Collaborative mechanism-based talent ability cultivation and evaluation system and method

Abstract

The invention discloses a talent ability cultivation and evaluation system and method based on a collaborative mechanism, and the method comprises the following steps: constructing an ability label system, and generating an ability label set; acquiring learning behavior data of the talents in the course learning process; obtaining task execution data of the talents in the business practice process; collecting a lecturer cooperation event, a tutor cooperation event and a supervisor cooperation event, and aggregating the three types of cooperation events according to a time window; executing cross-track synchronous adjustment on learning transfer parameters, guess parameters and error parameters in the initial parameter matrix; inputting the learning track sequence, the practice track sequence and the dynamic parameter matrix into a double-track Bayesian knowledge tracking model; and calculating a track consistency deviation based on the dual-track capability state vector set. According to the method, a double-track Bayesian knowledge tracking model is adopted, and talent ability cultivation and evaluation are achieved.

Description

Technical Field

[0001] This invention relates to the field of talent cultivation and intelligent assessment technology, and in particular to a talent capability cultivation and assessment system and method based on a collaborative mechanism. Background Technology

[0002] Existing talent cultivation and competency assessment technologies mostly rely on course learning records, completion of practical tasks, or manual evaluation results to construct talent competency profiles. They typically use a single-track mastery update model to track the learning process. Such methods update competency status based only on a single sequence of learning or practical behaviors, making it difficult to simultaneously capture the differences between the learning trajectory and the practical trajectory.

[0003] Furthermore, traditional knowledge tracking models often use fixed parameters to update mastery levels, failing to adapt dynamically to changes in the organization's training structure, instructor coaching intensity, or supervisor involvement. This leads to a disconnect between competency assessment results and the actual training process. Current technologies typically process learning behavior data and practical behavior data independently, without integrating the two types of data into a unified competency labeling system. They also lack structured methods for collecting and quantifying the behaviors of collaborating roles, making it difficult to comprehensively represent the multi-agent roles in the training process.

[0004] Furthermore, traditional methods often fail to establish an interactive update mechanism between the learning track and the practice track, relying solely on the mastery of independently outputted data as the assessment criterion. This fails to identify the impact of track consistency deviations on competency levels, limiting the accuracy of talent development planning and strategy generation.

[0005] Therefore, how to provide a talent competency training and assessment system and method based on a collaborative mechanism is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose a talent competency cultivation and assessment system and method based on a collaborative mechanism. This invention employs a dual-track Bayesian knowledge tracing model, a dynamic parameter adjustment mechanism, and a unified competency tagging encoding method to structurally model learning trajectory data, practical trajectory data, and collaborative events involving lecturers, mentors, and supervisors. It generates a dual-track competency state vector through cross-track interactive updates and constructs competency assessment results and personalized cultivation strategies based on track consistency deviations. This invention can simultaneously characterize competency changes on both the learning and practical tracks, possessing advantages such as fine-grained modeling, sufficient expression of the collaborative mechanism, high accuracy in competency assessment, and strong executableness in generating cultivation strategies.

[0007] A talent competency cultivation and assessment method based on a collaborative mechanism according to an embodiment of the present invention includes the following steps:

[0008] Construct a capability tag system, generate a capability tag set, and establish an initial parameter matrix based on the capability tag set;

[0009] Acquire learning behavior data of talents during the course learning process, perform structured processing and serialization encoding on the learning behavior data, and generate learning trajectory sequences;

[0010] Acquire task execution data of talents in the process of business practice, perform structured processing and quantitative encoding on task execution data, and generate practice trajectory sequences;

[0011] Collect lecturer collaboration events, tutor collaboration events, and supervisor collaboration events, aggregate the three types of collaboration events according to time windows, and generate a collaboration factor sequence;

[0012] Based on the synergistic factor sequence, cross-track synchronous adjustment is performed on the learning transfer parameters, guess parameters and error parameters in the initial parameter matrix to construct a dynamic parameter matrix;

[0013] The learning trajectory sequence, the practice trajectory sequence, and the dynamic parameter matrix are input into the dual-track Bayesian knowledge tracing model to form a dual-track capability state vector set.

[0014] The track consistency deviation is calculated based on the dual-track capability state vector set, a capability assessment result set is generated, and collaborative adjustment suggestions, training track strategies, and practice track strategies are generated.

[0015] Optionally, the construction of the capability tag system, the generation of a capability tag set, and the establishment of an initial parameter matrix based on the capability tag set specifically include:

[0016] Obtain organizational capability requirements, assign numbers to the capability tags corresponding to each capability requirement to form a capability tag set, and assign a unique tag index to each capability tag in the set;

[0017] A unified encoding space is constructed based on the set of capability tags. A capability tag encoding vector is generated for each capability tag. All capability tag encoding vectors are arranged in order according to the tag index. The arranged encoding vectors are stacked column by column to form a capability tag encoding matrix.

[0018] For each ability tag in the ability tag set, learning transfer parameters, guessing parameters, and error parameters are established respectively. The transfer parameters, guessing parameters, and error parameters corresponding to the same ability tag are combined to form an ability parameter vector. All ability parameter vectors are arranged in order according to the tag index. The arranged ability parameter vectors are stacked row by row to form an ability parameter matrix.

[0019] The capability tag encoding matrix and the capability parameter matrix are combined to generate an initial parameter matrix containing capability tag encoding information and parameter information.

[0020] Optionally, the step of acquiring learning behavior data of talents during the course learning process, performing structured processing and serialization encoding on the learning behavior data, and generating a learning trajectory sequence specifically includes:

[0021] Obtain course learning behavior records of target talents within a preset observation period, add timestamp information to each course learning behavior record, and arrange them in chronological order to form a sequence of learning behavior events;

[0022] For each learning behavior event in the learning behavior event sequence, the fields are parsed to extract the course identifier, knowledge unit identifier, grade information and completion status information. The knowledge unit identifier is mapped to the target ability tag in the ability tag set to obtain a sample pair set containing time index and ability tag index.

[0023] Based on the capability tag encoding matrix, the capability tag in the sample pair set is referenced to the corresponding capability tag encoding vector, and the time index is combined with the corresponding capability tag encoding vector to form the learning behavior encoding vector.

[0024] The learning behavior encoding vectors are sorted according to the time index to construct a learning behavior encoding sequence arranged in chronological order;

[0025] The learning behavior encoding sequence is divided into learning trajectory subsequences that correspond one-to-one with the set of ability labels according to the ability label index, thus obtaining the learning trajectory sequence set.

[0026] The set of learning trajectory sequences is indexed and aligned with the learning transition parameters, guessing parameters and error parameters in the initial parameter matrix. Each learning trajectory subsequence is associated with the capability parameter vector of the corresponding capability label in the initial parameter matrix to form a learning trajectory sequence corresponding to the learning trajectory parameters in the initial parameter matrix.

[0027] Optionally, the step of acquiring task execution data of talents in the process of business practice, performing structured processing and quantitative encoding on the task execution data, and generating a practice trajectory sequence specifically includes:

[0028] Obtain the task execution records of the target talent within a preset business cycle, attach a task timestamp to each task execution record, and arrange them in chronological order to form a sequence of task execution events;

[0029] For each task execution event in the task execution event sequence, the fields are parsed to extract the task identifier, the business unit identifier to which the task belongs, the task evaluation result and the task completion status information. The business unit identifier is mapped to the target capability tag in the capability tag set to obtain a set of task sample pairs containing time index and capability tag index.

