Intelligent scheduling method and system based on standardized training of resident physicians, and medium

Abstract

The invention discloses an intelligent scheduling method and system based on resident physician standardized training and a medium, and relates to the technical field of medical informatization aided design, and the method comprises the steps: generating an initial scheduling scheme according to the training progress, department acceptance and standardized training requirements of resident physicians, and optimizing a scheduling result through a linear programming model, dynamic rotary shift arrangement is realized; in the scheduling process, the initial scheduling scheme is matched with a preset rotation template in combination with a cross-department rotation rule, a rotation department matching table is generated, and the rotation department matching table is associated with the training progress of the resident physician; the authority control rule and the data security management and control mechanism are used for performing level-to-level management on user roles; associating a rotation shift arrangement result with an examination plan of an online examination system, and dynamically adjusting the content and form of the standardized training examination according to the rotation stage of the resident physician; the invention provides an intelligent scheduling solution which can dynamically adapt to complex constraint conditions and realize global resource optimization.

Description

Technical Field

[0001] The present application relates to the field of medical information management technology, and in particular to an intelligent scheduling method, system and medium based on standardized training of resident physicians. Background Art

[0002] In medical human resource management, rationally arranging the rotation departments of residents is crucial to improving their professional skills and hospital operation efficiency.

[0003] In the standardized training management of resident physicians, existing scheduling solutions often rely on manual experience or simple automated tools based on fixed rules (such as Excel templates). While these solutions can meet basic scheduling needs, they cannot adaptively optimize scheduling based on dynamic adjustments to the national training syllabus (such as the addition of new training subjects and changes in rotation duration). Traditional algorithms also easily lead to uneven distribution of human resources between departments (such as some departments being overloaded while others are idle). This makes it difficult to simultaneously meet national standards, departmental needs, and personalized physician training goals. There is an urgent need for an intelligent scheduling solution that can dynamically adapt to complex constraints and achieve global resource optimization. Summary of the Invention

[0004] In order to provide an intelligent scheduling solution that can dynamically adapt to complex constraints and achieve global resource optimization, this application provides an intelligent scheduling method, system and medium based on standardized training of resident physicians.

[0005] In the first aspect, the invention objectives of this application are achieved by adopting the following technical solutions: Intelligent scheduling method based on standardized resident training, including: Based on the training progress of residents, the department's acceptance capacity and standardized training requirements, a genetic algorithm is used to generate the initial scheduling plan, and the scheduling results are optimized through a linear programming model to achieve dynamic rotation scheduling; During the scheduling process, a rotation department matching table is generated based on the initial scheduling plan combined with the cross-department rotation rules and the preset rotation template, and the rotation department matching table is associated with the resident physician's training progress; Permission control rules and data security management mechanisms for hierarchical management of user roles; According to the rotation department matching table and the authority control rules, the rotation scheduling results are associated with the assessment plan of the online examination system, and the content and form of the standardized training examination are dynamically adjusted according to the rotation stage of the resident physician.

[0006] By adopting the above technical solutions, an intelligent scheduling solution is provided that can dynamically adapt to complex constraints and achieve global resource optimization. The cross-department rotation rules include cross-department rotation sequence and department capacity restrictions; the standardized training requirements are the requirements of the national syllabus (Standardized Training Syllabus for Resident Physicians); based on the initial scheduling plan and the personalized cross-department rotation rules adopted by the corresponding hospital, further adaptation is carried out, and an executable rotation department matching table is generated through a preset rotation template (such as undergraduate / master's differentiated rotation table); the rotation department matching table directly determines the rotation stage of the resident physician (such as "internal medicine rotation period" and "surgery rotation period"), and the examination plan is dynamically adjusted based on this stage (such as the internal medicine stage tests the medical history collection ability, and the surgical stage tests the surgical operation ability), the department Balanced rotation teaching and flexible scheduling that can flexibly adapt to changes in the national syllabus (standardized training syllabus for resident physicians) are conducive to achieving optimal resource allocation; based on the RBAC permission model, user accounts of different user roles are set up with permission control rules (such as only department directors can modify scheduling rules) and a data security management system (such as transaction processing to prevent scheduling interruptions) to provide a dynamic, compliant and personalized intelligent scheduling method; dynamic adjustment can cope with complex and changing training needs, and human resources can be optimized and utilized, thereby improving the training quality of standardized training for resident physicians and improving hospital operation efficiency.

[0007] In a preferred embodiment of the present application, the genetic algorithm is used to generate an initial scheduling plan, and the linear programming model is used to optimize the scheduling results, specifically including: Receive resident training progress information, department admission capacity, and historical scheduling data, and establish scheduling constraints; Genetic algorithms are used to generate multiple candidate scheduling schemes, and the initial scheduling scheme that satisfies department load balance is selected through crossover and mutation operations; A linear programming model is used to perform secondary optimization on the initial scheduling plan to minimize the overtime hours of residents and output a scheduling result that meets the requirements of standardized training.

