Intelligent recommendation scheduling method and system for training resources

Abstract

The invention provides an intelligent recommendation scheduling method and system for training resources, and relates to the technical field of data processing, and the method comprises the steps: obtaining a position data set and a resource data set, determining a subject association relationship and a position mapping relationship according to the position data set and the resource data set, extracting subject constraint information and position dependence information, performing time matching on the training demand duration and the resource available duration of each subject to obtain a candidate scheduling pair group set, performing resource conflict detection, removing conflict pair groups in the candidate scheduling pair group set, sorting the candidate scheduling pair groups according to position dependence information, extracting a candidate resource set corresponding to each subject, calculating a score value of each candidate resource to obtain a resource score set, and performing resource scheduling on the candidate resource set according to the score value of each candidate resource. And selecting the resource identifier with the maximum score value in each subject as a recommended resource to obtain scheduling plan data. According to the invention, the resource scheduling can be comprehensively judged under the condition of limited resources.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to an intelligent recommendation and scheduling method and system for training resources. Background Art

[0002] In the existing basic training system, the scheduling of training resources usually depends on manually formulating training plans and subject arrangements. After importing the training content of the current level based on the superior plan, it is necessary to manually formulate an implementation plan, match the subjects according to positions, and configure the training activities one by one. The arrangement of subjects mainly follows the training syllabus, and a training resource system is established from multiple dimensions such as test papers, samples, and teaching plans in combination with the training requirements of positions. Although some systems have introduced a subject priority management mechanism, the basis for resource allocation is still mainly static tables, lacking dynamic adaptability and intelligent recommendation capabilities. Especially in the case of diverse training resources and complex position intersections, it may be difficult to achieve accurate and rapid scheduling responses.

[0003] For example, in the automatic generation of subject plans, existing solutions mostly rely on simplified priority algorithms, using elements such as the start and end times of subjects and the daily class hours as the basic basis for scheduling, and adopting rule-based logic to advance the training schedule day by day. However, such methods generally ignore the analysis of resource conflicts between multiple positions and the mutual dependence of training tasks, and are prone to problems such as subject conflicts, uneven training intensity, and repeated resource allocation. In the case of overlapping position coverage, the same resource may be contended by multiple subjects, and the system may lack a comprehensive judgment and dynamic adjustment mechanism for resource priorities. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent recommendation and scheduling method and system for training resources, aiming to solve the problems mentioned in the background art.

[0005] To solve the above technical problems, the technical solution of the present invention is as follows:

[0006] In a first aspect, an intelligent recommendation and scheduling method for training resources, the method includes:

[0007] Obtain a position data set and a resource data set, determine the subject association relationship and position mapping relationship according to them, extract subject constraint information and position dependence information, and obtain a constraint data set;

[0008] According to the constraint data set, perform time matching on the training demand duration and resource available duration of each subject to obtain a set of candidate scheduling pair groups;

[0009] Perform resource conflict detection on the set of candidate scheduling pair groups to determine whether there is the same resource identifier corresponding to different subjects and time overlap, and obtain a conflict filtering set;

[0010] Remove the conflicting pair groups according to the conflict filter set, use the remaining candidate scheduling pair groups as the valid scheduling set, and sort them according to the position dependency information to obtain a sorted data sequence;

[0011] Extract the candidate resource sets corresponding to each subject according to the sorted data sequence, and calculate the scoring values of each candidate resource to obtain a resource scoring set;

[0012] Select the resource identifier with the largest scoring value in each subject as the recommended resource according to the resource scoring set, and extract its time period label and position number in the valid scheduling set to obtain scheduling plan data.

[0013] Furthermore, obtain the position data set and the resource data set, determine the subject association relationship and the position mapping relationship according to them, and extract the subject constraint information and the position dependency information to obtain a constraint data set, including:

[0014] Obtain the position numbers of all positions, the subject sets corresponding to each position, and the training requirement durations of each subject to obtain a position data set; obtain the resource numbers of all resources, the available resource durations, the subject types supported by the resources, and the position identifiers to which the resources belong to obtain a resource data set;

[0015] Determine the occurrence of each subject in different positions according to the position data set, and extract the continuous training relationship and non-parallel constraint between subjects to obtain subject association relationship data;

[0016] Determine the matching structure between the position identifier and the subject set according to the position data set to obtain position mapping relationship data;

[0017] Extract the sequence relationship, mutual exclusion execution flag, and training stage overlap relationship between each subject and other subjects according to the subject association relationship data to obtain subject constraint information data;

[0018] Analyze the sharing and preemption situations between different positions in the resource dimension according to the position mapping relationship data and the position identifiers to which the resources belong in the resource data set to obtain position dependency information data;

[0019] Merge the subject constraint information data and the position dependency information data to obtain a rule information set, and extract the subject identifiers, position numbers, time conditions, and execution relationship labels in it to obtain a constraint data set.

[0020] Furthermore, perform time matching on the training requirement durations of each subject and the available resource durations according to the constraint data set to obtain a candidate scheduling pair group set, including:

[0021] Obtain the subject identifier, total training duration, daily training duration requirement, and flag indicating whether segmented training is allowed to obtain the subject duration requirement data; obtain the resource identifier, continuously available time period, and set of subject identifiers supported by the resource to obtain the resource time segment data;

[0022] According to the subject constraint information in the constraint dataset, filter out other subjects that have a mutually exclusive relationship with it during the target subject arrangement to obtain a set of matchable subjects;

[0023] According to the resource time segment data, extract all the resource identifiers required for each subject identifier, and combine the continuously available time periods of each resource to obtain a set of resource schedulable windows;

[0024] For each subject in the subject duration requirement data, combine the flag indicating whether segmented training is allowed, and sequentially select a set of time periods that can continuously cover the total training duration from the set of resource schedulable windows to obtain time matching items;

[0025] For each time matching item, determine the subject identifier, matching resource identifier, proposed start time, proposed end time, and daily training duration value, generate candidate records for the scheduling pair group, and summarize all the candidate records for the scheduling pair group to obtain a set of candidate scheduling pair groups.

[0026] Further, perform a resource conflict detection on the set of candidate scheduling pair groups to determine whether there is a situation where different subjects correspond to the same resource identifier and the time overlaps, and obtain a conflict filtering set, including:

[0027] Group the candidate records for all the scheduling pair groups in the set of candidate scheduling pair groups by resources, cluster all the scheduling pair groups according to the resource identifier to obtain a set of resource pair groups;

[0028] Analyze any two scheduling pair groups in the set of resource pair groups to determine whether there is an overlapping interval between their corresponding start time and end time. When the result is yes, mark it as a time overlapping pair group;

[0029] According to the time overlapping pair groups, further determine whether their corresponding subject identifiers are different. When the result is yes, mark the pair group as a conflict pair group to obtain a conflict identification list;

[0030] Remove each scheduling pair group in the conflict identification list from the set of candidate scheduling pair groups, and use the remaining part of the set of candidate scheduling pair groups as a set of resource conflict-free pair groups;

[0031] Merge the conflict identification list and the set of resource conflict-free pair groups, and mark them as a conflict scheduling subset and a non-conflict scheduling subset respectively to obtain a conflict filtering set.

[0032] Further, according to the conflict filtering set, remove the conflicting pair groups therein, use the remaining candidate scheduling pair groups as the effective scheduling set, and sort them according to the position dependency information to obtain a sorted data sequence, including:

[0033] According to the conflict filtering set, remove the scheduling pair groups marked as conflict scheduling subsets from the candidate scheduling pair group set, and retain the remaining scheduling pair groups to obtain an effective scheduling set;

[0034] According to the effective scheduling set, extract the subject identifier and position number information in each scheduling pair group to construct subject-position comparison data;

[0035] According to the position dependency information, determine the set of precedence relationships between different positions to form a position dependency graph;

[0036] Map each position number in the subject-position comparison data to the position dependency graph to establish a transitive dependency relationship between subjects and obtain a subject scheduling directed graph;

[0037] Sort the subject scheduling directed graph, determine the execution order list of subjects according to the direction of the dependency edges in the graph, and obtain a sorted path sequence;

[0038] Map the subject identifiers in the sorted path sequence back to the corresponding scheduling pair groups in the effective scheduling set and arrange them in order to obtain a sorted data sequence.

