Deep Learning-Based Endocrine Nursing Teaching Decision System

Abstract

This invention relates to the field of artificial intelligence technology and provides a deep learning-based teaching decision-making system for endocrine nursing. The system includes a data acquisition module, a data preprocessing module, a feature extraction module, a teaching evaluation module, and a teaching decision-making module. The teaching evaluation module introduces a fairness-constrained dual-BiGRU model, which improves the consistency and stability of teaching evaluation across different student groups through subgroup partitioning and fairness constraint mechanisms. The teaching decision-making module introduces a risk-aware D3QN model, constructs a risk-sensitive Q-value function, and combines a static spectral risk measurement mechanism to perform segmented weighting of teaching action rewards, optimizing the decision-making and risk control capabilities of the strategy network. This system can achieve deep fusion of multi-source heterogeneous teaching data and personalized strategy recommendations, making it particularly suitable for complex medical teaching scenarios such as endocrine nursing, and exhibiting good fairness, adaptability, and practicality.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a deep learning-based endocrine nursing teaching decision system. Background Technology

[0002] With the rapid development of artificial intelligence technology, deep learning models are increasingly widely used in fields such as healthcare and smart education. Currently, intelligent teaching systems have been initially applied in medical and nursing education. Common methods include knowledge graph-based knowledge push systems, student behavior prediction models based on convolutional neural networks (CNN) or long short-term memory networks (LSTM), and teaching strategy generation methods based on reinforcement learning. These systems can, to some extent, model the correlation between teaching content and student behavior data, assisting teachers in teaching evaluation and strategy recommendation. However, existing intelligent nursing teaching systems generally have the following limitations: First, most systems are based on the average characteristics of the overall student population, lacking in-depth modeling of the differences in characteristics among subgroups, making it difficult to identify specific behavioral patterns (such as "fast-paced skipping learning"). The personalized teaching needs of students (either "video-dependent" or "video-dependent") can easily lead to distorted assessment results for some groups, thus affecting the suitability and fairness of teaching. Secondly, the current application of reinforcement learning in optimizing teaching strategies generally uses the traditional Q-value function as the core objective, ignoring the potential risk differences brought by teaching actions to different learning paths. This makes the generated teaching strategies lack stability and interpretability in complex teaching situations, making it difficult to meet the comprehensive requirements of nursing education for strategy safety and expected benefits. Finally, some systems have insufficient capabilities in the fusion and modeling of multi-source heterogeneous teaching data (such as case texts, answer records, video behavior, etc.), failing to fully extract the deep-seated correlation features between teaching content and student behavior, affecting the accuracy and responsiveness of subsequent assessment and decision-making modules, resulting in a lack of stability and relevance in teaching paths. Summary of the Invention

[0003] This invention aims to overcome the shortcomings of existing intelligent nursing teaching systems, such as insufficient personalized assessment, significant bias in group fairness, and limited risk control capabilities in teaching decisions. It proposes a deep learning-based endocrine nursing teaching decision-making system that integrates multidimensional data modeling, fairness constraints, and risk perception optimization strategies to improve the fairness of teaching assessment and the scientific rigor of teaching decisions. The system introduces a fairness-constrained dual-BiGRU model to cluster student groups with different learning behavior characteristics and applies subgroup fairness constraints during training, effectively suppressing systematic biases of the assessment model towards specific groups. Simultaneously, to enhance the risk control capabilities of teaching strategies, a static spectral risk measurement mechanism is introduced to segmentally weight and regulate the reward distribution of teaching actions. This allows for the construction of a risk-sensitive Q-value function during reinforcement learning, thereby optimizing the stability and adaptability of the strategy network. Based on multidimensional heterogeneous teaching data, this system can output personalized, risk-controllable, and fair optimized teaching decision paths, making it particularly suitable for complex teaching scenarios such as endocrine nursing.

[0004] This invention provides a deep learning-based endocrine nursing teaching decision system, which includes a data acquisition module, a data preprocessing module, a feature extraction module, a teaching evaluation module, and a teaching decision module.

[0005] The data acquisition module collects multi-dimensional endocrine nursing teaching data and constructs the original teaching dataset;

[0006] The data preprocessing module cleans, encodes, and standardizes the original teaching dataset to generate a cleaned teaching dataset.