[0030] Based on the capability tag encoding matrix, the capability tags in the task sample set are referenced by the corresponding capability tag encoding vectors, and the time index is combined with the capability tag encoding vector to form the task execution encoding vector.

[0031] Sort the task execution encoding vectors according to the time index to construct a task execution encoding sequence arranged in chronological order;

[0032] The task execution encoding sequence is divided into practice trajectory subsequences that correspond one-to-one with the set of capability labels according to the capability label index, thus obtaining a set of practice trajectory sequences.

[0033] The set of practice trajectory sequences is indexed and aligned with the learning transfer parameters, guessing parameters and error parameters in the initial parameter matrix. Each practice trajectory subsequence is associated with the capability parameter vector corresponding to the capability label in the initial parameter matrix, forming a practice trajectory sequence corresponding to the practice trajectory parameters in the initial parameter matrix.

[0034] Optionally, the collection of lecturer collaboration events, tutor collaboration events, and supervisor collaboration events, and the aggregation of these three types of collaboration events according to time windows to generate a collaboration factor sequence specifically includes:

[0035] Obtain teaching collaboration event records of lecturers during the preset observation period, obtain tutoring collaboration event records of mentors during the observation period, and obtain supervision collaboration event records of supervisors during the observation period, and arrange them in chronological order to form a collaboration event sequence.

[0036] For each collaborative event record in the collaborative event sequence, the field is parsed to extract the event type field, the event timestamp field, and the event intensity field. Based on the event timestamp field, the three types of collaborative events are divided into windows within a preset time window to obtain a set of collaborative events within the window.

[0037] Obtain the event intensity field value corresponding to the collaborative event set in each window, and arrange the event intensity field values ​​of the three types of collaborative events in the same window in a preset order to form a collaborative event intensity vector;

[0038] Arrange the collaborative event intensity vectors of all windows in window index order to construct a collaborative event intensity sequence;

[0039] Each collaborative event intensity vector in the collaborative event intensity sequence is quantized to generate a collaborative factor value. All collaborative factor values ​​are then arranged in window order to form a collaborative factor sequence, which is used to characterize the collaborative intensity within different time windows.

[0040] Optionally, the step of performing cross-track synchronous adjustment of the learning transfer parameters, guessing parameters, and error parameters in the initial parameter matrix based on the cooperating factor sequence to construct the dynamic parameter matrix specifically includes:

[0041] Obtain the sequence of synergistic factors, and for each synergistic factor in the sequence, generate a learning trajectory parameter adjustment coefficient vector and a practice trajectory parameter adjustment coefficient vector based on the synergistic factor value;

[0042] For each capability tag in the capability tag set and the synergy factor index, read the capability parameter vector corresponding to the capability tag from the initial parameter matrix, and adjust the capability parameter vector according to the learning trajectory parameter adjustment coefficient vector. Generate the learning trajectory adjusted capability parameter vector in the order of adjusted learning transfer parameters, adjusted guess parameters and adjusted error parameters.

[0043] Among them, the adjusted learning transfer parameter is obtained by multiplying the corresponding adjustment coefficient by the original learning transfer parameter, the adjusted guess parameter is obtained by multiplying the corresponding adjustment coefficient by the original guess parameter, and the adjusted error parameter is obtained by multiplying the corresponding adjustment coefficient by the original error parameter.

[0044] For each capability tag and synergy factor index in the capability tag set, the capability parameter vector corresponding to the capability tag in the initial parameter matrix is ​​adjusted according to the practice trajectory parameter adjustment coefficient vector. The practice trajectory adjusted capability parameter vector is generated in the order of adjusted learning transfer parameters, adjusted guess parameters, and adjusted error parameters.

[0045] Among them, the adjusted learning transfer parameter is obtained by multiplying the learning transfer parameter by the learning transfer adjustment coefficient in the practical trajectory parameter adjustment coefficient vector; the adjusted guessing parameter is obtained by multiplying the guessing parameter by the guessing adjustment coefficient in the practical trajectory parameter adjustment coefficient vector; and the adjusted error parameter is obtained by multiplying the error parameter by the error adjustment coefficient in the practical trajectory parameter adjustment coefficient vector.

[0046] For the synergy factor index, the adjusted capability parameter vectors of the learning trajectories corresponding to all capability labels are stacked row by row according to the capability label index order to form a dynamic parameter matrix of the learning trajectory. The adjusted capability parameter vectors of the practice trajectories corresponding to all capability labels are also stacked row by row according to the capability label index order to form a dynamic parameter matrix of the practice trajectory.

[0047] The dynamic parameter matrix of the learning trajectory obtained under each coordinating factor index is combined with the dynamic parameter matrix of the practice trajectory to construct a dynamic parameter matrix for the learning trajectory and the practice trajectory.

[0048] Optionally, the step of inputting the learning trajectory sequence, practice trajectory sequence, and dynamic parameter matrix into the dual-track Bayesian knowledge tracing model to form a dual-track capability state vector set specifically includes:

[0049] For each capability tag in the capability tag set, initialize the learning track hidden state vector and the practice track hidden state vector respectively, as the initial hidden state of the dual-track Bayesian knowledge tracing model;

[0050] The trajectory elements of the learning trajectory subsequence under the time index, together with the dynamic capability parameter vector of the learning track corresponding to the capability label in the dynamic parameter matrix set, are used as the input of the learning track update unit.

[0051] The trajectory elements of the practice trajectory subsequence under the time index, together with the dynamic capability parameter vector of the practice track corresponding to the capability label in the dynamic parameter matrix set, are used as the input of the practice track update unit.

[0052] Under the time index, the hidden state vector of the learning track is updated based on the input of the learning track update unit to obtain the probability of mastering the learning track ability;

[0053] Under the time index, the hidden state vector of the practice track is updated based on the input of the practice track update unit to obtain the probability of mastering the practice track capability;

[0054] Based on the cross-track coupling weight vector in the dynamic parameter matrix set, the learning track capability mastery probability is used as the modulation input of the practice track capability mastery unit, and the practice track capability mastery probability is used as the modulation input of the learning track update unit. Synchronous cross-track interactive update is performed on the hidden state vectors of the two tracks to obtain the learning track capability mastery probability and the practice track capability mastery probability after interactive processing.

[0055] The cross-track interactive update includes obtaining the probability of mastering the learning track ability and the probability of mastering the practice track ability under the time index;

[0056] The probability of mastering the learning track ability and the cross-track coupling weight vector are combined in a preset order to form a cross-track modulation vector for the practice track, and the update input vector of the practice track is modulated based on the cross-track modulation vector.

[0057] The probability of mastering the practical track capability and the cross-track coupling weight vector are combined in a preset order to form a cross-track modulation vector for the learning track, and the learning track update input vector is modulated based on the cross-track modulation vector.

[0058] The hidden state vector of the learning track is synchronously updated based on the updated input vector of the learning track after modulation, and the hidden state vector of the practice track is synchronously updated based on the updated input vector of the practice track after modulation, so as to obtain the probability of mastering the learning track ability after interactive processing and the probability of mastering the practice track ability after interactive processing.

