is written into the item occupancy result graph, and the shared surplus vector and linkage transmission vector corresponding to the stable path are written into the remaining available graph. The item occupancy result graph is written into the declared item node, occupied piece node and inter-piece call edge according to the experience identifier, and the remaining available graph is written into the shared piece surplus field and linkage piece status field according to the experience identifier. The output is the item occupancy result graph and the remaining available graph and is written into the result generation module read table. The abnormal or missing handling is as follows: when any matrix in the occupancy verification matrix, surplus verification matrix or transmission verification matrix has an empty row, the stable path is deleted and a verification missing mark is written. When there are two or more stable paths that are verified under the same experience identifier, one is retained and the rest are deleted according to the rule of the standard path code dictionary order.

[0098] In practical applications: For example, if the same student applies for the same competition experience simultaneously as an innovation and entrepreneurship credit item, a comprehensive quality item, and an award item, where the innovation and entrepreneurship credit item and the comprehensive quality item share the same common segment, while the award item occupies a unique segment, and the school-level award item is also subject to the linkage segment transmission constraint of the college-level award item; the pressure calculation module first establishes the item segment matrix, segment continuation matrix, mutual exclusion matrix, and sequence matrix, then generates a candidate segment sequence according to the application order and deletes candidate segment sequences where mutual exclusion items coexist, the sequence positions are reversed, or unique segments appear repeatedly; subsequently, the shared deduction is performed sequentially on the retained candidate segment sequences. Exclusive locking and linkage transmission are implemented, and standardized path codes are generated through singular value decomposition and orthogonal projection of the event expansion matrix. Candidate paths with the same status are merged. Finally, the event correspondence, shared deduction results, and linkage transmission results of stable paths are checked item by item. Only the stable paths that have been verified are written into the event occupancy result diagram, and the shared piece balance and linkage piece status are written into the remaining available diagram. Through this implementation process, the occupancy relationship of the same experience among multiple declared events is solved into a set of stable results that satisfy mutual exclusion, order, and occupancy constraints, providing a direct basis for the subsequent result generation module to output a unique and valid identification result.

[0099] The results generation module is used to read the occupancy result map and the remaining available map of the matter, perform conflict cancellation, sequential review and map writing back, and output a unique corresponding valid identification result and its occupancy trajectory.

[0100] In this implementation, the purpose of the result generation module is to further consolidate the already established over-occupancy results in the over-occupancy result map and the remaining available map into a unique valid identification result and its over-occupancy trajectory under the same experience identifier. This ensures that the multiple feasible paths obtained from the previous over-occupancy calculation first complete conflict cancellation, then sequential review, and finally map back-writing. This avoids retaining multiple sets of results, inverted order results, or distorted write-back results for the same second-classroom experience across multiple application items. The implementation process first aggregates over-occupancy paths according to experience identifiers and deletes conflicting paths, then performs sequential review on the retained paths and generates valid identification results and over-occupancy trajectories, and finally writes the valid identification results and over-occupancy trajectories back to the result nodes and connecting edges. This implementation process includes the following steps:

[0101] First, conflict resolution is performed on multiple occupied paths under the same experience identifier, retaining only candidate results that meet basic consistency. The input quantities are the application item node, occupied piece node, inter-piece call edge, and occupied path in the item occupied result graph, as well as the shared reserve and linkage transmission results in the remaining available graph. The processing action is as follows: according to the same experience identifier, the occupied path, occupied piece sequence, shared reserve, and linkage transmission results corresponding to each application item under that experience identifier are aggregated to form a path aggregation table. Then, conflict resolution is performed on each occupied path in the path aggregation table. The determination rule for the existence of mutually exclusive edge connections in the occupied piece sequence is that the element value of any two application items corresponding to the same occupied path after mapping to the item mutual exclusion relationship table is one, or there is an inter-piece relationship between any two occupied pieces connected by mutually exclusive edges. The determination rule for duplicate deductions after the shared reserve is that the initial callable number of the same shared piece minus the total number of deductions for the shared piece in the occupied path, and then minus the remaining number corresponding to the shared piece in the remaining available graph, results in a result that is not zero.