[0007] The feature extraction module extracts teaching content features and student behavior features from the cleaned teaching dataset to obtain a comprehensive feature set. The teaching content features include text semantic features, case difficulty labels, and knowledge point associations. The student behavior features include time-series learning trajectories, question mastery assessments, and interaction preference distributions.

[0008] The teaching evaluation module introduces the Fair Constrained Dual-BiGRU model. This model is constructed by building a BiGRU model and incorporating a K-Means subgroup partitioning mechanism and a subgroup fairness constraint mechanism during BiGRU model training to optimize the evaluation performance of the BiGRU model. The Fair Constrained Dual-BiGRU model processes the comprehensive feature set to generate teaching evaluation results, including a knowledge point matrix, a weakness diagnosis report, and a teaching adaptability curve. The Fair Constrained Dual-BiGRU model is a BiGRU model.

[0009] The instructional decision-making module defines the instructional action space and introduces a risk-aware D3QN model. This model is constructed by optimizing the traditional Q-value function of the D3QN model during its training process using a static spectral risk measurement mechanism. The risk-aware D3QN model is then used to assess the risk sensitivity of instructional actions within the instructional action space and optimize strategies, thereby obtaining instructional decision-making results. The risk-aware D3QN model comprises the D3QN model itself.

[0010] Furthermore, the process of generating teaching evaluation results by processing the comprehensive feature set through a fair constraint dual-BiGRU model specifically includes the following steps:

[0011] Step S1: Input the comprehensive feature set into the BiGRU model for initial training and generate initial teaching evaluation results;

[0012] Step S2: Set the error tolerance threshold and reference evaluation benchmark value. Perform multi-dimensional feature division on the comprehensive feature set using K-Means to obtain typical subgroups. Input the feature data of each individual in the typical subgroup into the BiGRU model to generate subgroup evaluation results. In the model training process, introduce a subgroup fairness constraint mechanism to constrain the deviation between the mean of the subgroup evaluation results and the reference evaluation benchmark value, and generate subgroup error constraints.

[0013] Step S3: Use the subgroup error constraint as a training constraint term to formalize the training task of the BiGRU model into a multi-fairness constraint-nonconvex optimization problem; introduce the primal-dual optimization algorithm to transform the multi-fairness constraint-nonconvex optimization problem into a dual objective function for joint training; use the dual objective function during training to iteratively optimize the principal parameters and dual variables of the BiGRU model.

[0014] Furthermore, the process of obtaining teaching decision-making results using the risk-perception D3QN model specifically includes the following steps:

[0015] Step B1: Construct a teaching status representation based on the teaching assessment results. The teaching action space includes basic content review, targeted training delivery, case video guidance, and assessment task arrangement; execute the teaching actions in the teaching action space, monitor changes in student status, and construct state transition pairs.

[0016] Step B2: Design a reward function, combine state transition pairs to generate immediate reward values; input the state transition pairs and immediate reward values ​​into the D3QN model for training, use the traditional Q-value function to evaluate the long-term benefits of the teaching actions, obtain the state-action reward distribution, and obtain the initial policy network;

[0017] Step B3: During training, a static spectral risk measurement mechanism is introduced to replace the training method in the D3QN model that uses the traditional Q-value function as the optimization objective. Specifically, the state-action reward distribution is modeled as a complete reward distribution. Based on the complete reward distribution, a risk spectral function is introduced to construct a piecewise linear approximation function as a risk-weighted mapping. The piecewise linear approximation function is embedded into the traditional Q-value function to construct a risk-sensitive Q-value function. This enables the perception and control of different risk quantiles in the teaching action reward distribution, generates a risk-sensitive Q-value, optimizes the policy network parameters in the D3QN model, and obtains the current teaching state and the optimized policy network.

[0018] Step B4: Based on the current teaching status and the optimization strategy network, select the optimal teaching action from the teaching action space and generate the teaching decision result.

[0019] By adopting the above solution, the beneficial effects achieved by the present invention are as follows:

[0020] This invention introduces a fairness-constrained dual-BiGRU model to achieve subgroup identification and structured modeling of various student learning behaviors in endocrine nursing education. It can dynamically calibrate the teaching assessment path for typical learning groups such as high-frequency interactive learners, video-dependent learners, and low-score review learners. During the training process, the model introduces a subgroup fairness constraint mechanism and dynamically compresses the assessment bias of different groups through primal-dual optimization, thereby significantly improving the consistency and fairness of teaching assessment results among diverse student groups. This solves the assessment distortion problem caused by traditional models ignoring group differences and enhances the system's ability to accurately reflect personalized learning status in complex nursing teaching environments.