[0059] Obtain the final ability mastery probability of the learning track and the final ability mastery probability of the practice track after interactive processing, and form a dual-track ability state vector with ability labels in a preset order.

[0060] Arrange the dual-track capability state vectors corresponding to all capability tags in the order of capability tag index to construct a set of dual-track capability state vectors to represent the mastery status of dual-track capabilities.

[0061] Optionally, the step of calculating track consistency deviation based on the dual-track capability state vector set, generating a capability assessment result set, and generating collaborative adjustment suggestions, training track strategies, and practice track strategies specifically includes:

[0062] For each capability tag in the capability tag set, obtain the dual-track capability state vector corresponding to the capability tag in the dual-track capability state vector set;

[0063] For each capability tag, the track difference parameter is calculated based on the final capability mastery probability of the learning track and the final capability mastery probability of the practice track in the dual-track capability state vector, and the track consistency deviation parameter is formed based on the track difference parameter.

[0064] The track difference parameter is the difference between the probability of mastering the final ability of the learning track and the probability of mastering the final ability of the practice track, and the track consistency deviation parameter is the absolute value of the track difference.

[0065] For each ability tag, the probability of mastering the final ability on the learning track, the probability of mastering the final ability on the practice track, and the track consistency deviation parameter are combined in order to form an ability assessment result vector. The ability assessment result vectors corresponding to all ability tags are arranged in the order of ability tag index to form an ability assessment result set.

[0066] Based on the set of competency assessment results and the preset instructor collaboration weight vector, a set of instructor collaboration adjustment suggestion vectors corresponding one-to-one with the competency tag set is generated, and the set of instructor collaboration adjustment suggestion vectors is used as collaboration adjustment suggestions for instructors.

[0067] Based on the set of competency assessment results and the preset mentor collaboration weight vector, a set of mentor collaboration adjustment suggestion vectors corresponding one-to-one with the competency tag set is generated, and the set of mentor collaboration adjustment suggestion vectors is used as the collaboration adjustment suggestions for mentors.

[0068] Based on the set of competency assessment results and the preset supervisor collaboration weight vector, a set of supervisor collaboration adjustment suggestion vectors corresponding one-to-one with the competency label set is generated, and the set of supervisor collaboration adjustment suggestion vectors is used as the collaboration adjustment suggestions for supervisors.

[0069] Based on the set of capability assessment results and the training track strategy generation rules, a set of training track strategy vectors corresponding one-to-one with the set of capability labels is generated to form the training track strategy.

[0070] Based on the set of capability assessment results and the rules for generating practice track strategies, a set of practice track strategy vectors corresponding one-to-one with the set of capability labels is generated to form the practice track strategy.

[0071] The aforementioned suggestions for collaborative adjustment of lecturers, mentors, and supervisors, along with the aforementioned training track strategy and practice track strategy, will be distributed to the subsequent talent development execution process to coordinate and control the trajectory of the talent development process.

[0072] A talent competency development and assessment system based on a collaborative mechanism according to an embodiment of the present invention includes the following modules:

[0073] The capability tag construction module is used to build a capability tag system, generate a capability tag set, and establish an initial parameter matrix based on the capability tag set.

[0074] The learning trajectory processing module is used to acquire learning behavior data of talents during the course learning process, perform structured processing and serialization encoding, and generate learning trajectory sequences.

[0075] The practice trajectory processing module is used to acquire task execution data of talents in the process of business practice, perform structured processing and quantitative coding, and generate practice trajectory sequences;

[0076] The collaborative factor generation module is used to collect lecturer collaborative events, tutor collaborative events, and supervisor collaborative events, aggregate them according to time windows, and generate a collaborative factor sequence.

[0077] The dynamic parameter adjustment module is used to perform cross-track synchronous adjustment of the initial parameter matrix based on the cooperating factor sequence to construct the dynamic parameter matrix;

[0078] The dual-track knowledge tracing module is used to input the learning trajectory sequence, the practice trajectory sequence, and the dynamic parameter matrix into the dual-track Bayesian knowledge tracing model to form a dual-track capability state vector set.

[0079] The capability assessment module is used to generate collaborative adjustment suggestions, training track strategies, and practice track strategies based on the dual-track capability state vector set.

[0080] The beneficial effects of this invention are:

[0081] This invention constructs a capability tagging system and establishes an initial parameter matrix based on a unified encoding space, enabling learning trajectories and practice trajectories to be expressed under the same capability dimension. This achieves alignment and co-modeling of capability states even with heterogeneous data sources. This approach can simultaneously link course learning behaviors and business practice behaviors in actual training scenarios, providing a unified capability expression basis for subsequent dynamic parameter adjustments and dual-track capability updates, reducing capability calculation deviations caused by differences in data sources.

[0082] This invention employs a mechanism for generating a dynamic parameter matrix driven by a synergistic factor sequence, enabling learning transfer parameters, guessing parameters, and error parameters to change synchronously based on the synergistic intensity of lecturers, mentors, and supervisors. This mechanism allows the knowledge tracing model to reflect the impact of multi-role collaborative behavior during organizational training, achieving dynamic adjustment of competency mastery status over time and overcoming the shortcomings of existing fixed-parameter models that cannot reflect synergistic differences. By introducing a cross-track interactive update unit into the dual-track Bayesian knowledge tracing model, this invention can synchronously update the competency status of the learning track and the practice track, reflecting the relationship between tracks in the model calculation process and improving the completeness of competency status representation.

[0083] This invention further generates track consistency deviation based on dual-track capability state vectors and constructs a capability assessment result set to quantitatively express the differences between the learning trajectory and the practice trajectory. Through structured processing of the capability assessment results, this invention can generate collaborative adjustment suggestions for lecturers, tutors, and supervisors, while simultaneously generating training track strategies and practice track strategies, providing an actionable operational basis for subsequent talent development. This method forms a continuous link in the combination of capability assessment and strategy generation, enabling the talent development process to be adjusted in real time according to capability changes, improving the adaptability and accuracy of the training process. Attached Figure Description

[0084] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0085] Figure 1 This is a flowchart of a talent competency cultivation and evaluation method based on a collaborative mechanism proposed in this invention;

[0086] Figure 2 This is a structural diagram of the dual-track Bayesian knowledge tracing model in the talent competency cultivation and evaluation system and method based on a collaborative mechanism proposed in this invention;

[0087] Figure 3 This is a schematic diagram of the structure of a talent competency cultivation and evaluation system based on a collaborative mechanism proposed in this invention. Detailed Implementation

[0088] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0089] refer to Figure 1 and Figure 2 A talent competency development and assessment method based on a collaborative mechanism includes the following steps:

[0090] Construct a capability tag system, generate a capability tag set, and establish an initial parameter matrix based on the capability tag set;

[0091] Acquire learning behavior data of talents during the course learning process, perform structured processing and serialization encoding on the learning behavior data, and generate learning trajectory sequences;

[0092] Acquire task execution data of talents in the process of business practice, perform structured processing and quantitative encoding on task execution data, and generate practice trajectory sequences;

[0093] Collect lecturer collaboration events, tutor collaboration events, and supervisor collaboration events, aggregate the three types of collaboration events according to time windows, and generate a collaboration factor sequence;

[0094] Based on the synergistic factor sequence, cross-track synchronous adjustment is performed on the learning transfer parameters, guess parameters and error parameters in the initial parameter matrix to construct a dynamic parameter matrix;

[0095] The learning trajectory sequence, the practice trajectory sequence, and the dynamic parameter matrix are input into the dual-track Bayesian knowledge tracing model to form a dual-track capability state vector set.