[0039] Further, according to the sorted data sequence, extract the candidate resource sets corresponding to each subject, and calculate the scoring values of each candidate resource to obtain a resource scoring set, including:

[0040] According to each subject in the sorted data sequence, extract all the resources associated with it from the effective scheduling set to form a course candidate resource set and obtain subject-resource mapping data;

[0041] According to the subject-resource mapping data, for each candidate resource, count the number of times it appears in the candidate resource set to obtain the resource appearance times, calculate the proportion of it that has been selected by other subjects currently, extract the proportion of the usage duration of this resource by the current subject, and calculate the resource usage frequency term through the interaction of the three;

[0042] For each resource, count the set of position numbers it can serve, extract the subject coverage ratio under each position, and normalize it based on the graph depth of the position node in the position dependency graph to obtain the position priority weight, and calculate the position coverage frequency term according to the subject coverage ratio and the position priority weight;

[0043] For each subject in the sorted data sequence, record its position in the sorting, combine it with the total number of subjects to calculate its priority coefficient, and at the same time extract its completed training duration and planned total training duration to calculate the subject completion term;

[0044] Perform a weighted sum of the resource usage frequency item, the position coverage frequency item, and the subject completion item, calculate the scoring value of different resources for each subject, and form a scoring record by combining each scoring value with its corresponding subject identifier and resource identifier to obtain a resource scoring set.

[0045] Furthermore, according to the resource scoring set, select the resource identifier with the largest scoring value in each subject as the recommended resource, and extract its time period label and position number in the effective scheduling set to obtain scheduling plan data, including:

[0046] For each subject identifier, filter out the resource identifier with the highest scoring value from the resource scoring set and determine it as the recommended resource for the subject;

[0047] In the effective scheduling set, find the scheduling pair group corresponding to the subject and its recommended resource, and extract the time period label in this scheduling pair group as the training start and end time;

[0048] Extract the corresponding position number from this scheduling pair group as the associated position identifier of the subject;

[0049] According to the subject identifier, recommended resource identifier, training start and end time, and position number, obtain a resource allocation detail record, and summarize all the resource allocation detail records to obtain scheduling plan data.

[0050] In a second aspect, an intelligent recommendation scheduling system for training resources, the system includes:

[0051] A constraint module, configured to obtain a position data set and a resource data set, determine a subject association relationship and a position mapping relationship according to them, extract subject constraint information and position dependency information, and obtain a constraint data set;

[0052] A matching module, configured to perform time matching on the training requirement duration and resource available duration of each subject according to the constraint data set to obtain a candidate scheduling pair group set;

[0053] A detection module, configured to perform resource conflict detection on the candidate scheduling pair group set, and determine whether there is the same resource identifier corresponding to different subjects and time overlap, to obtain a conflict filtering set;

[0054] A sorting module, configured to remove the conflicting pair groups from the conflict filtering set according to the conflict filtering set, use the remaining candidate scheduling pair groups as an effective scheduling set, and sort it according to the position dependency information to obtain a sorted data sequence;

[0055] A scoring module, configured to extract the candidate resource set corresponding to each subject according to the sorted data sequence, and calculate the scoring value of each candidate resource to obtain a resource scoring set;

[0056] The scheduling module is used to select the resource identifier with the largest score value in each subject as the recommended resource according to the resource score set, and extract its time period label and position number in the effective scheduling set to obtain the scheduling plan data.

[0057] The above solution of the present invention has at least the following beneficial effects:

[0058] By obtaining the position data set and the resource data set, the system of the present invention establishes the mapping relationship between the subjects and the positions, and extracts information such as the training duration of the subjects and the available time of the resources to form a complete constraint data set. In the scheduling stage, the system accurately matches the required duration of the resources by the subjects with the time window of the continuous availability of the resources to ensure that each training task can be completed within the required time, avoiding the situation of forced allocation when the resources do not fully meet the training requirements. This matching mechanism supports the flexible scheduling of training tasks under the premise of meeting the conditions, and at the same time combines the setting of whether the subject supports segmented training to achieve configurable scheduling granularity. In the case of limited resource configuration or intensive training, this method can significantly improve the scheduling success rate of the system and the completion quality of training tasks, and enhance the overall response ability of the training system.

[0059] By setting up a special resource conflict detection mechanism, the present invention judges the resource overlap of the candidate scheduling pair group. Especially when it is detected that more than two subjects request the same resource and there is an overlapping interval in time, this group of scheduling schemes is identified as a conflict pair group and eliminated. Different from the prior art, this mechanism not only relies on static rule judgment, but also combines the time label and resource identifier generated in the scheduling process for real-time analysis. The result of such processing is that the system can dynamically screen out unreasonable scheduling options, ensuring that each resource is only occupied by a single subject at the same time period, thus avoiding the common resource "preemption" problem in traditional training scheduling and enhancing the stability of the system and the controllability during actual execution. This operation is particularly crucial for training tasks with tight position resources and high subject parallelism, ensuring the operability of the scheduling plan during implementation.

[0060] The present invention constructs a position dependence graph through position dependence information, converts it into a directed graph of subject scheduling, and then performs topological sorting to obtain the subject sorting path. This sorting logic not only considers the training sequence within a single position, but also integrates the priority and dependence relationship between different positions. For example, some subjects can only be executed after specific basic subjects are completed. This mechanism effectively solves the problem of chaotic training order caused by the lack of task dependence judgment in traditional scheduling systems, ensuring that the scheduling result has a rigorous logical closed loop and training rationality. By driving the scheduling execution with the sorted data, the system can guide the subjects to be activated at the correct time point, ensuring the continuity of the training process and the integrity of the connection of position tasks, thus realizing a more scientific training organization method and improving the coordination and docking degree of the overall training system.

[0061] By introducing a resource scoring mechanism, the present invention comprehensively scores each resource from multiple dimensions, including resource usage frequency, position coverage frequency, and subject completion rate. The resource usage frequency reflects the current utilization rate of the resource in scheduling. The position coverage frequency takes into account the adaptability of the resource in multiple positions and its priority weight in the position dependency graph. The subject completion rate evaluates the proportion of the remaining training duration of the subject. The system sums up these three indicators with weights to generate a score value, which is used to select the most suitable resource for the subject. Compared with the static weight or priority decision logic, this scoring mechanism fully embodies the principle of optimizing resource scheduling in a dynamic environment, achieving a balance among reasonable resource allocation, maximizing utilization, and ensuring training objectives. The final recommendation result not only has a high matching degree but also has stronger interpretability and goal orientation, providing a solid decision-making basis for the intelligent scheduling system.

[0062] The present invention can support the segmented execution of training subjects. During the scheduling phase, it can identify whether a subject is allowed to be split into multiple training segments for execution and accordingly match the continuous available time periods of resources. This design greatly improves the flexibility of resource scheduling, enabling the system to still maximize the completion of training task allocation under the conditions of fragmented resources or local conflicts, and alleviating the impact of resource tension on the overall scheduling success rate. In the case of uneven time distribution of training subjects or a high degree of position intersection, the system can flexibly splice low-frequency idle resources through segmented training, improve resource utilization, and reduce training idle time. At the same time, for subjects that do not support segmented training, a complete time window can be forcibly matched to ensure training continuity. This differential processing strategy fully adapts to various complex training modes and management requirements, improving the system's adaptability to diverse scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 is a flowchart of an intelligent recommendation scheduling method for training resources provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0064] Hereinafter, exemplary embodiments of the present disclosure will be described in more detail with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

[0065] As Figure 1 shown, an embodiment of the present invention proposes an intelligent recommendation scheduling method for training resources, and the method includes:

[0066] Obtain the position dataset and the resource dataset, determine the subject association relationship and the position mapping relationship according to them, extract the subject constraint information and the position dependency information, and obtain the constraint dataset;

[0067] According to the constraint dataset, perform time matching on the training demand duration and the resource available duration of each subject to obtain a set of candidate scheduling pair groups;

[0068] Perform resource conflict detection on the set of candidate scheduling pair groups, and determine whether there is the same resource identifier corresponding to different subjects and time overlap, to obtain a conflict filtering set;

[0069] According to the conflict filtering set, remove the conflicting pair groups therein, use the remaining candidate scheduling pair groups as the effective scheduling set, and sort them according to the position dependency information to obtain a sorted data sequence;

[0070] According to the sorted data sequence, extract the candidate resource sets corresponding to each subject, and calculate the score values of each candidate resource to obtain a resource score set;

[0071] According to the resource score set, select the resource identifier with the largest score value in each subject as the recommended resource, and extract its time period label and position number in the effective scheduling set to obtain the scheduling plan data.