[0021] Meanwhile, this invention introduces a static spectral risk measurement mechanism into the teaching strategy optimization module, constructing a risk-sensitive Q-value function for piecewise weighted control of the distribution of teaching action rewards. Unlike traditional reinforcement learning models that only aim at expected returns, this invention can perceive the risk quantile differences brought about by different actions in the teaching path and dynamically adjust the selection priority of teaching behaviors. Through the synergistic effect of the risk spectral function and the piecewise linear approximation function, the system can take into account both teaching benefits and learning risks when generating teaching strategies, effectively improving the stability and controllability of the teaching path output, and providing reliable guarantees for scenarios such as pushing high-difficulty content and recommending key training tasks in nursing education.

[0022] In summary, this invention integrates core technologies such as multidimensional data-driven modeling, fairness constraint optimization, and risk-aware strategy control to construct an intelligent teaching decision-making system with high adaptability and controllability. This system not only enhances the comprehensive perception of students' learning status but also strengthens its intervention value in complex teaching decisions through a refined strategy output mechanism. Particularly in professional fields such as endocrine nursing, which have complex knowledge systems and high requirements for teaching feedback, this invention can significantly improve the accuracy and adaptability of teaching, significantly address the problems of uneven distribution of traditional teaching resources and lagging strategy control, and promote the development of nursing education towards a more intelligent, equitable, and safe direction. Attached Figure Description

[0023] Figure 1 This is a comparison chart of the training effects of the teaching decision-making models proposed in Examples 4 and 5;

[0024] Figure 2 This is the teaching adaptability curve proposed in Example 6;

[0025] Figure 1 In the diagram, the horizontal axis represents the number of training rounds (1-20), and the vertical axis represents the average reward value. The results show that the risk-aware D3QN model (blue curve) converges faster in the early stages and achieves a higher average reward value in the later stages, outperforming the traditional D3QN model (orange curve) overall.

[0026] Figure 2 In the middle, the left side shows the teaching adaptability score, and the bottom shows the week. Detailed Implementation

[0027] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0028] Example 1: This invention provides a deep learning-based endocrine nursing teaching decision-making system, which includes a data acquisition module, a data preprocessing module, a feature extraction module, a teaching evaluation module, and a teaching decision-making module.

[0029] The data acquisition module collects multi-dimensional endocrine nursing teaching data and constructs the original teaching dataset;

[0030] Multidimensional teaching data for endocrine nursing includes:

[0031] Teaching content data: courseware, case studies, test questions, video explanations;

[0032] Nursing knowledge graph data: classification of endocrine diseases, nursing standards;

[0033] Student learning behavior data: learning duration, answer records, and interaction frequency;

[0034] Learning assessment data: daily performance, exam scores, and subjective evaluations;

[0035] The data preprocessing module cleans, encodes, and standardizes the original teaching dataset to generate a cleaned teaching dataset.

[0036] The feature extraction module extracts teaching content features and student behavior features from the cleaned teaching dataset to obtain a comprehensive feature set. The teaching content features include text semantic features, case difficulty labels, and knowledge point associations. The student behavior features include time-series learning trajectories, question mastery assessments, and interaction preference distributions.

[0037] The teaching evaluation module introduces the Fair Constrained Dual-BiGRU model. This model is constructed by building a BiGRU model and incorporating a K-Means subgroup partitioning mechanism and a subgroup fairness constraint mechanism during BiGRU model training to optimize its evaluation performance. The Fair Constrained Dual-BiGRU model processes the comprehensive feature set to generate teaching evaluation results, including a knowledge point matrix, a weakness diagnosis report, and a teaching adaptability curve. The Fair Constrained Dual-BiGRU model is a BiGRU model.

[0038] The teaching decision-making module defines the teaching action space and introduces a risk-aware D3QN model. This model is constructed by optimizing the traditional Q-value function of the D3QN model during its training process using a static spectral risk measurement mechanism. The risk-aware D3QN model is then used to assess the risk sensitivity of teaching actions within the teaching action space and optimize strategies, thereby obtaining teaching decision-making results. The risk-aware D3QN model includes the D3QN model itself.