[0096] The track consistency deviation is calculated based on the dual-track capability state vector set, a capability assessment result set is generated, and collaborative adjustment suggestions, training track strategies, and practice track strategies are generated.

[0097] In this embodiment, the construction of the capability tag system, the generation of a capability tag set, and the establishment of an initial parameter matrix based on the capability tag set specifically include:

[0098] Obtain organizational capability requirements, assign numbers to the capability tags corresponding to each capability requirement to form a capability tag set, and assign a unique tag index to each capability tag in the set;

[0099] A unified encoding space is constructed based on the set of capability tags. A capability tag encoding vector is generated for each capability tag. All capability tag encoding vectors are arranged in order according to the tag index. The arranged encoding vectors are stacked column by column to form a capability tag encoding matrix.

[0100] For each ability tag in the ability tag set, learning transfer parameters, guessing parameters, and error parameters are established respectively. The transfer parameters, guessing parameters, and error parameters corresponding to the same ability tag are combined to form an ability parameter vector. All ability parameter vectors are arranged in order according to the tag index. The arranged ability parameter vectors are stacked row by row to form an ability parameter matrix.

[0101] The capability tag encoding matrix and the capability parameter matrix are combined to generate an initial parameter matrix containing capability tag encoding information and parameter information.

[0102] In this embodiment, the step of acquiring learning behavior data of talents during the course learning process, performing structured processing and serialization encoding on the learning behavior data, and generating a learning trajectory sequence specifically includes:

[0103] Obtain course learning behavior records of target talents within a preset observation period, add timestamp information to each course learning behavior record, and arrange them in chronological order to form a sequence of learning behavior events;

[0104] Data anonymization processing is performed on the learning behavior records in the learning behavior event sequence. Fields involving talent identity identifiers, account identifiers, and identifiable personal information are anonymized or mapped to generate an anonymized learning behavior event sequence.

[0105] For each learning behavior event in the desensitized learning behavior event sequence, the fields are parsed to extract the course identifier, knowledge unit identifier, grade information and completion status information. The knowledge unit identifier is mapped to the target ability tag in the ability tag set to obtain a sample pair set containing time index and ability tag index.

[0106] Based on the capability tag encoding matrix, the capability tag in the sample pair set is referenced to the corresponding capability tag encoding vector, and the time index is combined with the corresponding capability tag encoding vector to form the learning behavior encoding vector.

[0107] The learning behavior encoding vectors are sorted according to the time index to construct a learning behavior encoding sequence arranged in chronological order;

[0108] The learning behavior encoding sequence is divided into learning trajectory subsequences that correspond one-to-one with the set of ability labels according to the ability label index, thus obtaining the learning trajectory sequence set.

[0109] The set of learning trajectory sequences is indexed and aligned with the learning transition parameters, guessing parameters and error parameters in the initial parameter matrix. Each learning trajectory subsequence is associated with the capability parameter vector of the corresponding capability label in the initial parameter matrix to form a learning trajectory sequence corresponding to the learning trajectory parameters in the initial parameter matrix.

[0110] In this embodiment, the step of acquiring task execution data of talents in the process of business practice, performing structured processing and quantitative encoding on the task execution data, and generating a practice trajectory sequence specifically includes:

[0111] Obtain the task execution records of the target talent within a preset business cycle, attach a task timestamp to each task execution record, and arrange them in chronological order to form a sequence of task execution events;

[0112] For each task execution event in the task execution event sequence, the fields are parsed to extract the task identifier, the business unit identifier to which the task belongs, the task evaluation result and the task completion status information. The business unit identifier is mapped to the target capability tag in the capability tag set to obtain a set of task sample pairs containing time index and capability tag index.

[0113] Based on the capability tag encoding matrix, the capability tags in the task sample set are referenced by the corresponding capability tag encoding vectors, and the time index is combined with the capability tag encoding vector to form the task execution encoding vector.

[0114] Sort the task execution encoding vectors according to the time index to construct a task execution encoding sequence arranged in chronological order;

[0115] The task execution encoding sequence is divided into practice trajectory subsequences that correspond one-to-one with the set of capability labels according to the capability label index, thus obtaining a set of practice trajectory sequences.

[0116] The set of practice trajectory sequences is indexed and aligned with the learning transfer parameters, guessing parameters and error parameters in the initial parameter matrix. Each practice trajectory subsequence is associated with the capability parameter vector corresponding to the capability label in the initial parameter matrix, forming a practice trajectory sequence corresponding to the practice trajectory parameters in the initial parameter matrix.

[0117] In this embodiment, the collection of lecturer collaboration events, tutor collaboration events, and supervisor collaboration events, and the aggregation of these three types of collaboration events according to time windows to generate a collaboration factor sequence specifically includes:

[0118] Obtain teaching collaboration event records of lecturers during the preset observation period, obtain tutoring collaboration event records of mentors during the observation period, and obtain supervision collaboration event records of supervisors during the observation period, and arrange them in chronological order to form a collaboration event sequence.

[0119] For each collaborative event record in the collaborative event sequence, the field is parsed to extract the event type field, the event timestamp field, and the event intensity field. Based on the event timestamp field, the three types of collaborative events are divided into windows within a preset time window to obtain a set of collaborative events within the window.

[0120] Obtain the event intensity field value corresponding to the collaborative event set in each window, and arrange the event intensity field values ​​of the three types of collaborative events in the same window in a preset order to form a collaborative event intensity vector;

[0121] Arrange the collaborative event intensity vectors of all windows in window index order to construct a collaborative event intensity sequence;

[0122] Each collaborative event intensity vector in the collaborative event intensity sequence is quantized to generate a collaborative factor value, and all collaborative factor values ​​are arranged in window order to form a collaborative factor sequence, which is used to characterize the collaborative intensity within different time windows.

[0123] The quantization process includes: linear normalization of the collaborative event intensity vector based on its maximum and minimum values; piecewise quantization mapping of the collaborative event intensity vector to piecewise discrete values ​​according to a preset intensity level; standardization transformation of the collaborative event intensity vector based on the mean and standard deviation; interval compression mapping of the collaborative event intensity vector to a target numerical interval according to a preset interval boundary; weighted transformation of the collaborative event intensity vector based on a preset weight vector; and any quantization process capable of converting the collaborative event intensity into a unified numerical form that can be input into a model.

[0124] In this embodiment, the step of performing cross-track synchronous adjustment of the learning transfer parameters, guessing parameters, and error parameters in the initial parameter matrix based on the cooperating factor sequence to construct the dynamic parameter matrix specifically includes:

[0125] Obtain the sequence of synergistic factors, and for each synergistic factor in the sequence, generate a learning trajectory parameter adjustment coefficient vector and a practice trajectory parameter adjustment coefficient vector based on the synergistic factor value;

[0126] The process of generating the learning trajectory parameter adjustment coefficient vector includes: mapping the collaborative factor value to three coefficients used to adjust the learning transfer parameter, guess parameter and error parameter, and assembling them into a learning trajectory parameter adjustment coefficient vector in a preset order;

[0127] The process of generating the practice trajectory parameter adjustment coefficient vector includes: mapping the collaborative factor value to three coefficients used to adjust the learning transfer parameter, guessing parameter and error parameter, and assembling the practice trajectory parameter adjustment coefficient vector in a preset order;

[0128] The set of parameter adjustment coefficients is formed by combining all learning trajectory parameter adjustment coefficient vectors generated by the co-factor index with the practice trajectory parameter adjustment coefficient vector.