[0072] In the embodiment of the present invention, the position dataset and the resource dataset are obtained, the subject association relationship and the position mapping relationship are determined according to them, the subject constraint information and the position dependency information are extracted to obtain the constraint dataset, and through systematic extraction of structured data, a comprehensive dependency and constraint logic between subjects and positions is formed, effectively supporting the formulation of subsequent scheduling strategies;

[0073] According to the constraint dataset, perform time matching on the training demand duration and the resource available duration of each subject to obtain a set of candidate scheduling pair groups, ensure that each training task can be reasonably arranged on the resource time, reduce scheduling holes and fragmentation, and improve resource utilization;

[0074] Perform resource conflict detection on the set of candidate scheduling pair groups, and determine whether there is the same resource identifier corresponding to different subjects and time overlap, to obtain a conflict filtering set, and pre-check the resource overlap problem before scheduling decisions, avoid the situation of repeated use of resources in the same time period, so as to avoid the scheduling failure or interruption caused by resource conflicts in actual training execution, and improve the execution reliability of the scheduling plan;

[0075] According to the conflict filtering set, remove the conflicting pair groups therein, take the remaining candidate scheduling pair groups as the effective scheduling set, and sort them according to the position dependency information to obtain a sorted data sequence, ensuring that the execution of training subjects conforms to the dependency order requirements in the organizational management structure, avoiding cross - chaos in the training process, improving the systematicness, scientificity and dependency controllability of the training process, and enhancing the internal consistency of the scheduling logic;

[0076] According to the sorted data sequence, extract the candidate resource sets corresponding to each subject, and calculate the scoring values of each candidate resource to obtain a resource scoring set, making the resource recommendation process more scientific and interpretable, avoiding "blind selection" of resources or over - reliance on static priorities. In the context of multi - subject competition for resources, the optimal choice can be made according to the scoring values, improving the task fitness of the scheduling result and the overall resource utilization efficiency of the system;

[0077] According to the resource scoring set, select the resource identifier with the largest scoring value in each subject as the recommended resource, and extract its time period label and position number in the effective scheduling set to obtain the scheduling plan data, completing the closed - loop output of the entire scheduling chain, providing a data basis for the informatization construction of the training management system, and ensuring that the output scheduling result has high adaptability, high execution feasibility and traceability.

[0078] In a preferred embodiment of the present invention, obtain a position data set and a resource data set, and determine the subject association relationship and position mapping relationship according to them, extract subject constraint information and position dependency information to obtain a constraint data set, including:

[0079] Obtain the position numbers of all positions, the set of subjects corresponding to each position, and the training requirement duration of each subject to obtain the position data set; obtain the resource numbers of all resources, the available duration of resources, the resource - supported subject types, and the position identifiers to which the resources belong to obtain the resource data set;

[0080] According to the position data set, determine the occurrence of each subject in different positions, and extract the continuous training relationship and non - parallel constraint between subjects to obtain subject association relationship data;

[0081] According to the position data set, determine the matching structure between the position identifier and the subject set to obtain position mapping relationship data;

[0082] According to the subject association relationship data, extract the sequential relationship, mutually exclusive execution flag, and training stage overlap relationship between each subject and other subjects to obtain subject constraint information data;

[0083] According to the position mapping relationship data and the position identifiers to which the resources belong in the resource data set, analyze the sharing and preemption situations between different positions in the resource dimension to obtain position dependency information data;

[0084] Merge the subject constraint information data and the position dependency information data to obtain a rule information set, and extract the subject identifier, position number, time condition, and execution relationship label from it to obtain a constraint data set.

[0085] In the embodiment of the present invention, obtain the position numbers of all positions, the set of corresponding subjects for each position, and the training requirement duration of each subject to obtain a position data set, which provides a direct mapping relationship between positions and subjects, helps to quickly locate the position training requirements, and thus supports the generation of scheduling objectives; obtain the resource numbers of all resources, the available duration of resources, the types of subjects supported by resources, and the position identifiers to which the resources belong to obtain a resource data set, which establishes a training support constraint between subjects and resources to avoid problems such as resource mismatch or unavailable training resources; according to the position data set, determine the occurrence of each subject in different positions, and extract the continuous training relationship and non-parallel constraint between subjects to obtain subject association relationship data. By clearly defining the execution constraints of subjects, it is possible to avoid the generation of invalid or conflicting scheduling plans in the subsequent time matching and sorting stages, ensure the order and dependency consistency of task execution, and improve the controllability of the scheduling process; according to the position data set, determine the matching structure between the position identifier and the subject set to obtain position mapping relationship data, which reveals the cross-position distribution of subjects, and can determine whether the training resources have the potential for reuse in multiple positions, providing support for subsequent resource conflict detection and sorting scheduling.

[0086] According to the subject association relationship data, extract the sequential relationship, mutually exclusive execution flag, and training stage overlap relationship between each subject and other subjects to obtain subject constraint information data, explicitly model all abstract rules, enhance the decision-making accuracy of the scheduling algorithm, and avoid scheduling failures or unachieved training objectives caused by hidden constraints; according to the position mapping relationship data and the position identifiers to which the resources belong in the resource data set, analyze the sharing and preemption situations of resources among different positions to obtain position dependency information data, which can understand the resource relationships across positions, give priority to meeting the requirements of positions with high dependency intensity, reduce scheduling conflicts between positions, and improve the fairness of resource allocation and the overall scheduling quality of the system; merge the subject constraint information data and the position dependency information data to obtain a rule information set, and extract the subject identifier, position number, time condition, and execution relationship label from it to obtain a constraint data set, which integrates and abstracts the analysis results of multiple dimensions into a unified scheduling control input, improves the organization and automation of the entire scheduling system, and supports resource scheduling in concurrent, multi-objective, and high-complexity scenarios.

[0087] Among them, according to the position data set, determine the occurrence of each subject in different positions, and extract the continuous training relationship and non-parallel constraint between subjects to obtain subject association relationship data, which specifically includes:

[0088] First, by traversing each job record in the job dataset, count the frequency of each subject in each job and establish a subject-job reverse mapping relationship to identify whether there is the same set of subjects in multiple jobs. The system analyzes the combination order of multiple subjects within the same job. If it is found that subject A and subject B are in a fixed order in most jobs (for example, A is always before B), then record that there is a "continuous training relationship" between them; at the same time, if certain subjects are clearly defined as not schedulable simultaneously in the training rules (such as due to resource conflicts or repeated training stages), or they have never had overlapping schedules in historical tasks, the system identifies them as "non-parallel subjects" and marks them as mutually exclusive pairs. In addition, to improve the accuracy of the model, prior rules or expert rules can be further introduced. For example, referring to the training syllabus, define subject pairs with basic dependency relationships as belonging to an "order execution group", and subject pairs using the same type of key resources are marked as a "non-parallel group". Finally, the system structures and outputs all subject pairs with continuity or mutual exclusion characteristics to form subject association relationship data, which serves as the basis for subsequent subject scheduling order judgment and mutual exclusion screening mechanisms.