[0039] Example 2, based on Example 1, describes the process of generating teaching evaluation results by processing the comprehensive feature set using the Fair Constraint Dual-BiGRU model. The specific steps include:

[0040] Step S1: Input the comprehensive feature set into the BiGRU model for initial training and generate initial teaching evaluation results;

[0041] Step S2: Set the error tolerance threshold and reference evaluation benchmark. Use K-Means to perform multi-dimensional feature partitioning on the comprehensive feature set to obtain typical subgroups. Input the feature data of each individual in the typical subgroup into the BiGRU model to generate subgroup evaluation results. In the model training process, introduce a subgroup fairness constraint mechanism to constrain the deviation between the mean of the subgroup evaluation results and the reference evaluation benchmark, generate subgroup error constraints, and ensure that the deviation does not exceed the error tolerance threshold to suppress the systematic evaluation bias generated by the model among different types of students.

[0042] The formula for the subgroup fairness constraint mechanism is as follows:

[0043] ;

[0044] in, Indicates the typical subgroup number, This indicates the sample number within a typical subgroup. Indicates the first The number of samples in a typical subgroup Indicates the first Among the typical subgroups, the first Input data consisting of comprehensive features of each sample; These represent the main parameters of the Bi-GRU model. Indicates the Bi-GRU model for the th Sample The predicted results of teaching assessment, i.e., the subgroup assessment results; Indicates the first The average of the prediction evaluation results of all samples in a typical subgroup; Indicates the first Reference assessment benchmark values ​​for a typical subgroup; Represents the distance function; Indicates the first Error tolerance thresholds for a typical subgroup; This indicates that the calibration constraint is applied to all typical subgroups;

[0045] Typical subgroups include: high-frequency interactive students, low-score review-oriented students, video-dependent students, fast-paced learning-skipping students, and exam-oriented students;

[0046] Step S3: Using the subgroup error constraint as a training constraint, the training task of the BiGRU model is formalized into a multi-fairness constraint-nonconvex optimization problem. A primal-dual optimization algorithm is introduced to transform the multi-fairness constraint-nonconvex optimization problem into a jointly trained dual objective function. During training, the dual objective function is used to iteratively optimize the principal parameters and dual variables of the BiGRU model. The principal parameters are used to minimize the overall teaching error, while the dual variables dynamically adjust the fairness constraint strength for each subgroup. Through continuous iterative optimization, the model can automatically identify and suppress the path with the largest evaluation bias for each subgroup, effectively improving the fairness, stability, and generalization ability of the teaching evaluation model, optimizing the initial teaching evaluation results, and generating the final teaching evaluation results.

[0047] The optimization objective of a nonconvex optimization problem with multiple fairness constraints is:

[0048] ;

[0049] in, This represents the mean squared error function, which serves as the loss function for instructional evaluation. This represents the total number of training samples. Indicates the first The actual teaching assessment value corresponding to each sample This indicates that by adjusting the main parameters Minimize the overall teaching assessment error; Indicates the first A measure of the average difference between model predictions and reference evaluation benchmarks in a typical subgroup;

[0050] Transforming the multi-fairness-constrained non-convex optimization problem into a dual objective function helps to reconstruct the multi-constrained non-convex training task, which is difficult to solve directly, into a scalable and adjustable joint optimization problem. By introducing dual variables, the model can automatically adjust the constraint strength of each subgroup, thereby effectively suppressing subgroup prediction bias while optimizing the overall teaching performance, and improving the fairness, robustness and generalization ability of the model.

[0051] Dual objective function formula:

[0052] ;

[0053] in, Describe the dual objective function. This represents the total number of typical subgroups. Indicates the first The dual variables of a typical subgroup.

[0054] Example 3, based on Example 1, specifically includes the following steps in generating teaching evaluation results: inputting the comprehensive feature set into the BiGRU model for initial training to generate teaching evaluation results.

[0055] Example 4, according to Figure 1 This embodiment is based on Embodiment 2. In this embodiment, the risk perception D3QN model is used to obtain the teaching decision results. The specific steps include:

[0056] Step B1: Construct a teaching status representation based on the teaching assessment results. The teaching action space includes basic content review, targeted training delivery, case video guidance, and assessment task arrangement; execute the teaching actions in the teaching action space, monitor changes in student status, and construct state transition pairs.

