Evaluation method of medical and preventive integration skills based on causal-machine model

Abstract

The present invention relates to the field of educational technology, and in particular to a method for evaluating medical-preventive integration skills based on a causal-machine model, comprising the following steps: generating a question feature vector and a knowledge point semantic vector using natural language processing technology combined with a deep learning model, and calculating the correlation between the feature vectors using cosine similarity; generating a question difficulty coefficient based on the causal-machine model; generating a student knowledge mastery feature vector based on the student's answer accuracy and answering speed; recommending the most suitable homework questions to the student based on the deep learning model; and providing feedback and learning suggestions by quantitatively analyzing the student's answering performance through the causal-machine model. The present invention can identify students' weak links, effectively improve the efficiency of preventive medicine professional learning, assist teachers in making teaching decisions, and promote the personalized training of medical-preventive integration and medical-preventive collaborative talents.

Description

Technical Field

[0001] The present invention relates to the field of educational technology, and in particular to a method for evaluating medical-prevention integrated skills based on a causal-machine model. Background Art

[0002] In today's era, medical education faces unprecedented challenges and opportunities. With the deepening of health concepts and the growing demand for public health services, the integration of medical care and prevention has become an indispensable component of modern medical education. This process requires students to not only master traditional fields such as basic medicine and clinical medicine, but also to gain a deep understanding of related disciplines such as preventive medicine. However, students' understanding of these courses varies greatly, exhibiting distinct individual characteristics.

[0003] With the rapid development of information technology, artificial intelligence and big data technologies have gradually been introduced into the education field, giving rise to intelligent teaching systems. However, while many current online education platforms have a certain degree of homework push and automatic grading capabilities, they rely on fixed rules or simple question difficulty distribution models, or methods based on knowledge graphs and simple machine learning, making it difficult to effectively meet students' personalized learning needs. Homework push methods based on preset rules cannot adjust the push content in real time based on students' knowledge mastery. When recommending homework, these systems often have difficulty accurately analyzing the degree of match between questions and course knowledge points, and are unable to dynamically track students' knowledge mastery. This results in recommended questions that are either too simple to effectively improve learning outcomes, or too complex to be handled by students.

[0004] While knowledge graph-based recommendation methods attempt to link knowledge points and questions with students' learning behaviors through the structure of nodes and edges, and can provide a breadth of information in certain scenarios, they have various limitations in practical applications. First, the creation of a knowledge graph requires a large amount of manual annotation and expert knowledge, involving the definition of knowledge points and the establishment of relationships. This is not only time-consuming, but may also introduce subjective biases, resulting in insufficient accuracy and timeliness of the graph. Secondly, knowledge graphs are difficult to update in real time, and their fixed structure cannot effectively adapt to students' learning progress and personalized needs. Finally, they also have deficiencies in reasoning ability, and simple similarity calculations may not accurately reflect students' real learning needs and knowledge weaknesses.

[0005] The "Causal-Machine Learning" model constructs a causal relationship graph to clarify the dependency paths between questions and knowledge points. It also incorporates fuzzy theory to generate causal prior knowledge and optimizes the network structure using a partitioned MCMC algorithm to estimate the conditional probability distribution between variables. Based on the causal inference results, it can quantify the impact of different questions on student learning outcomes, avoiding the "spurious correlation" problem in traditional machine learning. Combined with the nonlinear modeling capabilities of deep learning, the "Causal-Machine Learning" model can dynamically deliver exercises that best suit students' current learning needs and provide teachers with more accurate learning feedback and personalized guidance. Deep learning models can process large-scale educational data, revealing the complex relationships between questions and knowledge points, and making personalized recommendations based on students' learning behavior and knowledge level. Furthermore, natural language processing technology can extract features from the text and question content of course materials to accurately identify the relevance between questions and knowledge points. However, a comprehensive and integrated approach for intelligent exercise delivery and feedback evaluation in the integration of medical and preventive medicine and collaborative talent development is currently lacking. Therefore, there is an urgent need for an innovative intelligent system that fully combines "causal-machine learning" and natural language processing technology, which can dynamically push exercises that best suit students' current learning needs based on their knowledge mastery, while providing intelligent learning feedback to improve students' learning outcomes in the integration of medical and preventive care and the cultivation of collaborative talents in medical and preventive care, and assist teachers in providing personalized teaching guidance. Summary of the Invention

[0006] The purpose of the present invention is to provide a medical and preventive integration skills evaluation method based on a causal-machine model, which can intelligently recommend homework questions suitable for students' knowledge level and generate targeted learning feedback.