[0102] The judgment rule for the non-correspondence between the linkage transmission result and the order of the edges is that the status code corresponding to the preceding item in the linkage transmission result is not written before the status code corresponding to the following item, or the status code of the following item has been rewritten while the status code of the preceding item remains in the untransmitted state; when a certain occupied path meets any of the above deletion conditions, the occupied path is deleted; when the deletion conditions are not met, the occupied path is retained; the output is a pending result group and written to the pending result table for sequential review steps to read; the abnormal or missing handling is as follows: when no occupied path is gathered under the same experience identifier, an empty result mark is written and the subsequent result generation of the experience identifier is stopped; when the shared margin or linkage transmission result is missing, the corresponding occupied path is written to the pending review result table instead of being directly deleted;

[0103] Secondly, the order of the retained paths in the pending result group is reviewed, and valid identification results and occupation trajectories are generated. The input is the pending result group. The processing action is as follows: for each retained occupation path under the same experience identifier, the order of the declared items and the order of the occupied pieces are read one by one, and the position of the declared item is matched with the position of the occupied piece written. The rule for the positional correspondence is that the writing order position of the i-th declared item in the occupation path corresponds one-to-one with the calling order position of the i-th occupied piece in the occupied piece sequence. If two declared items have a sequential relationship in the item sequence table, the occupied piece corresponding to the preceding item must be located before the occupied piece corresponding to the following item in the occupied piece sequence. When an occupation path meets the above correspondence rules, the occupation path is retained as a valid path. When any declared item order and the occupied piece sequence order are reversed in an occupation path, the occupation path is deleted.

[0104] For the retained valid paths, valid identification results and occupation trajectories are generated based on the occupied segment sequence, shared margin, and linkage transmission results. The valid identification results include at least the experience identifier, declaration item, call establishment status, and corresponding occupied segment identifier. The occupation trajectory includes at least the occupied segment sequence and the inter-segment connection order. When there are two or more valid paths under the same experience identifier, one is retained according to the rule of the standard path code dictionary order, and the rest are deleted. The output is the valid identification results and occupation trajectory, which are written to the valid results table and trajectory table respectively for the map back-write step to read. The abnormal or missing handling is as follows: when all occupation paths under the same experience identifier are deleted, a conflict result mark is written. When there is a retained path but the standard path code is missing, a replacement code is generated by splicing the segment identifiers of the occupied segment sequence before performing unique retention.

[0105] Next, the valid identification results and occupation trajectories are written back to the graph result layer, forming a unique output result that can be directly called for querying, verification, and subsequent statistics; the input quantities are the valid identification results and occupation trajectories; the processing actions are as follows: locate the result node set and the set of occupancy piece connection edges according to the experience identifier, write the valid identification results into the result nodes of the corresponding experience identifier and the declared item, wherein the written fields of the result nodes include the experience identifier, the declared item, the identification status, and the corresponding occupancy piece identifier; write the occupation trajectories into the connection edges between the corresponding occupancy pieces, wherein the written fields of the connection edges include the previous occupancy piece identifier, the next occupancy piece identifier, and the connection order position;

[0106] Next, the shared surplus is written back to the remaining number of times field of the corresponding shared piece node, and the linkage transmission result is written back to the status field of the corresponding linkage piece edge. After the write-back is completed, the old result nodes and old connection edges that already exist under the same experience identifier are overwritten. The overwriting rule is to delete the old result nodes and old connection edges and then write the current valid identification result and the current occupation trajectory. The output is a unique corresponding valid identification result and its occupation trajectory, which is written to the result output table for the platform query end, review end and statistics end to read directly. The abnormal or missing handling is as follows: when the result node fails to be located, the result node is regenerated according to the experience identifier and the node primary key is filled in. When the connection edge fails to be located, the connection edge is reconstructed according to the occupation piece sequence in the occupation trajectory and then the write-back is performed.