[0129] For each capability tag in the capability tag set and the synergy factor index, read the capability parameter vector corresponding to the capability tag from the initial parameter matrix, and adjust the capability parameter vector according to the learning trajectory parameter adjustment coefficient vector. Generate the learning trajectory adjusted capability parameter vector in the order of adjusted learning transfer parameters, adjusted guess parameters and adjusted error parameters.

[0130] Among them, the adjusted learning transfer parameter is obtained by multiplying the corresponding adjustment coefficient by the original learning transfer parameter, the adjusted guess parameter is obtained by multiplying the corresponding adjustment coefficient by the original guess parameter, and the adjusted error parameter is obtained by multiplying the corresponding adjustment coefficient by the original error parameter.

[0131] For each capability tag and synergy factor index in the capability tag set, the capability parameter vector corresponding to the capability tag in the initial parameter matrix is ​​adjusted according to the practice trajectory parameter adjustment coefficient vector. The practice trajectory adjusted capability parameter vector is generated in the order of adjusted learning transfer parameters, adjusted guess parameters, and adjusted error parameters.

[0132] Among them, the adjusted learning transfer parameter is obtained by multiplying the learning transfer parameter by the learning transfer adjustment coefficient in the practical trajectory parameter adjustment coefficient vector; the adjusted guessing parameter is obtained by multiplying the guessing parameter by the guessing adjustment coefficient in the practical trajectory parameter adjustment coefficient vector; and the adjusted error parameter is obtained by multiplying the error parameter by the error adjustment coefficient in the practical trajectory parameter adjustment coefficient vector.

[0133] For the synergy factor index, the adjusted capability parameter vectors of the learning trajectories corresponding to all capability labels are stacked row by row according to the capability label index order to form a dynamic parameter matrix of the learning trajectory. The adjusted capability parameter vectors of the practice trajectories corresponding to all capability labels are also stacked row by row according to the capability label index order to form a dynamic parameter matrix of the practice trajectory.

[0134] The dynamic parameter matrix of the learning trajectory obtained under each coordinating factor index is combined with the dynamic parameter matrix of the practice trajectory to construct a dynamic parameter matrix for the learning trajectory and the practice trajectory.

[0135] In this embodiment, the step of inputting the learning trajectory sequence, the practice trajectory sequence, and the dynamic parameter matrix into the dual-track Bayesian knowledge tracing model to form a dual-track capability state vector set specifically includes:

[0136] For each capability tag in the capability tag set, initialize the learning track hidden state vector and the practice track hidden state vector respectively, as the initial hidden state of the dual-track Bayesian knowledge tracing model;

[0137] The trajectory elements of the learning trajectory subsequence under the time index, together with the dynamic capability parameter vector of the learning track corresponding to the capability label in the dynamic parameter matrix set, are used as the input of the learning track update unit.

[0138] The trajectory elements of the practice trajectory subsequence under the time index, together with the dynamic capability parameter vector of the practice track corresponding to the capability label in the dynamic parameter matrix set, are used as the input of the practice track update unit.

[0139] Under the time index, the hidden state vector of the learning track is updated based on the input of the learning track update unit to obtain the probability of mastering the learning track ability;

[0140] Specifically, this includes: under the time index, assembling the learning trajectory elements and the learning track dynamic capability parameter vector in a preset order to form a learning track update input vector; calculating the learning track state transition factor based on the learning track update input vector, and applying the learning track state transition factor to the learning track latent state vector of the previous time index to generate an updated learning track latent state vector; calculating the learning track capability mastery probability based on the updated learning track latent state vector and the learning track dynamic capability parameter vector to obtain the learning track capability mastery probability of the time index;

[0141] Under the time index, the hidden state vector of the practice track is updated based on the input of the practice track update unit to obtain the probability of mastering the practice track capability;

[0142] Specifically, this includes: under the time index, assembling the practice trajectory elements and the practice track dynamic capability parameter vector into a practice track update input vector according to a preset order; calculating the practice track state transition factor based on the practice track update input vector, and applying the practice track state transition factor to the practice track hidden state vector of the previous time index to generate an updated practice track hidden state vector; calculating the practice track capability mastery probability based on the updated practice track hidden state vector and the practice track dynamic capability parameter vector to obtain the practice track capability mastery probability of the time index;

[0143] Based on the cross-track coupling weight vector in the dynamic parameter matrix set, the learning track capability mastery probability is used as the modulation input of the practice track capability mastery unit, and the practice track capability mastery probability is used as the modulation input of the learning track update unit. Synchronous cross-track interactive update is performed on the hidden state vectors of the two tracks to obtain the learning track capability mastery probability and the practice track capability mastery probability after interactive processing.

[0144] The cross-track interactive update includes obtaining the probability of mastering the learning track ability and the probability of mastering the practice track ability under the time index;

[0145] The probability of mastering the learning track ability and the cross-track coupling weight vector are combined in a preset order to form a cross-track modulation vector for the practice track, and the update input vector of the practice track is modulated based on the cross-track modulation vector.

[0146] The probability of mastering the practical track capability and the cross-track coupling weight vector are combined in a preset order to form a cross-track modulation vector for the learning track, and the learning track update input vector is modulated based on the cross-track modulation vector.

[0147] The hidden state vector of the learning track is synchronously updated based on the updated input vector of the learning track after modulation, and the hidden state vector of the practice track is synchronously updated based on the updated input vector of the practice track after modulation, so as to obtain the probability of mastering the learning track ability after interactive processing and the probability of mastering the practice track ability after interactive processing.

[0148] Obtain the final ability mastery probability of the learning track and the final ability mastery probability of the practice track after interactive processing, and form a dual-track ability state vector with ability labels in a preset order.

[0149] Arrange the dual-track capability state vectors corresponding to all capability tags in the order of capability tag index to construct a set of dual-track capability state vectors to represent the mastery status of dual-track capabilities.

[0150] In this embodiment, the step of calculating track consistency deviation based on the dual-track capability state vector set, generating a capability assessment result set, and generating collaborative adjustment suggestions, training track strategies, and practice track strategies specifically includes:

[0151] For each capability tag in the capability tag set, obtain the dual-track capability state vector corresponding to the capability tag in the dual-track capability state vector set;

[0152] For each capability tag, the track difference parameter is calculated based on the final capability mastery probability of the learning track and the final capability mastery probability of the practice track in the dual-track capability state vector, and the track consistency deviation parameter is formed based on the track difference parameter.

[0153] The track difference parameter is the difference between the probability of mastering the final ability of the learning track and the probability of mastering the final ability of the practice track, and the track consistency deviation parameter is the absolute value of the track difference.