[0007] To achieve the above object, the present invention adopts the following technical solutions:

[0008] A method for evaluating medical and preventive integrated skills based on a causal-machine model includes the following steps:

[0009] Step 1: Obtain the text of each question in the core skills question bank for preventive medicine talent training, as well as the text of the knowledge points in the professional training program course materials. Use natural language processing technology combined with a pre-trained BERT model to quantify the question text and knowledge point text to generate question semantic vectors and knowledge point semantic vectors. Use cosine similarity to calculate the correlation between the question feature vector and the knowledge point semantic vector. Based on this correlation, obtain the associated knowledge feature vector for the question.

[0010] Step 2: Based on the historical completion rate, historical answer accuracy rate, and teacher scoring difficulty of the question, the question difficulty coefficient is generated based on the causal-machine model. The difficulty coefficient, question semantic vector, and associated knowledge semantic vector are weighted and fused using the attention mechanism to generate the question feature vector for each question.

[0011] Step 3: Obtain students' answer results in the core skills question bank of the preventive medicine professional training program. Generate a student knowledge mastery feature vector based on the student's answer accuracy and answering speed, combined with the question semantic vector. This vector reflects the student's mastery of a certain knowledge point. Accumulate the student knowledge mastery feature vectors of related questions based on the associated knowledge feature vectors to form a comprehensive mastery feature vector. Extract the comprehensive mastery feature vectors for different knowledge points to form a coverage feature.

[0012] Step 4: Based on the question feature vector, knowledge point semantic vector, and student's comprehensive mastery feature vector, DeepFM, a deep learning model, recommends the most suitable homework questions for students, prioritizing questions that can help students consolidate their weaknesses or improve their current level.

[0013] Step 5: Automatically analyze students’ performance based on the causal-machine model and provide intelligent feedback to teachers; generate learning suggestions for students’ weak points in knowledge to help teachers adjust teaching plans in a timely manner.

[0014] Furthermore, in step 1, the BERT model is used to perform context-related semantic modeling on each word in the question. Specifically, the question text and knowledge point text are first segmented into smaller units using the BERT model's preprocessor; then, the segmented words are processed through the BERT model to generate a context-related embedding vector for each word; and then the generated context-related embedding vectors are aggregated to finally generate a question semantic vector.

[0015] Furthermore, the difficulty coefficient of the question generated based on the causal-machine model in step 2 includes:

[0016] First, by analyzing the completion rate, accuracy rate, and scoring difficulty of the questions, a causal relationship diagram between question characteristics and knowledge point characteristics was constructed;

[0017] Then, fuzzy prior weights are generated based on domain expert knowledge and historical data to generate fuzzy membership for edges in the causal graph;

[0018] Finally, Bayesian network optimization uses the partitioned MCMC algorithm to optimize the structure of the causal relationship graph, and conditional probability distribution estimation estimates the conditional probability distribution of question features based on the optimized Bayesian network and generates the question difficulty coefficient.

[0019] Furthermore, in step 3, the student's knowledge mastery feature vector is regularly updated to dynamically update the student's knowledge mastery status.

[0020] Furthermore, the factorization machine model of the DeepFM model captures the direct interaction between the question feature vector and the comprehensive mastery feature vector to obtain the matching degree of a certain type of question to a specific student; the dot product between the two vectors is calculated to evaluate the basic matching degree between the student and the question; the multi-layer neural network of the DeepFM model captures the complex nonlinear relationship between the question feature vector and the comprehensive mastery feature vector, and obtains accurate recommendation results from a more complex pattern.