[0107] In practical applications: For example, if the same student uses a competition experience for innovation and entrepreneurship credits, comprehensive quality assessment, and awards simultaneously, the occupancy calculation module outputs two feasible occupancy paths. One path has mutually exclusive edges connecting the comprehensive quality assessment and awards, while the other path, although lacking mutually exclusive edges, shows that the school-level awards are written before the college-level awards. The result generation module first performs conflict cancellation on the two paths, deleting the occupancy path with mutually exclusive edges, and then performs sequence verification on the remaining occupancy paths to confirm that the college-level awards occupy the occupancy piece before the school-level awards. Subsequently, it generates the valid assessment result and occupancy trajectory under this experience identifier, and writes the result nodes corresponding to innovation and entrepreneurship credits, comprehensive quality assessment, and awards into the graph result layer. At the same time, it writes back the remaining number of shared pieces and the status code of linked pieces to the corresponding nodes and edges. Through this implementation process, only one set of queryable, verifiable, and traceable final assessment results is retained under the same experience identifier, avoiding inconsistencies caused by multiple paths coexisting.

[0108] Working Principle: This solution first organizes the experience records, application records, and event records uploaded by students, teachers, and administrators into a unified record set. Based on this, a second-classroom knowledge graph is constructed, uniformly representing the time, role, achievements, proof, and event relationships corresponding to the same experience within the same relational structure. Furthermore, an experience is broken down into shareable, exclusive, and dynamically changing credential parts, and then further divided into occupancy pieces that can be called by specific application events. Subsequently, the system, according to the application order, allows each application event to sequentially call its corresponding occupancy piece, simultaneously processing the remaining uses of shared pieces, the locked status of exclusive pieces, and the impact of linked pieces on the availability of subsequent events. Finally, conflicting paths are deleted, and only one result path with the correct order and valid relationship is retained, generating a unique and valid identification result and its occupancy trajectory. In other words, this solution does not simply reuse a second-classroom experience for multiple events; instead, it first determines which content within the experience can be shared, which can only be used independently, and which will affect subsequent events, and then arrives at a unique and consistent identification result accordingly.

[0109] For example, a student participates in an innovation and entrepreneurship competition. This experience may be used for innovation and entrepreneurship credit recognition, comprehensive quality evaluation, and awards. After the student submits award materials, the teacher submits guidance records, and the administrator submits the rules for the use, mutual exclusion, and sequence of these items, the system first groups this information into the same experience. Then, it determines which content in the competition certificate can support both credits and comprehensive quality evaluation, which content can only be used for awards, and which content must be recognized at the college level before it can be used for university-level recognition. After that, the system calculates item by item according to the application order: which items are valid, which items will conflict with each other, and which items will occupy the subsequent available parts. Finally, it retains only one set of non-conflicting, correctly ordered, and traceable recognition results. In this way, what the administrator sees is no longer several scattered records, but the complete process of how an experience is used among multiple recognition items, how much usable content remains, and why the final result is formed.

[0110] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An AI-driven intelligent management cloud platform for extracurricular activities, characterized in that: include: The data acquisition terminal module is used to receive experience records, application records, and event records uploaded by students, teachers, and administrators. It extracts student identifiers, experience identifiers, time periods, participating roles, achievement content, proof content, application items, application order, concurrent application relationships, mutually exclusive application relationships, and sequential application relationships, and outputs a unified record set. The knowledge graph construction module is used to read a unified record set, using student identifiers, experience identifiers, occurrence time periods, participating roles, output content, proof content, and application items as nodes, and using attribution relationships, pointing relationships, limiting relationships, usage relationships, mutual exclusion relationships, and sequential relationships as edges to construct a second classroom knowledge graph and output experience subgraphs and item subgraphs. The credential expansion module is used to read the experience subgraph and, based on the limiting edges, belonging edges, and pointing edges between the occurrence time node, participating role node, result content node, and proof content node, expand the same extracurricular experience into an experience credential surface composed of shared parts, exclusive parts, and linked parts. The slice generation module is used to read the experience voucher surface and the event sub-graph, slice the experience voucher surface into shared slices, exclusive slices and linked slices, and output the occupied slice graph; The occupancy calculation module is used to read the occupancy map and declaration records, push each declaration item node into the corresponding slice according to the declaration order, perform occupancy locking on exclusive slices, perform remaining share deduction on shared slices, perform subsequent available status rewriting on linked slices, cut off the occupancy path connected by mutual exclusion edges, and output the item occupancy result map and remaining available map. The results generation module is used to read the occupancy result map and the remaining available map of the event, perform conflict cancellation, sequential review and map writing back, and output the unique corresponding valid identification result and its occupancy trajectory.