[0089] Among them, according to the subject association relationship data, extract the sequential relationship, mutual exclusion execution flag, and training stage overlap relationship between each subject and other subjects to obtain subject constraint information data, specifically including:

[0090] Based on the previously constructed subject association relationship data, establish a subject dependency graph and perform semantic classification and labeling on the graph structure, further refining it into three core constraint information: sequential relationship, mutual exclusion execution, and stage overlap relationship. For the sequential relationship, the system analyzes the existence of a directed path structure of "preceding → succeeding" in the subject dependency path, automatically marks that subject A must be executed before subject B, and records its priority level; for the mutual exclusion execution flag, the system checks the subject pairs with mutual exclusion marks. If there is a physical overlap in the resources used by the two, a conflict in the training tasks in the same time period, or they are restricted in their definitions from being carried out simultaneously, then clearly mark this subject as "not schedulable simultaneously"; for the stage overlap relationship, the system learns from previous training records or identifies through a rule engine to determine which subjects, although there is no conflict in the training content, may have overlapping time periods. For example, one subject only has theoretical lectures in the first two hours, and another subject conducts practical exercises in the last two hours. Since they do not overlap at the resource level, they can be marked as "stage-overlappable". The system uniformly outputs these three types of information in the form of structured constraint entries. Each entry includes subject identifier, associated subject identifier, constraint type, constraint direction (such as whether it is unidirectional dependency), and time reference information, and finally forms subject constraint information data, providing key logical conditions for subsequent steps such as resource-time matching and scheduling order sorting.

[0091] Among them, according to the position mapping relationship data and the position identifiers of the resources in the resource dataset, analyze the sharing and preemption situations among different positions in terms of resources, and obtain the position dependency information data, which specifically includes:

[0092] First, based on the position mapping relationship data, determine in which positions there is a training demand for each subject. Then, in combination with the resource dataset, extract the position identifier to which each resource belongs and the set of subjects it supports, so as to determine the attribution relationship between resources and positions. By comparing the subjects supported by the resources with their affiliated position identifiers, as well as the structure of the subject distribution in different positions, the system identifies which positions have cross-demand for a certain resource. If the system finds that two or more positions schedule the same resource and there is a potential risk of overlapping scheduling periods, it is determined that there is a "resource preemption" problem among these positions, and the system establishes an edge of resource contention relationship between these positions. For those cases where the resource serves only a single position or the service subjects do not overlap, they are marked as resource isolation states. Further, the system can construct a position priority index based on the position training level, position function priority, or training task urgency, map it to the edges of the resource contention graph in the form of weights, and generate a position dependency graph. In this dependency graph, nodes represent positions, edges represent resource sharing relationships, the directionality of the edges represents the resource allocation priority, and the edge weights can be used as sorting indicators. Finally, the system structures and outputs all the dependency paths formed by resource intersections among positions as position dependency information data, including fields such as position pairs, resource intersections, contention probabilities, priority level labels, etc. This information data will be used to determine the task sequence and scheduling urgency judgment during the scheduling and sorting stage.

[0093] In a preferred embodiment of the present invention, according to the constraint dataset, perform time matching on the training demand duration of each subject and the available duration of resources to obtain a set of candidate scheduling pairs, including:

[0094] Obtain the subject identifier, total training duration, daily training duration requirement, and whether segmented training is allowed flag to obtain the subject duration demand data; obtain the resource identifier, continuous available time period, and the set of subject identifiers supported by the resource to obtain the resource time segment data;

[0095] According to the subject constraint information in the constraint dataset, screen out other subjects that have a mutually exclusive relationship with it during the target subject arrangement to obtain a set of matchable subjects;

[0096] According to the resource time segment data, extract all the resource identifiers required by each subject identifier, and in combination with the continuous available time periods of each resource, obtain the resource schedulable window set;

[0097] For each subject in the subject duration requirement data, in combination with the flag indicating whether segmented training is allowed, sequentially select a set of time periods that continuously cover the total training duration from the resource schedulable window set to obtain time matching items;

[0098] For each time matching item, determine the subject identifier, matching resource identifier, proposed start time, proposed end time, and daily training duration value, generate candidate records for the scheduling pair group, and summarize all candidate records for the scheduling pair group to obtain the candidate scheduling pair group set.

[0099] In the embodiment of the present invention, obtaining the subject identifier, total training duration, daily training duration requirement, and the flag indicating whether segmented training is allowed to obtain the subject duration requirement data realizes the structured and standardized processing of the subject training requirements and provides an input basis for subsequent time matching; obtaining the resource identifier, continuous available time periods, and the set of subject identifiers supported by the resource to obtain the resource time segment data standardizes the expression of the time dimension information of resource usage and lays a data foundation for subsequent matching processing based on the time window; screening out other subjects that have a mutually exclusive relationship with the target subject during the arrangement of the target subject according to the subject constraint information in the constraint dataset to obtain the set of matchable subjects, and ensuring the correctness of the scheduling logic and providing a legal input set for subsequent candidate resource screening and time window calculation by eliminating subjects that do not meet the scheduling conditions based on the execution constraints, avoiding logical conflicts between subjects from the source and improving the scheduling feasibility; extracting all resource identifiers required for each subject identifier according to the resource time segment data and combining the continuous available time periods of each resource to obtain the resource schedulable window set, transforming the resource from an abstract entity into a measurable time slice, realizing the transformation of the scheduling problem from "resource allocation" to "time block filling", providing an operable search space for subsequent specific matching, and enhancing the scheduling flexibility and resource utilization efficiency; for each subject in the subject duration requirement data, in combination with the flag indicating whether segmented training is allowed, sequentially select a set of time periods that continuously cover the total training duration from the resource schedulable window set to obtain time matching items, realizing the precise matching between the subject requirements and the resource capabilities, providing multiple scheduling options on the premise of ensuring legal logic, and leaving sufficient optimization space for the subsequent scoring and conflict judgment stages of the system; for each time matching item, determine the subject identifier, matching resource identifier, proposed start time, proposed end time, and daily training duration value, generate candidate records for the scheduling pair group, and summarize all candidate records for the scheduling pair group to obtain the candidate scheduling pair group set. By centrally managing all scheduling candidate solutions, the system can provide a high-quality input data source for subsequent operations, improving the coherence of the overall scheduling logic and the modularity of system execution.

[0100] Among them, screening out other subjects that have a mutually exclusive relationship with the target subject during the arrangement of the target subject according to the subject constraint information in the constraint dataset to obtain the set of matchable subjects specifically includes:

[0101] First, an estimation range of the scheduling time period is established for each target subject. Based on the total training duration of the subject and the daily training duration requirement, the minimum required time span is estimated, and combined with the windows in the resource time segments that may support the subject, an initial candidate set of arranged time periods is formed. Subsequently, the system extracts the constraint information related to the subject from the constraint dataset, specifically including the mutually exclusive execution flag, the sequential dependency label, the phase overlap identifier, and other contents. By parsing these labels, the system identifies the set of other subjects that cannot be scheduled simultaneously with this subject in terms of time. For example, when there is a mutually exclusive flag between two subjects, or the training phases to which they belong have non-overlapping attributes, or it is clearly marked that they need to be executed sequentially (for example, subject B must be after subject A), then this subject is temporarily excluded from the candidate space for parallel scheduling. The system analyzes and screens all subjects with such constraint relationships one by one to form a set of subjects that can theoretically coexist within the pre-arranged time period of the target subject Ti, which is called the set of matchable subjects. This set will be used as the input for the next resource scheduling match to ensure that there are no logical conflicts between subjects in the subsequent steps. During the whole process, each elimination has a constraint basis for tracing, and the finally formed set has logical closure and scheduling security.

[0102] Among them, for each subject in the subject duration requirement data, combined with the flag indicating whether segmented training is allowed, a set of time periods that can continuously cover the total training duration is sequentially selected from the resource schedulable window set to obtain time matching items, specifically including:

[0103] The system traverses each subject T in the subject duration requirement data, extracts its total training duration, the minimum daily training duration, and the flag indicating whether segmented training is allowed, and screens the set of relevant resources R in the resource schedulable window set. For each resource R, the system obtains its list of time segments, and each time segment includes a start time and an end time. The system first judges whether the length of the time segment is greater than or equal to the minimum daily training duration to ensure that the basic daily training requirements are met, and then makes a policy branch judgment based on the flag indicating whether segmented training is allowed.