[0021] Furthermore, reinforcement learning is introduced into the DeepFM model, and each student's current knowledge mastery and homework completion status are used as the state in reinforcement learning to recommend a series of questions to students. After each recommendation, the recommendation strategy is adjusted according to the student's answer performance.

[0022] Furthermore, the reinforcement learning adopts an off-policy learning method based on a value function.

[0023] Furthermore, a reward mechanism is introduced to evaluate the recommendation results; when students perform well on the recommended topics, positive rewards are given; if students perform poorly on the recommended topics, negative feedback is given.

[0024] Furthermore, the homework questions are recommended to students to consolidate their weak points or to improve their current level.

[0025] Furthermore, the analysis of students' answering performance in step 5 refers to the assessment of students' knowledge and the assessment of question quality.

[0026] Furthermore, in step 5, the automatic analysis of students' answering performance using the causal-machine model refers to: the evaluation of students' knowledge is to assess the accuracy of identifying students' knowledge levels and the accuracy of identifying students' knowledge weaknesses, calculated as the proportion of successfully identified weak knowledge points;

[0027] The quality assessment of questions is to calculate the completion rate of students after the recommended homework, reflecting the attractiveness and applicability of the recommendation; and calculate the overlap rate between the semantic vectors of the knowledge points involved in the recommended homework and the semantic vectors of the knowledge points that students have not yet mastered.

[0028] The present invention has the following beneficial effects:

[0029] (1) Combining the "causal-machine" model with natural language processing, relying on a data-driven model, it can automatically learn and adapt to students' learning patterns, thereby reducing labor costs. At the same time, it can comprehensively consider multi-dimensional student data, including answering speed and answering accuracy, and then provide more comprehensive personalized recommendations and accurately identify students' weak links. Compared with traditional methods based on user behavior analysis, this invention uses reinforcement learning technology to achieve intelligent optimization of homework recommendation strategies, making the recommendation process more intelligent and personalized, and reducing the problems caused by collecting complex data.

[0030] (2) Personalized homework recommendations for the medical and preventive medicine integration talent training course can intelligently recommend homework topics that match students' knowledge level and learning needs by analyzing their understanding of preventive medicine course knowledge points;

[0031] (3) Improve the effectiveness of preventive medicine learning. By dynamically adjusting the recommended homework content, students can learn within the correct knowledge framework, ensuring that their basic knowledge, data analysis skills and application capabilities in the field of preventive medicine are improved, thereby improving the overall learning effect.

[0032] (4) Assist teachers to optimize teaching strategies, gain a deeper understanding of students’ learning progress and weak links in the cultivation of integrated medical and preventive medicine talents, and thus adjust teaching plans and strategies. For example, if many students perform poorly in the “regression analysis” section, teachers can strengthen the teaching of relevant knowledge points in a targeted manner. By collecting and analyzing data such as students’ homework completion rates and answer accuracy, a basis for educational decision-making is provided, helping educational administrators to formulate more reasonable preventive medicine course designs and teaching arrangements, thereby improving the overall teaching quality.

[0033] (5) Promote the intelligent transformation of medical and preventive integrated talent training, combine deep learning and data mining technology, and provide new solutions for the intelligent development of preventive medicine education, which will help promote the personalized training of medical and preventive collaborative talents and improve the quality of education. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a schematic diagram of the process of the present invention.

[0035] Figure 2 This is a flowchart of the deep learning model processing of the present invention. DETAILED DESCRIPTION

[0036] like Figure 1 As shown, this embodiment provides a method for evaluating medical and preventive integrated skills based on a causal-machine model, comprising the following steps:

[0037] Step 1: Obtain the text of each question in the core skills question bank for preventive medicine talent training, as well as the knowledge point text of the professional training program course materials; use natural language processing (NLP) technology combined with a pre-trained deep learning model to quantify the question text and knowledge point text to generate question semantic vectors and knowledge point semantic vectors, use cosine similarity to calculate the correlation between the question feature vector and the knowledge point semantic vector, and obtain the associated knowledge feature vector of the question based on the correlation.