[0154] For each ability tag, the probability of mastering the final ability on the learning track, the probability of mastering the final ability on the practice track, and the track consistency deviation parameter are combined in order to form an ability assessment result vector. The ability assessment result vectors corresponding to all ability tags are arranged in the order of ability tag index to form an ability assessment result set.

[0155] Based on the set of competency assessment results and the preset instructor collaboration weight vector, a set of instructor collaboration adjustment suggestion vectors corresponding one-to-one with the competency tag set is generated, and this set of instructor collaboration adjustment suggestion vectors serves as the collaboration adjustment suggestions for instructors; based on the set of competency assessment results and the preset mentor collaboration weight vector, a set of mentor collaboration adjustment suggestion vectors corresponding one-to-one with the competency tag set is generated, and this set of mentor collaboration adjustment suggestion vectors serves as the collaboration adjustment suggestions for mentors; based on the set of competency assessment results and the preset supervisor collaboration weight vector, a set of supervisor collaboration adjustment suggestion vectors corresponding one-to-one with the competency tag set is generated, and this set of supervisor collaboration adjustment suggestion vectors serves as the collaboration adjustment suggestions for supervisors.

[0156] Based on the set of capability assessment results and the training track strategy generation rules, a set of training track strategy vectors corresponding one-to-one with the set of capability labels is generated to form the training track strategy.

[0157] The training track strategy generation rules include: obtaining the final ability mastery probability and track consistency deviation parameter of the learning track from the ability assessment result vector; combining the final ability mastery probability of the learning track with the preset ability threshold vector in a preset order to generate the learning track ability deviation vector; determining the learning intensity level corresponding to the ability label based on the learning track ability deviation vector; generating the training track content index based on the learning intensity level and the preset learning content index table; constructing the training track strategy vector based on the training track content index and the preset training content sequence; and arranging the training track strategy vectors corresponding to all ability labels in the order of the ability label index to form a training track strategy vector set.

[0158] Based on the set of capability assessment results and the rules for generating practice track strategies, a set of practice track strategy vectors corresponding one-to-one with the set of capability labels is generated to form the practice track strategy.

[0159] The practice track strategy generation rules include: obtaining the final capability mastery probability and track consistency deviation parameter of the practice track from the capability assessment result vector; combining the final capability mastery probability of the practice track with a preset practice threshold vector in a preset order to generate a practice track capability deviation vector; determining the practice task intensity level corresponding to the capability label based on the practice track capability deviation vector; generating a practice track task index based on the practice task intensity level and a preset practice task index table; constructing a practice track strategy vector based on the practice track task index and a preset practice task sequence; and arranging the practice track strategy vectors corresponding to all capability labels in the capability label index order to form a practice track strategy vector set.

[0160] The aforementioned suggestions for collaborative adjustment of lecturers, mentors, and supervisors, along with the aforementioned training track strategy and practice track strategy, will be distributed to the subsequent talent development execution process to coordinate and control the trajectory of the talent development process.

[0161] refer to Figure 3 A talent development and evaluation system based on a collaborative mechanism includes the following modules:

[0162] The capability tag construction module is used to build a capability tag system, perform unified encoding on capability tags to generate a capability tag encoding matrix, and establish a capability parameter matrix containing learning transfer parameters, guessing parameters and error parameters based on the capability tag encoding matrix to form an initial parameter matrix;

[0163] The learning trajectory processing module is used to acquire learning behavior data of talents during the course learning process, perform structured processing and serialization encoding on the learning behavior data, generate learning trajectory sequences corresponding to ability tags, and match the learning trajectory sequences with the learning trajectory parameters in the initial parameter matrix.

[0164] The practice trajectory processing module is used to acquire task execution data of talents in the process of business practice, perform structured processing and quantitative encoding on the task execution data, generate practice trajectory sequences corresponding to ability tags, and match the practice trajectory sequences with the practice trajectory parameters in the initial parameter matrix;

[0165] The collaborative factor generation module is used to collect collaborative events of lecturers, tutors, and supervisors, aggregate collaborative events by time windows, and generate a collaborative factor sequence to characterize the intensity of collaboration based on the aggregation results.

[0166] The dynamic parameter adjustment module is used to perform cross-track synchronous adjustment of the learning transfer parameters, guessing parameters and error parameters in the initial parameter matrix based on the cooperating factor sequence, forming a set of dynamic parameter matrices for the learning trajectory and the practice trajectory;

[0167] The dual-track knowledge tracing module is used to input the learning trajectory sequence, the practice trajectory sequence and the dynamic parameter matrix into the dual-track Bayesian knowledge tracing model, and perform hidden state updates and cross-track interaction updates on the learning track and the practice track in chronological order to generate a set of dual-track capability state vectors.

[0168] The competency assessment module is used to calculate the track consistency deviation based on the dual-track competency state vector, generate a competency assessment result set, and generate collaborative adjustment suggestions for lecturers, mentors, and supervisors based on the competency assessment result set.

[0169] The strategy generation module is used to generate training track strategies and practice track strategies corresponding to the ability tags based on the set of ability assessment results, and to distribute the training track strategies and practice track strategies to the talent training execution process.

[0170] Example 1:

[0171] To verify the feasibility of this invention in practice, it was applied to an internal job skills enhancement project within an organization. The organization needs to conduct continuous talent development for multiple job positions, encompassing both systematic course learning and long-term business practice, while also introducing three collaborative roles—instructors, mentors, and supervisors—to intervene in the development process. Traditional methods in such scenarios typically rely solely on course completion records or periodic assessment results, which are insufficient to reflect changes in practical abilities or quantify the impact of multi-role collaborative behavior, leading to delayed ability assessment results and insufficient basis for strategy adjustments. This invention, by constructing a unified ability tagging system and structurally modeling learning trajectories, practice trajectories, and collaborative events, effectively solves the problem of the disconnect between learning ability and practical ability.

[0172] Throughout the entire implementation process, the system continuously collects data on course learning behavior, including learning duration, assessment results, knowledge point coverage, and error distribution, generating a learning trajectory sequence. Simultaneously, it structurally records business practice processes, including task completion status, key step accuracy, task time consumption, and result stability, generating a practice trajectory sequence. Instructor lecture frequency, tutor Q&A sessions, and supervisor review activities are uniformly collected as collaborative events and aggregated by time windows to generate a collaborative factor sequence. Based on this collaborative factor sequence, the system dynamically adjusts learning transfer parameters, guessing parameters, and error parameters, ensuring that the capability update process reflects changes in collaborative intensity. Subsequently, a dual-track Bayesian knowledge tracing model updates the learning and practice tracks in parallel and performs cross-track interactions, forming a dual-track capability state vector set. Based on this, it generates capability assessment results and targeted training and practice track strategies.

[0173] In the lightweight implementation scheme, this invention simplifies the overall process for application environments with limited resources, short training cycles, or incomplete collaborative role configurations. The lightweight scheme retains the competency tagging system, learning trajectory processing module, and dual-track knowledge tracking module; however, in the collaborative factor generation stage, only lecturer collaborative events are collected as the primary source of collaboration, while tutor and supervisor events are processed using default weights. Simultaneously, during dynamic parameter adjustment, the update frequency of the collaborative factor sequence is reduced, with parameter synchronization adjustments triggered only at stage evaluation nodes, thereby reducing computational overhead. Regarding practice trajectory collection, the lightweight scheme selects key task nodes for sampling, rather than recording the entire practice process, to reduce data collection and storage costs.