[0104] If the flag indicating whether segmented training is allowed is no (i.e., segmented training is not allowed), the system only retains those time segments whose single time period length is greater than or equal to the minimum daily training duration as available candidates; for these time periods, the system directly generates scheduling candidates, which include the subject, the resource, the start time, and the end time, etc., indicating that the subject can complete the training goal by this resource in this time period at one time.

[0105] If the flag indicating whether segmented training is allowed is yes (i.e., segmented training is allowed), the system combines all available time periods of the resources. Using a greedy strategy or dynamic programming method, it starts from the forward traversal to select time periods and accumulates the training duration until the accumulated time reaches or exceeds the total training duration. During the combination process, it is also necessary to ensure that the length of each selected time period is not less than the total training duration to avoid overly fragmented training. After the combination is completed, T is bound to the combined time period sequence and R to form a set of scheduling matching items. Each time period combination is encapsulated as a candidate record for a scheduling pair group, including fields such as resource identifier, subject identifier, start and end time sequence, and daily training arrangement, forming multiple sets of scheduling matching paths for subsequent scoring and screening.

[0106] In a preferred embodiment of the present invention, resource conflict detection is performed on the candidate scheduling pair group set to determine whether there is time overlap for the same resource identifier corresponding to different subjects, and a conflict filtering set is obtained, including:

[0107] Resource grouping is performed on all candidate records of the scheduling pair groups in the candidate scheduling pair group set. All scheduling pair groups are clustered according to the resource identifier to obtain a resource pair group set;

[0108] Any two scheduling pair groups in the resource pair group set are analyzed to determine whether there is an overlapping interval between their corresponding start times and end times. When the result is yes, it is marked as a time-overlapping pair group;

[0109] Based on the time-overlapping pair groups, it is further determined whether their corresponding subject identifiers are different. When the result is yes, the pair group is marked as a conflict pair group to obtain a conflict identifier list;

[0110] Each scheduling pair group in the conflict identifier list is removed from the candidate scheduling pair group set, and the remaining part of the candidate scheduling pair group set is used as a resource conflict-free pair group set;

[0111] The conflict identifier list and the resource conflict-free pair group set are merged and marked as a conflict scheduling subset and a non-conflict scheduling subset respectively to obtain a conflict filtering set.

[0112] In an embodiment of the present invention, resource grouping is performed on all scheduling pair group candidate records in the candidate scheduling pair group set, and all scheduling pair groups are clustered according to resource identifiers to obtain a resource pair group set. By clustering through resource identifiers, potential time conflict problems are localized in advance, reducing the computational burden of irrelevant data; any two scheduling pair groups in the resource pair group set are analyzed to determine whether there is an overlapping interval between their corresponding start time and end time. When the result is yes, it is marked as a time-overlapping pair group, which can accurately identify the conflict situation where the same resource is concurrently requested in multiple subjects, avoiding the common problem of duplicate resource allocation in traditional training systems; according to the time-overlapping pair group, it is further determined whether their corresponding subject identifiers are different. When the result is yes, the pair group is marked as a conflict pair group to obtain a conflict identifier list, ensuring that the system only marks the scheduling pair groups that truly constitute resource competition, enhancing the intelligence and accuracy of scheduling judgment and avoiding resource utilization reduction caused by misjudgment; each scheduling pair group in the conflict identifier list is removed from the candidate scheduling pair group set, and the remaining part of the candidate scheduling pair group set is used as a resource conflict-free pair group set, ensuring the executability and consistency of the scheduling result in actual deployment. Removing the conflict pair groups can effectively reduce the probability of tasks being interrupted or reallocated in the subsequent execution stage, ensuring the complete implementation of training; the conflict identifier list and the resource conflict-free pair group set are merged and marked as a conflict scheduling subset and a non-conflict scheduling subset respectively to obtain a conflict filtering set, constructing a structured conflict filtering set. The system realizes data isolation between the conflict detection stage and the sorting and scoring stage, avoiding incorrect or invalid data from interfering with the subsequent scheduling logic.

[0113] Among them, resource grouping is performed on all scheduling pair group candidate records in the candidate scheduling pair group set, and all scheduling pair groups are clustered according to resource identifiers to obtain a resource pair group set, specifically including:

[0114] First, the candidate scheduling pair group set is received. The candidate scheduling pair group set is generated in the previous stage, and each scheduling pair group candidate record includes data fields such as subject identifier, matching resource identifier, proposed start time, proposed end time, and daily training duration. To facilitate the development of conflict detection operations, the system needs to perform structured classification processing on all scheduling pair group records in this set and group them with the resource identifier as the clustering dimension.

[0115] The system sequentially traverses each scheduling pair group record in the candidate scheduling pair group set and reads the resource identification field it contains. To ensure processing efficiency, the system creates a hash map table with the resource identification as the primary key during the initialization phase. Whenever a new scheduling pair group is read, it determines whether the resource identification already exists as the primary key in the map table. If the resource identification has not appeared yet, the system will create a corresponding primary key in the map table and initialize it with an empty list; if the resource identification already exists, it will directly append the current scheduling pair group record to the scheduling pair group list corresponding to the resource identification. In this way, the system can complete the resource clustering process of all scheduling pair groups in a single linear scan, thereby efficiently constructing the structural mapping relationship of "resource identification - scheduling pair group list".

[0116] After the traversal and classification operations of all scheduling pair group records are completed, the system extracts the scheduling pair group list corresponding to each resource identification in the hash map table to form a resource pair group set. Each element in this set is a resource pair group, and its content is all the time segments when the resource is scheduled in the candidate scheduling pair group set. The system can use this resource pair group set to perform subsequent concurrent conflict judgment and time intersection detection based on the resource dimension.

[0117] In a preferred embodiment of the present invention, according to the conflict filtering set, the conflicting pair groups are removed from it, and the remaining candidate scheduling pair groups are used as the effective scheduling set, and they are sorted according to the position dependency information to obtain a sorted data sequence, including:

[0118] According to the conflict filtering set, the scheduling pair groups marked as the conflicting scheduling subset are removed from the candidate scheduling pair group set, and the remaining scheduling pair groups are retained to obtain the effective scheduling set;

[0119] According to the effective scheduling set, the subject identification and position number information in each scheduling pair group are extracted to construct subject-position comparison data;

[0120] According to the position dependency information, the set of precedence relationships between different positions is determined to form a position dependency graph;

[0121] The position numbers in the subject-position comparison data are mapped to the position dependency graph to establish the transitive dependency relationship between subjects, and a subject scheduling directed graph is obtained;

[0122] The subject scheduling directed graph is sorted, and the execution order list of subjects is determined according to the direction of the dependency edges in the graph to obtain a sorted path sequence;

[0123] The subject identifications in the sorted path sequence are mapped back to the corresponding scheduling pair groups in the effective scheduling set and arranged in order to obtain a sorted data sequence.

[0124] In the embodiment of the present invention, according to the conflict filtering set, the scheduling pair groups marked as the conflict scheduling subset are removed from the candidate scheduling pair group set, and the remaining scheduling pair groups are retained to obtain an effective scheduling set, ensuring that each resource in the generated effective scheduling set has a clear time attribution and will not be occupied by multiple subjects in parallel, avoiding the problems of "resource preemption" and "subject mutex violation" that are prone to occur in traditional static scheduling; according to the effective scheduling set, the subject identifier and position number information in each scheduling pair group are extracted to construct subject-position comparison data, establishing a mapping path between the training task and the position, so that the sequential logical relationship between positions can be transmitted to the level of subject execution order, providing a logical basis for sorting; according to the position dependency information, the set of precedence relationships between different positions is determined to form a position dependency graph, which improves the structural integrity and logical consistency of the training scheduling, avoiding disordered subject execution or broken training processes caused by the failure to transmit position dependencies; mapping each position number in the subject-position comparison data to the position dependency graph to establish the transitive dependency relationship between subjects, obtaining a subject scheduling directed graph, which intuitively reflects the scheduling dependency path between subjects, can capture indirect dependency relationships, effectively prevent "circular dependencies" or "logical deadlocks" from occurring in the training process, and enhance the adaptability of the system under complex training plans; sorting the subject scheduling directed graph, determining the execution order list of subjects according to the direction of the dependency edges in the graph to obtain a sorted path sequence, so that the subject scheduling result strictly follows the position dependency order, significantly improving the clarity and execution correctness of the training scheduling logic, and avoiding training effect deviations caused by the advance or delay of subjects; mapping the subject identifiers in the sorted path sequence back to the corresponding scheduling pair groups in the effective scheduling set and arranging them in order to obtain a sorted data sequence, combining the subject execution order at the logical level with the scheduling records at the data level, and realizing a closed-loop mapping from the position dependency logic to the resource configuration execution.