[0038] This embodiment uses a pre-trained BERT model to perform context-sensitive semantic modeling on each word in the question, thereby obtaining a more accurate representation of the question's semantic features. First, the BERT model's preprocessor is used to segment the question text and knowledge point text into smaller units (into words or subwords, called "tokens"). Special markers [CLS] and [SEP] are added at the beginning and end of the text to indicate the start and end of the text. For example, "Question 1: Please briefly describe the difference between standard deviation and standard error." After tokenization, it becomes "[CLS], please, briefly describe, standard deviation, and, standard error, of, different, [SEP]".

[0039] The segmented words are processed through the BERT model to generate context-dependent embedding vectors for each word. The BERT model not only considers the semantics of the word itself, but also combines the context to generate more accurate embedding vectors, which can preserve subtle changes in vocabulary in different contexts. For example, in question 1, the BERT model will consider the relationship between "standard deviation" and "briefly describe" and "different", distinguishing this question from a simple calculation question on the standard deviation. For each question, the model summarizes the context-dependent vectors it generates and ultimately generates a question semantic vector. Similarly, for each knowledge point, a corresponding knowledge point semantic vector is also generated. These vectors can represent the semantic content of the question or knowledge point.

[0040] Each question usually involves one or more knowledge points. The similarity between the question and each knowledge point is expressed by calculating the cosine similarity between the question semantic vector and the knowledge point semantic vector. Cosine similarity calculates the cosine value of the angle between two vectors. The closer the value is to 1, the closer the semantics of the two vectors are, and thus the greater the correlation between the question semantic vector and the knowledge point semantic vector.

[0041] by Represents the semantic vector of question 1, The cosine similarity calculated for the semantic vector of knowledge point a "standard deviation" is approximately 0.95, indicating that students should be able to apply the content of knowledge point a "standard deviation definition" well when answering this question. At the same time, it also means that knowledge point a "standard deviation" is the core of understanding the question. Similarly, and The cosine similarity of the semantic vector of knowledge point b, "Definition of standard error," is 0.90. Although slightly lower than that of knowledge point a, it is still an important supplementary content. If students can combine the definitions of the two, they will be able to answer the question more comprehensively. However, compared with knowledge point c, "Describe the calculation method of standard deviation," the cosine similarity is 0.85. Although it is still relevant, it is slightly broader than the first two knowledge points. Students may know some related concepts, but they are not as specific as the first two. Therefore, the relevant knowledge points of Question 1 are more likely to be labeled as the definition of standard deviation and the definition of standard error, rather than the calculation method of standard deviation. After calculating the cosine similarity with multiple knowledge points, this embodiment obtains the relevant features of the question and the knowledge point, "Question 1: Please briefly describe the difference between standard deviation and standard error": {"definition of standard deviation": 0.95, "definition of standard error": 0.90, "method of calculating standard deviation": 0.80, ...}".

[0042] Step 2: Based on the historical completion rate, historical answer accuracy rate, and teacher scoring difficulty of the questions, the causal-machine model is used to generate the question difficulty coefficient. The causal relationship graph is used to construct the association path between the questions and knowledge points. Fuzzy theory is combined to generate prior weights. The attention mechanism is used to weightedly fuse the difficulty coefficient, question semantic vector, and associated knowledge semantic vector to generate the question feature vector for each question.

[0043] Causal relationship graph construction analyzes question completion rate, accuracy rate, and scoring difficulty to construct a causal relationship graph between question characteristics and knowledge point characteristics. For example, for "Question 1" {"Completion Rate": 0.75, "Accuracy Rate": 0.60, "Scoring Difficulty": 3.5} and "Question 2" {"Completion Rate": 0.85, "Accuracy Rate": 0.90, "Scoring Difficulty": 2.0}, the model identifies that completion rate and accuracy rate not only directly affect question difficulty but also indirectly influence difficulty estimates through associated knowledge point characteristics.

[0044] Fuzzy prior weight generation uses domain expert knowledge and historical data to generate fuzzy membership for edges in the causal graph. For example, the model generates a weight of 0.8 for the relationship "completion rate → difficulty coefficient" and a weight of 0.6 for the relationship "accuracy rate → difficulty coefficient." These prior weights ensure that the model focuses on critical paths during learning.