[0057] Step B2: Design a reward function, combine state transition pairs to generate immediate reward values; input the state transition pairs and immediate reward values ​​into the D3QN model for training, use the traditional Q-value function to evaluate the long-term benefits of the teaching actions, obtain the state-action reward distribution, and obtain the initial policy network;

[0058] Step B3: During training, a static spectral risk measurement mechanism is introduced to replace the training method in the D3QN model that uses the traditional Q-value function as the optimization objective. Specifically, the state-action reward distribution is modeled as a complete reward distribution. Based on the complete reward distribution, a risk spectral function is introduced to construct a piecewise linear approximation function as a risk-weighted mapping. The piecewise linear approximation function is embedded into the traditional Q-value function to construct a risk-sensitive Q-value function. This enables the perception and control of different risk quantiles in the teaching action reward distribution, generates a risk-sensitive Q-value, optimizes the policy network parameters in the D3QN model, and obtains the current teaching state and the optimized policy network.

[0059] The formula used in the static spectral risk measurement mechanism:

[0060] ;

[0061] in, This represents the risk adjustment function. Indicating in strategy Under, state and actions The risk-sensitive Q-value function; Indicates the scaling factor. Represents the mathematical expectation; Represents the risk spectrum function. This represents a piecewise linear approximation function. Indicates the offset item. Indicates the stretching factor. Represents the complete return distribution;

[0062] Piecewise linear approximation function formula:

[0063] ;

[0064] in, This represents the predicted return under the state-action combination. This indicates the number of segments in the state-action reward distribution. Indicates the first Each return quantile, Indicates the first The weighting coefficients for each segment, Indicates the first quantiles corresponding to the cumulative distribution function Indicates the first The segmented intermediate points are used for numerical integration approximation; This represents the risk spectrum function value at quantiles. Risk preference weights This represents the risk spectrum function value at quantiles. Risk preference weights; Represents the negative part of the function;

[0065] Step B4: Based on the current teaching status and the optimization strategy network, select the optimal teaching action from the teaching action space and generate the teaching decision result.

[0066] Example 5, according to Figure 1 This embodiment is based on Embodiment 2. In this embodiment, the risk perception D3QN model is used to obtain the teaching decision results. The specific steps include:

[0067] Step R1: Construct a teaching status representation based on the teaching assessment results. The teaching action space includes basic content review, targeted training delivery, case video guidance, and assessment task arrangement; execute the teaching actions in the teaching action space, monitor changes in student status, and construct state transition pairs.

[0068] Step R2: Design a reward function, combine it with state transition pairs to generate immediate reward values; input the state transition pairs and immediate reward values ​​into the D3QN model for training, use the traditional Q-value function to evaluate the long-term benefits of teaching actions, and obtain the current teaching state and policy network;

[0069] Step R3: Based on the current teaching state and the strategy network, select the optimal teaching action from the teaching action space and generate the teaching decision result.

[0070] Example 6, according to Figure 2 This embodiment is based on Embodiment 4. In this embodiment,

[0071] The teaching evaluation module introduces the Fair Constrained Dual-BiGRU model. This model is constructed by building a BiGRU model and incorporating a K-Means subgroup partitioning mechanism and a subgroup fairness constraint mechanism during BiGRU model training to optimize its evaluation performance. The Fair Constrained Dual-BiGRU model processes the comprehensive feature set to generate teaching evaluation results, including a knowledge point matrix, a weakness diagnosis report, and a teaching adaptability curve. The Fair Constrained Dual-BiGRU model is a BiGRU model.

[0072] The teaching evaluation results for student S_17 are shown in Table 1:

[0073] Table 1. Teaching Evaluation Results (Student S_17)

[0074] ;

[0075] Student S_17's Weakness Diagnosis Report:

[0076] High-risk knowledge points:

[0077] K4: Management of thyroid storm - Low correct answer rate (29%), high interruption rate of corresponding case video playback, and students' learning time in this module is much lower than the class average;

[0078] K2: Hypoglycemia emergency procedures - There is a tendency to confuse the correct options, with many errors concentrated in the questions about the order of operations;

[0079] The teaching adaptability curve for student S_17 from weeks 1 to 8 is shown in the attached figure; the left side represents the teaching adaptability score, and the bottom side represents the week.