[0125] Among them, mapping each position number in the subject-position comparison data to the position dependency graph to establish the transitive dependency relationship between subjects and obtaining a subject scheduling directed graph specifically includes:

[0126] First, the system extracts the subject identifiers and position numbers information contained in each scheduling pair group according to the effective scheduling set, and constructs a subject-position comparison data table. Each record in the data table represents the position number P corresponding to a certain subject M. Subsequently, the system combines the pre-generated position dependency graph, which is a directed graph structure, where each edge represents the execution order between the "preceding position" and the "succeeding position". For example, if the training task of position P1 must be completed before that of position P2, there is a directed edge from P1 to P2 in this graph. The system analyzes each pair of subjects (M1, M2) by traversing the subject-position comparison data. If the position to which M1 belongs is P1, the position to which M2 belongs is P2, and there is a directed edge from P1 to P2 in the position dependency graph, a directed edge from M1 to M2 is added at the subject level. The system repeats this operation until all possible subject pairs are analyzed, and finally forms a subject scheduling directed graph. Each node in this graph structure corresponds to a subject identifier, and each edge indicates that this subject must be executed before another subject, forming a subject scheduling dependency network extended based on the position relationship.

[0127] Among them, sorting the subject scheduling directed graph, determining the execution order list of subjects according to the direction of the dependency edges in the graph, and obtaining the sorted path sequence, specifically including:

[0128] After the subject scheduling directed graph is constructed, the system performs a topological sorting operation on it to determine the execution order of subjects. Topological sorting starts from the nodes with "in-degree 0", that is, the subjects that currently have no predecessor dependencies. The system first scans all subject nodes in the graph, identifies the set of subjects with in-degree 0, and adds them to the initialized execution sequence list. Then, these subject nodes and all their output edges are removed from the graph, and the in-degrees of the remaining nodes are recalculated. In the new graph, find the subject nodes with in-degree 0 again, and repeat the above process until all subject nodes are output to the execution sequence and the sorting is completed. If it is found during the sorting process that there are no nodes with in-degree 0, it means that there is a circular dependency relationship in the graph, and the system will trigger an exception handling mechanism to prompt the existence of a dependency conflict and terminate the scheduling. After successfully completing the topological sorting, the sorted path sequence output by the system is the global execution order of the subjects, ensuring that all dependency relationships are satisfied.

[0129] Among them, mapping the subject identifiers in the sorted path sequence back to the corresponding scheduling pair groups in the effective scheduling set, and arranging them in order to obtain the sorted data sequence, specifically including:

[0130] After the system obtains the subject sorting path sequence, it reads each subject identifier in turn and performs a matching search in the effective scheduling set based on this identifier. Each subject corresponds to one or more scheduling pair group candidate records in the effective scheduling set. The system preferentially selects the scheduling pair group that meets the following conditions according to the sorting order: the resource time period arrangement satisfies the sorting dependency relationship, that is, the time arrangement is later than the scheduling times of all predecessor subjects. For this reason, the system uses the training end time of the sorted subjects as the time boundary to filter and match the scheduling times of subsequent subjects to ensure that there is no reverse order conflict in time. When there are multiple eligible scheduling pair groups, the system preferentially selects the one with the highest score according to the score value in the resource score set. After the matching is completed, the system adds the record of this scheduling pair group to the sorting data sequence. This process continues until all subject identifiers in the sorting path sequence are mapped to their best scheduling pair groups in the effective scheduling set.

[0131] In a preferred embodiment of the present invention, according to the sorting data sequence, the candidate resource sets corresponding to each subject are extracted, and the score values of each candidate resource are calculated to obtain a resource score set, including:

[0132] According to each subject in the sorting data sequence, all resources associated with it are extracted from the effective scheduling set to form a course candidate resource set, and subject-resource mapping data is obtained;

[0133] According to the subject-resource mapping data, for each candidate resource, the number of times it appears in the candidate resource set is counted to obtain the resource appearance times, the proportion of it that has been selected by other subjects currently is calculated, the proportion of the usage duration of this resource by the current subject is extracted, and the resource usage frequency item is calculated by the interaction of the three;

[0134] For each resource, the set of job numbers that it can serve is counted, the subject coverage ratio under each job is extracted, and based on the graph depth of the job node in the job dependency graph, it is normalized to obtain the job priority weight, and the job coverage frequency item is calculated according to the subject coverage ratio and the job priority weight;

[0135] For each subject in the sorting data sequence, its position in the sorting is recorded, combined with the total number of subjects, its priority coefficient is calculated, and at the same time, its completed training duration and planned total training duration are extracted to calculate the subject completion item;

[0136] The resource usage frequency item, the job coverage frequency item and the subject completion item are weighted and summed to calculate the score values of different resources for each subject, and each score value is combined with its corresponding subject identifier and resource identifier to form a score record to obtain a resource score set.

[0137] In an embodiment of the present invention, according to each subject in the sorted data sequence, all resources associated with it are extracted from the effective scheduling set to form a course candidate resource set, and subject resource mapping data is obtained. All resource sets that may meet the current subject training needs in the scheduling are archived to provide an analysis basis for subsequent scoring; according to the subject resource mapping data, for each candidate resource, the number of times it appears in the candidate resource set is counted to obtain the number of resource appearances, the proportion of its current selection by other subjects is calculated, and the proportion of the current subject's use time of the resource is extracted. The resource usage frequency item is calculated based on the interaction of the three. By integrating the resource appearance frequency, usage overlap and adaptability to subject requirements, the system can evaluate whether the resource is scarce or exclusive, and then determine its scheduling priority, so as to solve the resource problem. Conflicts and duplications in allocation can be solved to improve the fairness and rationality of resource utilization; for each resource, the set of position numbers that it can serve is counted, the subject coverage ratio under each position is extracted, and the position priority weight is normalized based on the graph depth of the position node in the position dependency graph. The position coverage frequency item is calculated according to the subject coverage ratio and the position priority weight, which can measure the adaptability of resources to key positions in the system, thereby achieving the continuity of the position scheduling priority strategy and improving the overall resource allocation efficiency of the system; for each subject in the sorted data sequence, its position in the sort is recorded, and its priority coefficient is calculated in combination with the total number of subjects. At the same time, its completed training time and planned total training time are extracted, and the subject completion item is calculated to respond to the progress of the training tasks in a timely manner. For subjects that have not been completed and are ranked high, higher weights will be given for resource scheduling to ensure that key training tasks can be prioritized in the scheduling plan and improve the system's satisfaction and timeliness with training node goals; the resource usage frequency item, position coverage frequency item and subject completion item are weighted and summed, and the score value of each subject for different resources is calculated. Each score value and its corresponding subject identifier and resource identifier are combined into a score record to obtain a resource score set, realizing a multi-dimensional quantitative evaluation of the matching degree between resources and subjects, and providing a quantitative decision-making basis for subsequent recommended scheduling.