[0045] Bayesian network optimization uses a partitioned MCMC algorithm to optimize the structure of the causal relationship graph, ensuring efficient capture of complex dependencies between variables. For example, for Question 1, the model generated an indirect path through optimization: "Related Knowledge Points → Completion Rate → Difficulty Level," indicating that the coverage of related knowledge points in this question is an important implicit variable influencing the completion rate.

[0046] Conditional probability distribution estimation uses an optimized Bayesian network to estimate the conditional probability distribution of question features. For example, the model calculated a difficulty coefficient of 0.68 based on the completion and accuracy rates for Question 1, while the difficulty coefficient for Question 2 was 0.34, indicating that Question 1 was significantly more difficult than Question 2. This result is a comprehensive analysis of both explicit data (completion and accuracy rates) and implicit relationships (knowledge point coverage).

[0047] Through the above steps, after the model generates the question difficulty coefficient, it uses the attention mechanism to weightedly fuse the difficulty coefficient, question semantic vector and associated knowledge semantic vector to generate the question feature vector.

[0048] For example, in the essay question, "In nonparametric methods, what are the methods for comparing survival rates? What are their characteristics?" Knowledge point characteristics: {"Kaplan-Meier method": 0.95, "Log-rank test": 0.90, ...} The model uses a causal relationship diagram to identify the path from "knowledge point coverage → completion rate → difficulty coefficient," demonstrating that knowledge point coverage is the core factor affecting difficulty. The difficulty coefficient is estimated to be 0.80 using fuzzy weights and conditional probabilities.

[0049] Multiple-choice question: "A disease is affected by smoking (X1) and drinking (X2). Develop a Cox regression model and interpret the regression coefficients." Knowledge point characteristics: {"Cox regression model": 0.95, ...} The causal relationship diagram shows a high weight for the direct path from "completion rate → difficulty coefficient," indicating that completion rate is the primary influencing factor for this question. The difficulty coefficient is estimated to be 0.60.

[0050] In the final fusion process, the model assigns higher weights to difficulty features, highlighting features with lower complexity but higher problem-solving requirements.

[0051] Step 3: Obtain the students’ answer results in the core skills question bank of the preventive medicine professional training program, and use the students’ correct answer rate to and answering speed Combined with the question semantic vector Generate student knowledge mastery feature vector; , reflecting the degree of students’ mastery of a certain knowledge point; according to the associated knowledge feature vector, the student’s knowledge mastery feature vector of the related questions is accumulated to form a comprehensive mastery feature vector, and the comprehensive mastery feature vectors of different knowledge points are extracted to form the coverage feature.

[0052] For example, when students are doing the question feature vector If a student's speed or accuracy is low when answering questions related to a specific problem, even if they complete the question, it indicates that they have not mastered the knowledge point, which will be reflected in the semantic vector for that question. For example, for question 5, "Please explain the basic principles of the linear regression model," the student's accuracy is 0.5 and their speed is 0.4. For question 6, "When is the logistic regression model used? Please give an example," the student's accuracy is 0.8 and their speed is 0.2. Finally, the student's knowledge mastery feature vector is weighted by accuracy and speed. Although the student did cover the knowledge points of both questions 5 and 6, their performance on question 6 was very good (high accuracy and moderate speed), and their mastery feature vector reflects a good understanding of the relevant knowledge point. In contrast, for question 5, due to their poor performance (low accuracy and slow speed), the mastery feature vector is low. This correction mechanism helps accurately assess students' mastery of knowledge points and provides targeted support for subsequent learning.

[0053] For each knowledge point, students often complete multiple questions of different difficulty levels and types. Therefore, this embodiment needs to combine the answer results of multiple questions to generate the student's comprehensive mastery of a specific knowledge point. For example, the comprehensive mastery feature vector of a student's knowledge point "standard deviation" can be expressed as: By accumulating multiple related question feature vectors, it is possible to evaluate the student's overall mastery of a knowledge point.