[0080] The teaching decision module defines the teaching action space and introduces the risk-aware D3QN model. The risk-aware D3QN model is constructed by building a D3QN model and introducing a static spectral risk measurement mechanism to optimize the traditional Q-value function of the D3QN model during the training process. The risk-aware D3QN model is used to assess the risk sensitivity of teaching actions in the teaching action space and optimize strategies to obtain teaching decision results.

[0081] Table 2 shows the teaching decision results for S_17, as shown in Table 1:

[0082] Table 2. Results of Teaching Decisions (Student S_17)

[0083] .

[0084] The present invention and its embodiments have been described above. This description is not restrictive. The accompanying drawings are only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this description and designs a similar structure and embodiment without departing from the spirit of the present invention, such design should fall within the protection scope of the present invention.

Claims

1. A deep learning-based endocrine nursing teaching decision-making system, comprising a data preprocessing module, wherein the data preprocessing module generates a cleaning teaching dataset; characterized in that: The system also includes a feature extraction module, a teaching evaluation module, and a teaching decision-making module; The feature extraction module extracts teaching content features and student behavior features from the cleaned teaching dataset to obtain a comprehensive feature set; The teaching evaluation module introduces the Fair Constrained Dual-BiGRU model, which processes the comprehensive feature set to generate teaching evaluation results. The Fair Constrained Dual-BiGRU model includes the BiGRU model. The teaching decision module defines the teaching action space, introduces the risk perception D3QN model, and uses the risk perception D3QN model to conduct risk sensitivity assessment and strategy optimization of teaching actions in the teaching action space based on teaching evaluation results, thereby obtaining teaching decision results. The risk perception D3QN model includes the D3QN model; the teaching action space includes basic content review, targeted training push, case video guidance, and assessment task arrangement. The process of generating teaching evaluation results using a fair constraint dual-BiGRU model includes the following steps: Step S1: Input the comprehensive feature set into the BiGRU model for initial training and generate initial teaching evaluation results; Step S2: Set a reference evaluation benchmark value, perform multi-dimensional feature division on the comprehensive feature set using K-Means, obtain typical subgroups, and generate subgroup evaluation results; introduce a subgroup fairness constraint mechanism to constrain the deviation between the mean of the subgroup evaluation results and the reference evaluation benchmark value, and generate subgroup error constraints. Step S3: Using the subgroup error constraint as a training constraint, the training task of the BiGRU model is formalized into a multi-fairness constraint-nonconvex optimization problem; the primal-dual optimization algorithm is introduced to transform the multi-fairness constraint-nonconvex optimization problem into a dual objective function; during the training process, the dual objective function is used to iteratively optimize the principal parameters and dual variables of the Bi-GRU model, optimize the initial teaching evaluation results, and generate teaching evaluation results. The process of obtaining teaching decision results using the risk-perception D3QN model includes the following steps: Step B1: Construct a teaching state representation based on the teaching evaluation results, execute teaching actions, and construct state transition pairs; Step B2: Design a reward function, combine state transition pairs to generate immediate reward values; input the state transition pairs and immediate reward values ​​into the D3QN model for training, use the traditional Q-value function to evaluate the teaching actions, and obtain the state-action reward distribution; Step B3: During the training process, a static spectral risk metric mechanism is introduced to replace the traditional Q-value function, resulting in an optimized policy network; Step B4: Based on the optimized strategy network, select the optimal teaching action and generate teaching decision results.

2. The deep learning-based endocrine nursing teaching decision-making system according to claim 1, characterized in that: The teaching content features include textual semantic features, case difficulty labels, and knowledge point associations; the student behavior features include time-series learning trajectories, question mastery assessments, and interaction preference distributions.

3. The deep learning-based endocrine nursing teaching decision-making system according to claim 1, characterized in that: Typical subgroups include: high-frequency interactive students, low-score review students, video-dependent students, fast-paced learning students, and exam-oriented students.

4. The deep learning-based endocrine nursing teaching decision-making system according to claim 1, characterized in that: The static spectral risk measurement mechanism in step B3 is as follows: the state-action reward distribution is modeled as a complete reward distribution. Based on the complete reward distribution, a risk spectrum function is introduced to construct a piecewise linear approximation function. The piecewise linear approximation function is embedded into the traditional Q-value function to construct a risk-sensitive Q-value function, generate a risk-sensitive Q-value, optimize the policy network parameters in the D3QN model, and obtain the optimized policy network.

Images