[0138] The calculation formula of the score value is: ,

[0139] in, For resources Relative to subject The rating value of is the index of the resource, is the index of the subject, For resources The number of times it appears in the candidate resource set, For resources The proportion of subjects assigned, For Subject Usage of resources The duration ratio is the total number of positions is the index of the position is the resource available for the position The ratio is the position The number of subjects required is the position The priority weight is the subject The completed training duration is the subject The total training plan duration is the subject The priority coefficient is the coefficient

[0140] Among them is the weight coefficient, and their sum is 1. The specific weight values can be dynamically adjusted, and can be optimized based on the management strategy of the training system, the current task requirements, and the resource supply and demand situation to achieve the optimal matching of the scheduling results and maximize the execution effect

[0141] In the case of extremely limited training resources (such as a sharp increase in the number of trainees, only 1-2 sets of hardware resources, etc.), the usage frequency and preemption ratio of resources should be given priority. As much as possible, resources should be allocated to subjects with less competition and high exclusivity. At this time, the proportion of the first term (resource frequency) in the formula should be increased, that is, increase , to avoid the scheduling results concentrating on high-conflict resources. The universality of positions and the priority of subjects still retain reference significance, but their roles are weaker. Therefore is correspondingly decreased. At this time take the values of 0.6, 0.2, and 0.2 respectively; in the case where a training center supports the synchronous training of multiple different positions (such as parallel operations of management, R & D, operation and maintenance, etc.), the system needs to give priority to allocating resources to resources with high coverage of positions and applicable to key positions (large graph depth, high priority). Therefore, the weight term should be significantly increased to ensure the rationality of scheduling between positions. At the same time, reduce the impact of the resource conflict term (that is, reduce ), and allow a certain degree of repeated resource scheduling to improve the concurrency ability. At this time take the values of 0.2, 0.6, and 0.2 respectively; when some subjects are in the final stage of the training cycle, or a key assessment task is approaching, the system needs to give priority to allocating resources to subjects with low training completion but high priority (such as retraining of management skills). At this time, the third term of the formula (subject urgency) should be mainly considered, and increase The weights are used to ensure that the scheduling results prioritize the resource allocation for unfinished key subjects. The position coverage and resource conflicts are temporarily weakened in exchange for an improvement in the overall quality of training task completion. At this time The values are 0.2, 0.2, and 0.6 respectively.

[0142] Among them, is the priority weight of the position based on the graph structure depth of each position node in the position dependency graph. In the previous step, the system has constructed a directed position dependency graph based on the resource dependencies and subject continuity between positions. Each node in this graph represents a position, indicating that a certain position depends on the resources or task execution order of another position. For each position node , the system calculates its graph depth through depth-first traversal, that is, the maximum path length of this position node from the starting node in the dependency path. The greater the graph depth, the later the position is in the execution process, and usually the training tasks of the previous positions need to be completed first. In order to normalize the position graph depth into a weight item in the scoring, the system normalizes the maximum depth of all positions to obtain the priority weight calculation formula of the position : . This priority weight reflects the relative execution stage of the position in the global training process. The larger the value, the later it is, and usually the more it depends on the previous training content, which is suitable as an indicator for resource priority scheduling.

[0143] Among them, is the priority coefficient of the subject , which is an urgency indicator used to describe the current subject in the scheduling ranking, and it comes from the position of this subject in the sorting data sequence. This sorting data sequence was obtained after constructing a directed scheduling graph of subjects through position dependency relationships in the previous claim, and gives a sorted list of all subjects in the order of the dependency path in the graph. Assuming the total number of subjects is , the position of the current subject is (the position number starts from 0, and the smaller the number, the more forward), and the system calculates its priority coefficient according to its position in the overall sorting, using the following formula: . The result of this formula is a real number between [0,1]. The closer it is to 1, the more forward the current subject is, and the stronger the scheduling priority. It and the training completion degree of the subject together constitute the training task urgency item in the scoring formula, that is, the unfinished and forward-sorted subjects will get higher scores and priority scheduling recommendations for resources.

[0144] In a preferred embodiment of the present invention, according to the resource scoring set, the resource identifier with the largest score value in each subject is selected as the recommended resource, and the time period label and position number in the effective scheduling set are extracted to obtain scheduling plan data, including:

[0145] For each subject identifier, the resource identifier with the highest score value is screened out from the resource scoring set and determined as the recommended resource for the subject;

[0146] In the effective scheduling set, find the scheduling pair group corresponding to the subject and its corresponding recommended resource, and extract the time period label in the scheduling pair group as the training start and end time;

[0147] Extract the corresponding position number from the scheduling pair group as the associated position identifier of the subject;

[0148] According to the subject identifier, recommended resource identifier, training start and end time, and position number, obtain the resource allocation detail record, and summarize all the resource allocation detail records to obtain the scheduling plan data.

[0149] In the embodiment of the present invention, for each subject identifier, the resource identifier with the highest score value is screened out from the resource scoring set and determined as the recommended resource for the subject, realizing the optimal decision-making of resource recommendation, avoiding blind selection, and improving the matching degree between the recommendation result and the actual training demand; in the effective scheduling set, find the scheduling pair group corresponding to the subject and its corresponding recommended resource, and extract the time period label in the scheduling pair group as the training start and end time, ensuring the executability of the resource allocation time in the scheduling plan and directly associating with the effectiveness verification in the previous conflict detection and time matching stage of the system; extract the corresponding position number from the scheduling pair group as the associated position identifier of the subject, clarifying the responsible entity and operation position of subject training, and making the scheduling plan have position directivity; according to the subject identifier, recommended resource identifier, training start and end time, and position number, obtain the resource allocation detail record, and summarize all the resource allocation detail records to obtain the scheduling plan data, which not only has clear readability but also supports seamless connection between the system and other training platforms, and has high consistency, accuracy, and traceability, and can meet the requirements of various usage scenarios such as multi-level resource scheduling, task management, and data evaluation.

[0150] The embodiment of the present invention also provides a training resource intelligent recommendation scheduling system, and the system includes:

[0151] A constraint module, configured to obtain a position data set and a resource data set, and determine a subject association relationship and a position mapping relationship according to them, extract subject constraint information and position dependency information, and obtain a constraint data set;

[0152] A matching module, configured to perform time matching on the training requirement duration and resource available duration of each subject according to a constraint data set, so as to obtain a set of candidate scheduling pair groups;

[0153] A detection module, configured to perform resource conflict detection on the set of candidate scheduling pair groups, and determine whether there are different subjects corresponding to the same resource identifier and time overlap, so as to obtain a conflict filtering set;

[0154] A sorting module, configured to remove the conflicting pair groups therefrom according to the conflict filtering set, use the remaining candidate scheduling pair groups as an effective scheduling set, and sort the effective scheduling set according to position dependency information to obtain a sorted data sequence;

[0155] A scoring module, configured to extract a set of candidate resources corresponding to each subject according to the sorted data sequence, and calculate the scoring values of each candidate resource to obtain a resource scoring set;

[0156] A scheduling module, configured to select the resource identifier with the largest scoring value in each subject as a recommended resource according to the resource scoring set, and extract the time period label and position number thereof in the effective scheduling set to obtain scheduling plan data.

[0157] It should be noted that this system corresponds to the above method, and all implementation manners in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0158] An embodiment of the present invention further provides a computing device, including: a processor and a memory storing a computer program. When the computer program is run by the processor, the above method is executed. All implementation manners in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0159] An embodiment of the present invention further provides a computer-readable storage medium storing instructions. When the instructions are run on a computer, the computer is enabled to execute the above method. All implementation manners in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0160] The above is the preferred embodiment of the present invention. It should be pointed out that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. An intelligent recommendation and scheduling method for training resources, characterized in that The method includes: Obtain a position dataset and a resource dataset, determine subject association relationships and position mapping relationships based on them, extract subject constraint information and position dependency information, and obtain a constraint dataset; Based on the constraint dataset, perform time matching on the training requirement durations and resource available durations of each subject to obtain a set of candidate scheduling pair groups; Perform resource conflict detection on the set of candidate scheduling pair groups to determine whether there are different subjects corresponding to the same resource identifier and time overlap, and obtain a conflict filtering set; Based on the conflict filtering set, remove the conflicting pair groups therein, use the remaining candidate scheduling pair groups as a valid scheduling set, and sort it according to the position dependency information to obtain a sorted data sequence; Based on the sorted data sequence, extract the candidate resource sets corresponding to each subject, and calculate the scoring values of each candidate resource to obtain a resource scoring set; Based on the resource scoring set, select the resource identifier with the largest scoring value in each subject as the recommended resource, and extract its time period label and position number in the valid scheduling set to obtain scheduling plan data.