[0054] Dynamically update a student's knowledge mastery of a particular knowledge point. Because a student's knowledge mastery is not static but constantly changes over time and with practice, the student's knowledge mastery feature vector is regularly updated to more accurately reflect their learning progress. For example, when a student retakes a question or demonstrates improvement in a new exercise, the student's knowledge mastery feature vector is dynamically adjusted based on the latest results and time records. By generating a student's knowledge mastery feature vector for each knowledge point, we can form a comprehensive understanding of that student's learning status.

[0055] Step 4: Based on the question feature vector, knowledge point semantic vector and student's comprehensive mastery feature vector, DeepFM, a deep learning model, recommends the most suitable homework questions for students, giving priority to questions that can help students consolidate their weak links or improve their current level.

[0056] The linear component of the DeepFM model uses a factorization machine model to capture the direct interaction between the question feature vector and the comprehensive mastery feature vector, determining the matching degree of a particular question type for a specific student. The dot product between the two vectors is calculated to assess the basic matching degree between the student and the question. The factorization model accurately calculates the linear matching degree between the student and the question. For example, if a student performs poorly on questions related to "regression analysis," the factorization module will capture this relationship and provide foundational information for subsequent recommendations.

[0057] The deep part of the DeepFM model uses a multi-layer neural network to capture the complex nonlinear relationship between the question feature vector and the comprehensive mastery feature vector, ensuring that recommendations can derive accurate results from more complex patterns. Not only does it consider the question feature vector and the comprehensive mastery feature vector of the question, but it also deeply explores the high-order interactions between features. The question feature vector and the comprehensive mastery feature vector are input into multiple hidden layers, and nonlinear feature interactions are extracted layer by layer through activation functions. Through the processing of multi-layer neural networks, the complex nonlinear relationship between students and questions can be captured. For example, a student's performance on a data analysis question may have a high-order connection with their performance on a statistical inference question. The deep part identifies and utilizes this complex relationship to achieve more accurate personalized recommendations.

[0058] Building on the existing DeepFM model, this implementation incorporates reinforcement learning to optimize recommendations by continuously adjusting the recommendation strategy. Each student's current knowledge and homework completion status is used as the "state" in reinforcement learning, including information such as historical homework performance and knowledge point mastery. The "action" is to recommend a series of questions to the student. After each recommendation, the recommendation strategy is adjusted based on the student's answer performance (including accuracy rate, time spent answering questions, and learning fatigue). The effectiveness of the recommendations is evaluated through a reward mechanism.

[0059] Specifically, when students perform well on recommended questions (increased accuracy and improved knowledge mastery), they will be given positive rewards. Conversely, if students' performance on recommended questions fails to improve, negative feedback will be given, thereby adjusting the model parameters and optimizing the recommendation strategy. The focus is not only on short-term benefits, such as students' performance on the current question, but also on long-term benefits, that is, the students' continued progress throughout the learning process. Through dynamic adjustments, questions that best suit students' current learning status and long-term development can be gradually recommended, thereby maximizing learning outcomes. The introduction of reinforcement learning enables the recommendation system to be continuously optimized. It not only relies on static feature matching, but also can dynamically adjust the recommendation strategy according to students' learning progress. It can automatically identify students' weaknesses and gradually introduce more difficult questions as students' performance improves, helping students achieve continuous progress.

[0060] For example, if a student's mastery of "Cohort Study Concepts" is 0.65 and their mastery of "Clinical Trial Design" is 0.46, we will recommend difficult questions on "Cohort Study" and relatively simple questions on "Clinical Experimental Study Design." Suppose the student completed the "Clinical Trial Design" questions well (with high speed and accuracy) but had lower accuracy and speed on the difficult "Cohort Design" questions. Based on the reward mechanism, we will update the student's learning status and fine-tune their mastery of these two related knowledge points. The updated results will be used for the next recommended questions.

[0061] By combining the DeepFM model with the dynamic adjustment mechanism of reinforcement learning, we innovatively implement intelligent recommendations tailored to students' personalized learning needs. The DeepFM model accurately captures the interaction between question characteristics and student knowledge mastery, while the application of reinforcement learning ensures the effectiveness of the recommendation system. Feedback is generated for teachers based on students' responses, identifying questions where they performed poorly and, through the knowledge points associated with the questions, determining which knowledge points the student has yet to master. Feedback is generated based on the student's incorrect answers. The system lists the knowledge points involved in the questions they answered incorrectly and provides relevant learning suggestions.