2. The AI-driven intelligent management cloud platform for the second classroom according to claim 1, characterized in that: The data acquisition terminal module includes: The experience records, application records, and event records uploaded by students, teachers, and administrators are read separately by source. The experience identifier, student identifier, time period, participating role, result content, proof content, application items, application order, concurrent application relationship, mutually exclusive application relationship, and sequential application relationship are extracted according to the type of uploading terminal. Records from different uploading terminals but with the same experience identifier are merged into the same collection group, and the terminal-separated record group is output. For each terminal's separate record group, perform a group-based comparison of the occurrence time, participating roles, results content, and proof content. When the occurrence time is consistent and the results content and proof content are mutually referential, generate an experience master piece. When there is a correspondence between the declared items and the items used together, the items are mutually exclusive, or the items are sequential, generate an item constraint piece. Then, associate the experience master piece and the item constraint piece with the experience identifier and write them into a unified record set. The unified record set performs missing item filling and conflict annotation on each experience main piece and each matter constraint piece. When there are multiple participating roles or multiple proof contents under the same experience identifier, the original correspondence is preserved and written into the position identifier in the piece respectively. When the same declaration matter corresponds to multiple sequential positions, the results are generated in order of upload time and output as a unified record set for the knowledge graph construction module to read.

3. The AI-driven intelligent management cloud platform for the second classroom according to claim 2, characterized in that: The knowledge graph construction module includes: Read the student identifier, experience identifier, occurrence time, participating role, result content, proof content, and application item of each record in the unified record set. Based on the rules of consistent student identifier, overlapping occurrence time, consistent item name in the result content and item name in the proof content, participating role not being a conflicting role limited by the mutual exclusion relationship of items, and the sequential position of the same application item uniquely corresponding to one record, establish record corresponding edges, and group records with record corresponding edges into the same candidate node group, and output candidate node group and candidate relationship group; Based on candidate node groups and candidate relationship groups, each candidate node group is statistically analyzed for time period breaks, role conflicts, result proof mismatches, item sequence reversals, and item co-occurrence conflicts. The candidate node groups are then reconstructed in a fixed order of first splitting and then merging. The splitting rule is to separate records containing any conflict from the original candidate node group, and the merging rule is to merge candidate node groups with consistent student identifiers and no conflict. After each round of reconstruction, candidate relationship groups are regenerated based on experience identifier, occurrence time, participating role, result content, proof content, and application item, until the members of the candidate node groups obtained from two consecutive rounds of reconstruction are completely consistent and the edge items of the candidate relationship groups are completely consistent. A stable node set and a stable edge set are then output.

4. The AI-driven intelligent management cloud platform for the second classroom according to claim 3, characterized in that: The knowledge graph construction module also includes: Based on stable node sets and stable edge sets, consistency checks are performed on each edge from the perspectives of time period, role, result proof, and matter. Specifically, the time period perspective checks whether the occurrence time corresponds to the limiting relationship; the role perspective checks whether the participating roles correspond to the attribution relationship; the result proof perspective checks whether the result content corresponds to the pointing relationship; and the matter perspective checks whether the declared matter corresponds to the concurrent, mutually exclusive, and sequential relationships. Each edge is processed according to the following rules: if the number of supporting items is greater than the number of counter-evidence items, the edge is retained; if the number of supporting items is equal to the number of counter-evidence items, the edge is retained and a conflict flag is added; and if the number of supporting items is less than the number of counter-evidence items, the edge is deleted. Then, experience subgraphs containing attribution, pointing, and limiting relationships are extracted based on experience identifiers, and matter subgraphs containing concurrent, mutually exclusive, and sequential relationships are extracted based on declared matters. For each perspective check, a supported item is recorded if the check is successful, and a counter-evidence item is recorded if the check is unsuccessful.