[0174] In the comparative verification, both the complete and lightweight solutions used business review results as the capability benchmark. The results showed that, in terms of the two key indicators of process execution capability and problem analysis capability, the average evaluation deviation of the lightweight solution was 0.36 and 0.38, respectively. Although slightly higher than the complete solution's 0.34 and 0.36, it was still significantly lower than the traditional single-track evaluation method's 0.41 and 0.46. Meanwhile, the lightweight solution maintained an improvement of over 10% in practical task accuracy, reduced computational resource consumption by approximately 32%, and reduced data collection by approximately 45%, significantly reducing system deployment costs while ensuring the effectiveness of the evaluation.

[0175] Therefore, this invention is not only applicable to complete collaborative training scenarios, but can also be flexibly adapted to talent training environments of different scales and resource conditions through a lightweight implementation scheme, achieving an effective balance between the accuracy of capability assessment and system operating costs.

[0176] Table 1. Comparison of Capability Assessment and Implementation Results between the Complete Solution and the Lightweight Solution

[0177] Implementation scheme type Ability Tags Mean of evaluation deviation Improvement in practical accuracy Changes in data collection volume Changes in computing resource usage Traditional methods Process execution capability 0.41 3.2% benchmark benchmark Complete solution Process execution capability 0.34 11.5% +100% +100% Lightweight solution Process execution capability 0.36 10.2% −45% −32% Traditional methods Problem analysis skills 0.46 4.1% benchmark benchmark Complete solution Problem analysis skills 0.36 13.4% +100% +100% Lightweight solution Problem analysis skills 0.38 11.1% −42% −29%

[0178] As can be seen from the table above, this invention significantly outperforms the traditional single-track assessment method in both the complete and lightweight implementation schemes in terms of capability assessment accuracy and practical capability improvement. Firstly, regarding the key capability indicator of process execution capability, the traditional method's average assessment deviation is 0.41, while the complete scheme of this invention reduces this deviation to 0.34, and the lightweight scheme also controls the deviation at 0.36, both significantly better than the assessment results of the traditional method. This indicates that this invention, through a unified capability labeling system and a dual-track capability status modeling approach, can effectively reduce the deviation between assessment results and actual capabilities, making capability assessment closer to real business performance.

[0179] Secondly, regarding the assessment of problem analysis capabilities, the average evaluation bias of traditional methods is 0.46. The complete solution of this invention reduces this value to 0.36, a reduction of 21.7%. Although the lightweight solution simplifies the collection of collaborative factors and the frequency of parameter updates, it still controls the evaluation bias within 0.38, a reduction of more than 17% compared to traditional methods. These results demonstrate that even with reduced sources of collaborative events and lower data collection frequency, the dual-track Bayesian knowledge tracing model combined with a dynamic parameter adjustment mechanism can still stably depict the trend of ability changes in learning and practice trajectories, maintaining high evaluation effectiveness.

[0180] In terms of improving practical skills, this invention also demonstrates significant advantages. The accuracy rate of process execution improved by 11.5% with the complete solution, while the lightweight solution maintained a 10.2% improvement, whereas the traditional method only improved by 3.2%. Regarding problem analysis capabilities, the accuracy rates of the complete and lightweight solutions improved by 13.4% and 11.1% respectively, significantly higher than the 4.1% improvement of the traditional method. These data indicate that the training track strategy and practice track strategy generated by this invention based on the capability assessment results can effectively guide subsequent training activities, making capability improvement quantifiable and practically effective.

[0181] Furthermore, from the perspective of system implementation costs, the lightweight implementation scheme achieves a significant reduction in resource consumption while ensuring the evaluation effect and capability improvement. Compared with the complete scheme, the lightweight scheme reduces data collection by more than 40% and computational resource consumption by about 30%, while the evaluation bias and practical effect only show minor changes. This demonstrates that the present invention has good scalability and adaptability under different resource constraints, supporting both high-precision capability training scenarios and rapid deployment needs in resource-constrained environments, reflecting the flexibility and practicality of the method in real-world applications.

[0182] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A talent capacity training and evaluation method based on a synergy mechanism, characterized in that, The method comprises the following steps: constructing a capability label system, generating a capability label set, and establishing an initial parameter matrix according to the capability label set; obtaining learning behavior data of the talent in a course learning process, performing structured processing and serialized coding on the learning behavior data, and generating a learning track sequence; obtaining task execution data of the talent in a business practice process, performing structured processing and quantized coding on the task execution data, and generating a practice track sequence; collecting lecturer collaboration events, tutor collaboration events and supervisor collaboration events, aggregating the three types of collaboration events according to a time window, and generating a collaboration factor sequence; based on the collaboration factor sequence, performing cross-track synchronous adjustment on learning transfer parameters, guessing parameters and error parameters in the initial parameter matrix to construct a dynamic parameter matrix; inputting the learning track sequence, the practice track sequence and the dynamic parameter matrix into a double-track Bayesian knowledge tracing model to form a double-track capability state vector set; calculating track consistency deviation based on the double-track capability state vector set, generating a capability evaluation result set, and generating a collaborative adjustment suggestion, a training track strategy and a practice track strategy.

2. The talent capacity training and evaluation method based on a synergy mechanism according to claim 1, characterized in that, The method of constructing a capability label system, generating a capability label set, and establishing an initial parameter matrix based on the capability label set comprises: obtaining organizational capability requirements, numbering the capability labels corresponding to each capability requirement to form a capability label set, and assigning a unique label index to each capability label in the set; based on the capability label set, constructing a unified coding space, generating a capability label coding vector for each capability label, and forming a capability label coding matrix; for each capability label in the capability label set, respectively establishing learning transfer parameters, guessing parameters and error parameters to form a capability parameter vector; stacking the arranged capability parameter vectors by rows to form a capability parameter matrix, combining the capability label coding matrix and the capability parameter matrix to generate an initial parameter matrix.

3. The method of claim 1, wherein the method is based on a synergy mechanism. The method of obtaining learning behavior data of the talent in a course learning process, performing structured processing and serialized coding on the learning behavior data, and generating a learning track sequence comprises: obtaining course learning behavior records of the target talent within a preset observation period, adding timestamp information to each course learning behavior record, and arranging them in chronological order to form a learning behavior event sequence; performing field analysis on each learning behavior event in the learning behavior event sequence to obtain a sample pair set containing time index and capability label index; based on the capability label coding matrix, referring to the corresponding capability label coding vector for the capability label in the sample pair set, combining the time index and the corresponding capability label coding vector to form a learning behavior coding vector; sorting the learning behavior coding vectors according to the time index to construct a learning behavior coding sequence; dividing the learning behavior coding sequence into learning track subsequences corresponding to the capability label set according to the capability label index to obtain a learning track sequence set; associating each learning track subsequence with the capability parameter vector of the corresponding capability label in the initial parameter matrix to form a learning track sequence.