[0062] Step 5: The causal-machine model automatically analyzes students’ performance and provides intelligent feedback to teachers. It generates learning suggestions targeting students’ weaknesses, helping teachers adjust their teaching plans in a timely manner.

[0063] Automatically record and analyze student performance on homework and tests, focusing on accuracy, which tracks whether students answered each question correctly. Time spent answering questions reflects students' familiarity with the questions and their difficulty, analyzed based on the match between time spent answering questions and question complexity. Track students' historical answers, particularly to determine whether their performance on the same knowledge point has improved or declined over time.

[0064] To assess student knowledge, the "Cause-to-Machine" model combines causal analysis of predicted and actual test results to evaluate the accuracy of student knowledge identification. Causal inference is also used to identify weaknesses in students' knowledge and calculate the percentage of successfully identified weak points. For example, causal relationship diagrams are used to analyze students' performance on the knowledge points of "Experimental Design" and "Causal Inference," accurately pinpointing weak points and generating targeted supplementary suggestions.

[0065] To assess the quality of assignments, the "Cause-to-Machine" model calculates the completion rate of recommended assignments, reflecting the appeal and applicability of the recommendations. Furthermore, based on the causal graph, the overlap between the knowledge points covered in the recommended assignments and those that the students have not yet mastered is calculated to optimize the coverage of the recommended assignments. For example, if the majority of recommended assignments are completed, this indicates that the difficulty level is well matched to the student's learning level. However, some assignments fail to cover key knowledge points related to "causal inference," indicating that the system needs to further optimize this coverage.

[0066] The final evaluation knowledge constraint is C = {"{Student knowledge assessment: [Knowledge mastery accuracy rate, Knowledge weakness identification rate]}", "{Question quality assessment: [Homework completion rate, Knowledge point coverage rate]}". For example, a student has a good grasp of health statistics knowledge but still has some knowledge gaps in epidemiology. This can be used to determine their Knowledge mastery accuracy rate. Accurately identifying the student's weaknesses in "Experimental Design" and "Causal Inference" through data analysis can help to provide targeted supplementary recommendations and provide good feedback. This can be used to determine the student's Knowledge weakness identification rate. Most of the recommended homework questions are completed, indicating that the difficulty of the questions is well matched to the student's learning level. This can be used to determine the student's homework completion rate. Although the recommended questions are of high quality, some questions do not address the difficulty of "Causal Inference". Therefore, the system's question coverage of knowledge points needs further optimization. This can be used to determine the student's homework completion rate. The final knowledge assessment of the student was: [Knowledge mastery accuracy: 85%, knowledge weak link identification rate: 90%]", "Question quality assessment: [Homework completion rate: 95%, knowledge point coverage rate: 80%].

[0067] The above description is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any modification and replacement based on the technical solution and inventive concept provided by the present invention should be covered by the protection scope of the present invention.