5. The AI-driven intelligent management cloud platform for the second classroom according to claim 4, characterized in that: The credential expansion module includes: Read the occurrence time node, participating role node, result content node, and proof content node from the experience subgraph. Generate a shared part according to the rule that the same proof content node is connected to two or more result content nodes through pointing edges and falls into the same occurrence time node through limiting edges. Generate an exclusive part according to the rule that the same proof content node is connected to only one result content node through pointing edges and that the result content node corresponds to only one participating role node. Generate a linked part according to the rule that the result content node connected to the same proof content node through pointing edges corresponds to multiple participating role nodes and that any change in any participating role node will cause a change in the affiliation relationship of the remaining result content nodes. Output the initial voucher surface. Based on the initial credential surface, record the corresponding experience identifier, occurrence time node, participating role node, result content node, and proof content node for each shared part, each exclusive part, and each linked part. Then, perform a rearrangement according to the rules that the same proof content node cannot be written into both the shared part and the exclusive part at the same time, the same result content node can only be assigned to one part under the same occurrence time node, and the same participating role node can only retain one ownership chain under the same proof content node, and output the experience credential surface. Based on the experience credential surface, the limiting edge, attribution edge, and pointing edge of each part are checked item by item. The limiting edge is used to check the correspondence between the proof content node and the occurrence time node, the attribution edge is used to check the correspondence between the result content node and the participating role node, and the pointing edge is used to check the correspondence between the proof content node and the result content node. The part where the limiting edge, attribution edge, and pointing edge are all completely corresponding is retained in the experience credential surface, and the part with any missing correspondence is deleted from the experience credential surface. The experience credential surface is then output for the slice generation module to read.

6. The AI-driven intelligent management cloud platform for the second classroom according to claim 5, characterized in that: The slice generation module includes: Based on the shared, exclusive, and linked parts in the experience document surface, as well as the shared edges, mutually exclusive edges, and sequential edges in the item subgraph, each part is read item by item, including its corresponding proof content node, result content node, participating role node, and occurrence time node. The part that has a shared edge connection with two or more application item nodes and whose corresponding proof content node, result content node, participating role node, and occurrence time node are consistent is cut into a shared piece. The part that has a connection with only one application item node and whose application item node is separated from the other application item nodes by a mutually exclusive edge is cut into an exclusive piece. The part that has a sequential edge connection with two or more application item nodes and whose available relationship between the preceding application item node and the subsequent application item node is rewritten is cut into a linked piece. The initial occupied piece group is output. Based on the initial occupied piece groups, write the experience identifier, piece identifier, corresponding declaration item node, corresponding proof content node, corresponding result content node, corresponding participating role node, and corresponding occurrence time node to each shared piece, each exclusive piece, and each linked piece. Then, perform a rearrangement according to the following rules: the same proof content node is assigned to only one shared piece or one exclusive piece under the same occurrence time node; the same result content node is assigned to only one exclusive piece or one linked piece under the same participating role node; and the linked pieces corresponding to the same declaration item node under the same experience identifier are connected end to end in the order of their edges. Output the occupied piece table. Based on the occupancy map, the edge correspondence between each shared map, each exclusive map, and each linked map and the item subgraph is checked item by item. For shared maps, it is checked whether there are shared edges between the corresponding application item nodes. For exclusive maps, it is checked whether there are mutually exclusive edges between the corresponding application item node and other application item nodes. For linked maps, it is checked whether there are sequential edges between the corresponding application item nodes and whether the connection order is consistent with the arrangement order within the map. The complete map corresponding to each edge item and its interconnection relationship are written into the occupancy map, and the occupancy map is output for the pressure occupancy calculation module to read.