4. The talent capacity training and evaluation method based on a synergy mechanism according to claim 1, characterized in that, The task execution data of the talent in the business practice process is acquired, and structured processing and quantitative coding are performed on the task execution data to generate a practice track sequence, specifically including: Acquiring the task execution records of the target talent within a preset business cycle, appending task timestamp information to each task execution record, and arranging the task execution records in chronological order to form a task execution event sequence; Performing field analysis on each task execution event in the task execution event sequence to obtain a task sample pair set containing time index and ability label index; Based on the ability label coding matrix, the ability label in the task sample pair set is referenced to obtain the corresponding ability label coding vector, and the time index and the ability label coding vector are combined to form a task execution coding vector; The task execution coding vectors are sorted according to the time index to construct a task execution coding sequence arranged in chronological order; The task execution coding sequence is divided into practice track subsequences corresponding to the ability label set according to the ability label index to obtain a practice track sequence set; Each practice track subsequence is associated with the ability parameter vector of the corresponding ability label in the initial parameter matrix to form a practice track sequence.

5. The talent capacity training and evaluation method based on a synergy mechanism according to claim 1, characterized in that, The lecturer collaboration event, the tutor collaboration event, and the supervisor collaboration event are collected, and the three types of collaboration events are aggregated according to a time window to generate a collaboration factor sequence, specifically including: Acquiring the lecture collaboration event records of the lecturer within an observation period, acquiring the tutoring collaboration event records of the tutor within the observation period, and acquiring the supervision collaboration event records of the supervisor within the observation period, and arranging the collaboration event records in chronological order to form a collaboration event sequence; Performing field analysis on each collaboration event record in the collaboration event sequence, and dividing the three types of collaboration events according to the event timestamp field to obtain a collaboration event set within a window; Respectively obtaining the event strength field values corresponding to each collaboration event set within a window, and arranging the event strength field values of the three types of collaboration events within the same window in a preset order to form a collaboration event strength vector; Arranging the collaboration event strength vectors of all windows in chronological order to construct a collaboration event strength sequence; Quantitatively processing each collaboration event strength vector in the collaboration event strength sequence to generate a collaboration factor value, and arranging all the collaboration factor values in chronological order to form a collaboration factor sequence.

6. The talent capacity training and evaluation method based on a synergy mechanism according to claim 1, characterized in that, Based on the collaboration factor sequence, the learning transfer parameter, the guessing parameter, and the error parameter in the initial parameter matrix are adjusted across the tracks synchronously to construct a dynamic parameter matrix, specifically including: Acquiring the collaboration factor sequence, and for each collaboration factor in the sequence, generating a learning track parameter adjustment coefficient vector and a practice track parameter adjustment coefficient vector according to the collaboration factor value; For each ability label in the ability label set and the collaboration factor index, reading the ability parameter vector from the initial parameter matrix and adjusting it to generate an adjusted ability parameter vector for the learning track; Adjusting the ability parameter vector corresponding to the ability label in the initial parameter matrix according to the practice track parameter adjustment coefficient vector to generate an adjusted ability parameter vector for the practice track; Stack the learning trajectory adjusted capability parameter vectors corresponding to all capability labels in the order of capability label indexes to form a learning trajectory dynamic parameter matrix; Stack the practice trajectory adjusted capability parameter vectors corresponding to all capability labels in the order of capability label indexes to form a practice trajectory dynamic parameter matrix; Combine the learning trajectory dynamic parameter matrix and the practice trajectory dynamic parameter matrix obtained under each coordination factor index to construct a dynamic parameter matrix.

7. The method of claim 1, wherein the method is based on a synergy mechanism. The inputting the learning trajectory sequence, the practice trajectory sequence and the dynamic parameter matrix into the double-track Bayesian knowledge tracing model to form a double-track capability state vector set specifically includes: For each capability label in the capability label set, initialize the learning track hidden state vector and the practice track hidden state vector respectively as the initial hidden state of the double-track Bayesian knowledge tracing model; At the time index, update the learning track hidden state vector based on the learning track update unit input to obtain the learning track capability mastery probability; At the time index, update the practice track hidden state vector based on the practice track update unit input to obtain the practice track capability mastery probability; Based on the cross-track coupling weight vector in the dynamic parameter matrix set, perform synchronous cross-track interaction update on the hidden state vectors of the two tracks to obtain the learning track capability mastery probability and the practice track capability mastery probability after interaction processing; Combine the learning track capability mastery probability and the cross-track coupling weight vector in a preset order to form a cross-track modulation vector, and modulate the practice track update input vector based on the cross-track modulation vector; Combine the practice track capability mastery probability and the cross-track coupling weight vector in a preset order to form a cross-track modulation vector, and modulate the learning track update input vector based on the cross-track modulation vector; Synchronously update the learning track hidden state vector based on the modulated learning track update input vector, and synchronously update the practice track hidden state vector based on the modulated practice track update input vector; Obtain the learning track final capability mastery probability and the practice track final capability mastery probability after interaction processing, and form the double-track capability state vector of the capability label in a preset order to construct the double-track capability state vector set.

8. The talent capacity training and evaluation method based on a synergy mechanism according to claim 1, characterized in that, The calculating the track consistency deviation based on the double-track capability state vector set to generate the capability evaluation result set, and generating the coordination adjustment suggestion, the training track strategy and the practice track strategy specifically includes: For each capability label in the capability label set, obtain the double-track capability state vector corresponding to the capability label in the double-track capability state vector set; For each capability label, calculate the track difference parameter according to the learning track final capability mastery probability and the practice track final capability mastery probability in the double-track capability state vector, and form the track consistency deviation parameter based on the track difference parameter; For each capability label, combine the learning track final capability mastery probability, the practice track final capability mastery probability and the track consistency deviation parameter in order to form a capability evaluation result vector; Arrange the capability evaluation result vectors corresponding to all capability labels in the order of capability label indexes to form a capability evaluation result set; Generate lecturer collaborative adjustment suggestions, tutor collaborative adjustment suggestions and supervisor collaborative adjustment suggestions according to the capability evaluation result set and a preset weight vector; Generate a training track strategy vector set to form a training track strategy according to the capability evaluation result set and a training track strategy generation rule; Generate a practice track strategy vector set to form a practice track strategy according to the capability evaluation result set and a practice track strategy generation rule. 9.The talent capacity training and evaluation system based on a collaborative mechanism according to claim 1, which executes the talent capacity training and evaluation method based on a collaborative mechanism according to any one of claims 1 to 8, characterized in that, The method comprises the following modules: A capability label construction module is configured to construct a capability label system, generate a capability label set, and establish an initial parameter matrix according to the capability label set; A learning track processing module is configured to acquire learning behavior data of talents in a course learning process, perform structured processing and serialized encoding, and generate a learning track sequence; A practice track processing module is configured to acquire task execution data of talents in a business practice process, perform structured processing and quantized encoding, and generate a practice track sequence; A collaborative factor generation module is configured to collect lecturer collaborative events, tutor collaborative events and supervisor collaborative events, aggregate the events according to a time window, and generate a collaborative factor sequence; A dynamic parameter adjustment module is configured to perform cross-track synchronous adjustment on the initial parameter matrix based on the collaborative factor sequence, and construct a dynamic parameter matrix; A dual-track knowledge tracking module is configured to input the learning track sequence, the practice track sequence and the dynamic parameter matrix into a dual-track Bayesian knowledge tracking model, and form a dual-track capability state vector set; A capability evaluation module is configured to generate collaborative adjustment suggestions, a training track strategy and a practice track strategy based on the dual-track capability state vector set.

Images