Claims

1. A method for evaluating medical and preventive integrated skills based on a causal-machine model, characterized in that: The steps include: Step 1: Obtain the text of each question in the core skills question bank for preventive medicine talent training, as well as the text of the knowledge points in the professional training program course materials. Use natural language processing technology combined with a pre-trained BERT model to quantify the question text and knowledge point text to generate question semantic vectors and knowledge point semantic vectors. Use cosine similarity to calculate the correlation between the question semantic vector and the knowledge point semantic vector. Based on this correlation, obtain the associated knowledge feature vector for the question. Step 2: Based on the historical completion rate, historical answer accuracy rate, and teacher scoring difficulty of the question, the question difficulty coefficient is generated based on the causal-machine model. The difficulty coefficient, question semantic vector, and associated knowledge semantic vector are weighted and fused using the attention mechanism to generate the question feature vector for each question. Step 3: Obtain students' answer results in the core skills question bank of the preventive medicine professional training program. Generate a student knowledge mastery feature vector based on the student's answer accuracy and answering speed, combined with the question semantic vector. This vector reflects the student's mastery of a certain knowledge point. Accumulate the student knowledge mastery feature vectors of related questions based on the associated knowledge feature vectors to form a comprehensive mastery feature vector. Extract the comprehensive mastery feature vectors for different knowledge points to form a coverage feature. Step 4: Based on the question feature vector, knowledge point semantic vector, and student's comprehensive mastery feature vector, DeepFM, a deep learning model, recommends the most suitable homework questions for students, prioritizing questions that can help students consolidate their weaknesses or improve their current level. The factorization machine model of the DeepFM model captures the direct interaction between the question feature vector and the comprehensive mastery feature vector to obtain the matching degree of a certain question type to a specific student; The dot product between two vectors is calculated to assess the basic match between the student and the question. The DeepFM model's multi-layer neural network captures the complex nonlinear relationship between the question feature vector and the comprehensive mastery feature vector, deriving accurate recommendations from a more complex model. Introducing reinforcement learning into the DeepFM model, each student's current knowledge and homework completion status is used as the state in reinforcement learning to recommend a series of questions to students; After each recommendation, adjust the recommendation strategy based on the student's answer performance; Step 5: Automatically analyze students’ performance based on the causal-machine model and provide intelligent feedback to teachers; generate learning suggestions for students’ weak points in knowledge to help teachers adjust teaching plans in a timely manner.

2. The method for evaluating medical and preventive integrated skills based on the causal-machine model according to claim 1 is characterized in that: In step 1, the BERT model is used to perform contextual semantic modeling on each word in the question. Specifically, the BERT model preprocessor is used to segment the question text and knowledge point text into smaller units. Then, the segmented words are processed through the BERT model to generate context-related embedding vectors for each word; the generated context-related embedding vectors are then aggregated to finally generate a question semantic vector.

3. The method for evaluating medical and preventive integrated skills based on the causal-machine model according to claim 1 is characterized in that: The difficulty coefficients of the questions generated based on the causal-machine model described in step 2 include: First, by analyzing the completion rate, accuracy rate, and scoring difficulty of the questions, a causal relationship diagram between question characteristics and knowledge point characteristics was constructed; Then, fuzzy prior weights are generated based on domain expert knowledge and historical data to generate fuzzy membership for edges in the causal graph; Finally, Bayesian network optimization uses the partitioned MCMC algorithm to optimize the structure of the causal relationship graph, and conditional probability distribution estimation estimates the conditional probability distribution of question features based on the optimized Bayesian network and generates the question difficulty coefficient.

4. The method for evaluating medical and preventive integrated skills based on the causal-machine model according to claim 1 is characterized in that: In step 3, the student's knowledge mastery feature vector is updated regularly to dynamically update the student's knowledge mastery.

5. The method for evaluating medical and preventive integrated skills based on the causal-machine model according to claim 1 is characterized in that: The reinforcement learning adopts an off-policy learning method based on a value function.

6. According to the causal-machine model-based medical and preventive integration skill evaluation method described in claim 1, a reward mechanism is introduced to evaluate the recommendation results through the reward mechanism; when students perform well on the recommended topics, positive rewards are given; if students perform poorly on the recommended topics, negative feedback is given.

7. According to the method for evaluating medical and preventive integrated skills based on the causal-machine model in claim 1, homework questions are recommended to students to consolidate their weak links or to improve their current level.

8. According to the medical-prevention integrated skills evaluation method based on the causal-machine model of claim 1, the analysis of students' answering performance in step 5 refers to the assessment of students' knowledge and the assessment of question quality.

9. According to the method for evaluating medical and preventive integrated skills based on a causal-machine model in claim 1, in step 5, automatically analyzing the student's answering performance using the causal-machine model means: evaluating the student's knowledge level by accurately identifying the student's knowledge level and the accuracy of identifying the student's knowledge weaknesses, calculated as the proportion of successfully identified weak knowledge points; The quality assessment of questions is to calculate the completion rate of students after the recommended homework, reflecting the attractiveness and applicability of the recommendation; and calculate the overlap rate between the semantic vectors of the knowledge points involved in the recommended homework and the semantic vectors of the knowledge points that students have not yet mastered.

Images