7. The AI-driven intelligent management cloud platform for the second classroom according to claim 6, characterized in that: The occupancy calculation module includes: Based on the shared pieces, exclusive pieces, linked pieces, mutually exclusive edges, and sequential edges in the occupied piece map, as well as the application items and application order in the application records, establish a piece matrix, piece continuation matrix, mutually exclusive matrix, and sequential matrix according to the same experience identifier. Multiply the row vector of the piece matrix corresponding to the current application item with the piece continuation matrix to generate a candidate piece sequence. Delete candidate piece sequences with non-zero mutually exclusive matrix mapping values, non-corresponding sequential matrix positions, and duplicate exclusive pieces. Output the candidate path table. Based on the candidate path table, a path matrix, shared surplus vector, exclusive occupancy vector, and linkage transmission vector are established for each candidate piece sequence. Shared deduction, exclusive locking, and linkage transmission are sequentially performed on the path matrix. The path matrix is ​​then multiplied by the item piece matrix to obtain the item expansion matrix. Singular value decomposition is performed on the item expansion matrix to extract the column space corresponding to non-zero singular values. Orthogonal projection is performed on the column space to generate a standard path code. One candidate path with the same standard path code and consistent shared surplus vector, exclusive occupancy vector, and linkage transmission vector is retained. After each declared item is written sequentially, the process of shared deduction, exclusive locking, linkage transmission, multiplying the path matrix and item piece matrix to obtain the item expansion matrix, performing singular value decomposition on the item expansion matrix to extract the column space corresponding to non-zero singular values, performing orthogonal projection on the column space to generate a standard path code, and retaining candidate paths is repeated until the standard path code set, shared surplus vector set, exclusive occupancy vector set, and linkage transmission vector set obtained in two adjacent rounds are consistent. A stable path table is then output.

8. The AI-driven intelligent management cloud platform for the second classroom according to claim 7, characterized in that: The occupancy calculation module also includes: Based on the stable path table, a pressure check matrix, a surplus check matrix, and a transfer check matrix are constructed for each stable path. The pressure check matrix is ​​multiplied by the item piece matrix to check the item correspondence. The surplus check matrix is ​​checked against the shared surplus vector position by position to check the shared deduction result. The transfer check matrix is ​​checked against the linkage transfer vector position by position to check the linkage transfer result. The stable paths that have been checked are written into the item pressure result graph, and their corresponding shared surplus vector and linkage transfer vector are written into the remaining available graph.

9. The AI-driven intelligent management cloud platform for the second classroom according to claim 8, characterized in that: The result generation module includes: Based on the occupancy result diagram and the remaining available diagram, the occupancy paths, occupancy sequence, shared reserve and linkage transmission results corresponding to each declared item are collected according to the same experience identifier. Conflict cancellation is performed on each occupancy path under the same experience identifier. Occupancy paths with mutually exclusive edge connections, duplicate deductions after writing back the shared reserve, and linkage transmission results that do not correspond to the order of the edges are deleted. The pending result group is output. Based on the pending results group, the order of each occupied path retained under the same experience identifier is reviewed. The order of the declared items in each occupied path is matched with the order of the occupied area sequence, and the occupied path corresponding to each position is retained as a valid path. The occupied path with the reversed order is deleted. Then, the valid identification result and the occupied trajectory are generated according to the occupied area sequence, shared reserve and linkage transmission result in the valid path. Based on the valid identification results and the occupation trajectory, the valid identification results are written into the result nodes of the corresponding experience identifier and the declared items, the occupation trajectory is written into the connection edges between the corresponding occupied pieces, and the shared surplus and linkage transmission results are written back to the corresponding nodes and corresponding edges, outputting a unique valid identification result and its occupation trajectory.

Images