2. The intelligent recommendation scheduling method for training resources according to claim 1, wherein Obtain a position dataset and a resource dataset, determine subject association relationships and position mapping relationships based on them, extract subject constraint information and position dependency information, and obtain a constraint dataset, including: Obtain the position numbers of all positions, the subject sets corresponding to each position, and the training requirement durations of each subject to obtain a position dataset; obtain the resource numbers of all resources, the resource available durations, the resource-supported subject types, and the position identifiers to which the resources belong to obtain a resource dataset; Based on the position dataset, determine the occurrence of each subject in different positions, and extract the continuous training relationships and non-parallel constraints between subjects to obtain subject association relationship data; Based on the position dataset, determine the matching structure between position identifiers and subject sets to obtain position mapping relationship data; Based on the subject association relationship data, extract the sequential relationship, mutually exclusive execution flag, and training phase overlap relationship between each subject and other subjects to obtain subject constraint information data; Based on the position mapping relationship data and the position identifiers to which the resources belong in the resource dataset, analyze the sharing and preemption situations between different positions in the resource dimension to obtain position dependency information data; Merge the subject constraint information data and the position dependency information data to obtain a rule information set, and extract the subject identifiers, position numbers, time conditions, and execution relationship labels therein to obtain a constraint dataset.

3. The intelligent recommendation and scheduling method of training resources according to claim 2, characterized in that, Based on the constraint dataset, perform time matching on the training requirement durations and resource available durations of each subject to obtain a set of candidate scheduling pair groups, including: Obtain subject identifier, total training duration, daily training duration requirement, and flag indicating whether segmented training is allowed to obtain subject duration requirement data; obtain resource identifier, continuous available time period, and set of resource-supported subject identifiers to obtain resource time segment data; Based on the subject constraint information in the constraint dataset, filter out other subjects that have a mutually exclusive relationship with the target subject during the target subject arrangement period to obtain a set of matchable subjects; Based on the resource time segment data, extract all the resource identifiers required by each subject identifier, and combine the continuous available time periods of each resource to obtain a resource schedulable window set; For each subject in the subject duration requirement data, in combination with the flag indicating whether segmented training is allowed, sequentially select a set of time periods that can continuously cover the total training duration from the resource schedulable window set to obtain time matching items; For each time matching item, determine the subject identifier, matching resource identifier, proposed start time, proposed end time, and daily training duration value, generate candidate records for the scheduling pair group, and summarize all candidate records for the scheduling pair group to obtain a candidate scheduling pair group set.

4. The intelligent recommendation scheduling method for training resources according to claim 3, wherein Perform resource conflict detection on the candidate scheduling pair group set to determine whether there are different subjects corresponding to the same resource identifier and time overlap, and obtain a conflict filtering set, including: Group the candidate records for all scheduling pair groups in the candidate scheduling pair group set by resources, cluster all scheduling pair groups according to the resource identifier, and obtain a resource pair group set; Analyze any two scheduling pair groups in the resource pair group set to determine whether there is an overlapping interval between their corresponding start time and end time. When the result is yes, mark it as a time-overlapping pair group; Based on the time-overlapping pair group, further determine whether their corresponding subject identifiers are different. When the result is yes, mark the pair group as a conflict pair group to obtain a conflict identifier list; Remove each scheduling pair group in the conflict identifier list from the candidate scheduling pair group set, and use the remaining part of the candidate scheduling pair group set as a resource conflict-free pair group set; Merge the conflict identifier list and the resource conflict-free pair group set, and mark them as a conflict scheduling subset and a non-conflict scheduling subset respectively to obtain a conflict filtering set.

5. The intelligent recommendation and scheduling method of training resources according to claim 4, characterized in that According to the conflict filtering set, remove the conflict pair groups in it, use the remaining candidate records for the scheduling pair group as an effective scheduling set, and sort it according to the position dependency information to obtain a sorted data sequence, including: According to the conflict filtering set, remove the scheduling pair groups marked as the conflict scheduling subset from the candidate scheduling pair group set, and retain the remaining scheduling pair groups to obtain an effective scheduling set; According to the effective scheduling set, extract the subject identifier and position number information in each scheduling pair group to construct subject-position comparison data; According to the position dependency information, determine the set of precedence relationship between different positions to form a position dependency graph; Map each position number in the subject-position comparison data to the position dependency graph to establish a transitive dependency relationship between subjects and obtain a subject scheduling directed graph; Sort the subject scheduling directed graph, and determine the execution order list of subjects according to the direction of the dependency edges in the graph to obtain a sorted path sequence; Map the subject identifiers in the sorted path sequence back to the corresponding scheduling pair groups in the effective scheduling set, and arrange them in order to obtain a sorted data sequence.

6. The intelligent recommendation scheduling method of training resources according to claim 5, characterized in that, According to the sorted data sequence, extract the candidate resource set corresponding to each subject, and calculate the scoring value of each candidate resource to obtain a resource scoring set, including: According to each subject in the sorted data sequence, extract all resources associated with it from the effective scheduling set to form a course candidate resource set and obtain subject-resource mapping data; According to the subject-resource mapping data, for each candidate resource, count the number of times it appears in the candidate resource set to obtain the resource appearance count, calculate the proportion of it that has been selected by other subjects currently, extract the proportion of the usage duration of this resource by the current subject, and calculate the resource usage frequency item based on the interaction of the three; For each resource, count the set of job number identifiers it can serve, extract the subject coverage ratio under each job, and normalize it based on the graph depth of the job node in the job dependency graph to obtain the job priority weight. Calculate the job coverage frequency item based on the subject coverage ratio and the job priority weight; For each subject in the sorted data sequence, record its position in the sorting, combine it with the total number of subjects to calculate its priority coefficient, and at the same time extract its completed training duration and the total planned training duration to calculate the subject completion item; Perform a weighted sum of the resource usage frequency item, the job coverage frequency item, and the subject completion item, calculate the scoring value of different resources for each subject, and form a scoring record by combining each scoring value with its corresponding subject identifier and resource identifier to obtain the resource scoring set.

7. An intelligent recommendation and scheduling method for training resources according to claim 6, characterized in that, According to the resource scoring set, select the resource identifier with the largest scoring value in each subject as the recommended resource, and extract its time period label and job number in the effective scheduling set to obtain the scheduling plan data, including: For each subject identifier, screen out the resource identifier with the highest scoring value from the resource scoring set and determine it as the recommended resource for this subject; In the effective scheduling set, find the scheduling pair group corresponding to the subject and its corresponding recommended resource, and extract the time period label in this scheduling pair group as the training start and end time; Extract the corresponding job number from this scheduling pair group as the associated job identifier of the subject; According to the subject identifier, recommended resource identifier, training start and end time, and job number, obtain the resource allocation detail record, and summarize all the resource allocation detail records to obtain the scheduling plan data.

8. An intelligent recommendation and scheduling system for training resources, characterized in that, This system is used to execute the method described in any one of claims 1 to 7. The system includes: A constraint module for obtaining the job data set and the resource data set, and determining the subject association relationship and the job mapping relationship according to them, extracting the subject constraint information and the job dependency information to obtain the constraint data set; A matching module for performing time matching on the training requirement duration and the resource available duration of each subject according to the constraint data set to obtain the candidate scheduling pair group set; A detection module for performing resource conflict detection on the candidate scheduling pair group set to determine whether there are different subjects corresponding to the same resource identifier and time overlap, and obtaining the conflict filtering set; A sorting module for removing the conflicting pair groups from the conflict filtering set according to the conflict filtering set, taking the remaining candidate scheduling pair groups as the effective scheduling set, and sorting them according to the job dependency information to obtain the sorted data sequence; A scoring module for extracting the candidate resource set corresponding to each subject according to the sorted data sequence and calculating the scoring value of each candidate resource to obtain the resource scoring set; A scheduling module for selecting the resource identifier with the largest scoring value in each subject as the recommended resource according to the resource scoring set, and extracting its time period label and job number in the effective scheduling set to obtain the scheduling plan data.

9. A computing device, characterized in that, Including: One or more processors; A storage device for storing one or more programs, which when executed by the one or more processors cause the one or more processors to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, A program is stored in the computer-readable storage medium, and when the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images