[0008] By adopting the above technical solutions, the problem that traditional scheduling methods cannot balance department load and resident physician working hours, resulting in low scheduling efficiency and easy to cause fatigue, is solved. By obtaining the resident physician training progress, the real-time capacity of the department and the historical scheduling records, a constraint condition set containing the upper limit of department load and the upper limit of physician working hours is constructed. Multiple scheduling candidate schemes are generated through genetic algorithms, and the crossover mutation operation is used to screen out the initial scheme that meets the department load balancing rate. The linear programming model is used to perform secondary optimization of the initial scheme, with minimizing the resident physician overtime as the objective function, and the final schedule that meets the upper limit of daily working hours (e.g., ≤8 hours) is output to improve the utilization of department resources.

[0009] In a preferred embodiment of the present application, the method further includes: Dynamically generate a phased competency assessment report based on residents' mandatory viewing progress and correct answer rate in the medical video learning system; Match the resident physician's competency assessment results with the assessment indicators of the current rotation department, and automatically adjust the weights of teaching case discussions and practical tasks in subsequent rotation plans.

[0010] By adopting the above technical solution, the problem that the existing training system is unable to dynamically adjust the weight of practical tasks according to learning effects, resulting in a mismatch between capabilities and training content, is solved. By real-time capturing the viewing progress of the medical video learning system (such as a completion rate ≥ 98% for viewing completion), the accuracy of answering questions, and the knowledge point mastery labels, a phased capability assessment model is constructed based on the BP neural network to generate an assessment report including theoretical scores and operational capabilities. By cross-comparing the assessment results with department assessment indicators (such as the case discussion compliance rate), the weight of practical tasks corresponding to weaknesses is automatically increased, which is conducive to improving the pass rate of residents in key operation assessments.

[0011] In a preferred embodiment of the present application, the method further includes: Construct a teaching and learning interactive and mutual evaluation system based on multi-dimensional evaluation indicators, including: By weighted calculation of phased competency assessment indicators, a phased competency assessment matrix for residents is generated. The phased competency assessment indicators include WPBA behavior records, 360-degree evaluation feedback, and participation in teaching case discussions. Set dynamic association rules between phased capability assessment results and rotation plans. When the dimension score of a phased capability assessment indicator is lower than the preset score threshold, generate and push targeted reinforcement task packages. Collect residents' learning behavior data in the medical video learning system (such as key frame dwell time and interactive answer accuracy), and combine it with the department's work logs of rotation scheduling to construct a multi-source evaluation data set; Using a Markov decision process model, the probability of rotation path deviation is predicted based on the multi-source evaluation data set to dynamically adjust the subsequent department rotation sequence and training focus; The learning behavior data is associated with the knowledge graph of the medical video learning system to determine the knowledge points that the resident doctors have not mastered, and the push of targeted learning tasks is triggered based on the unmastered knowledge points.

[0012] By adopting the above technical solution, the problem that the multi-dimensional evaluation system in the existing technology lacks quantitative association rules, resulting in the inability to accurately locate capability shortcomings, is solved. Learning behavior data includes key frame dwell time, interactive question answering accuracy, etc.; by weighted calculation of WPBA behavior records (such as a weight of 40%), 360-degree evaluation (such as a weight of 30%), and case discussion participation (such as a weight of 30%), a phased capability assessment matrix (scoring range 0-100) is generated to set a scoring threshold (such as a single item ≤ 60 points), triggering the reinforcement task package generation logic. Task types include targeted literature reading (10 articles / week) and simulated operation training (3 times / week), so as to verify the effectiveness of reinforcement and improve the accuracy of locating capability shortcomings by comparing the department assessment results before and after the task execution.

[0013] In a preferred embodiment of the present application, after generating an initial shift scheduling plan using a genetic algorithm and optimizing the shift scheduling result using a linear programming model, the method further includes: Obtaining initial attribute data of the resident physician and standardized training constraints, and determining a first load prediction value of the resident physician before entering the department and a second load prediction value before leaving the department based on the initial attribute data of the resident physician, the cross-department rotation rules, and the standardized training constraints; Generating a scheduling comparison rule base for dynamically calibrating a rotation plan according to the standardized training constraint condition, the first load forecast value, and the second load forecast value; Constructing a scheduling parameter benchmark set including a load adjustment threshold, and combining the scheduling parameter benchmark set with the scheduling comparison rule base to establish a dynamic regulation model for resident scheduling; The work status data of the resident physicians is obtained in real time, the work status data is input into the dynamic scheduling control model, and the department rotation sequence adjustment instruction is output.

[0014] By adopting the above technical solutions, through the double-level optimization of genetic algorithm and linear programming, the time for generating scheduling plans is shortened; the dynamic control model based on the load sensitivity coefficient is used to reduce the daily average load fluctuation rate of the department; and the real-time input model of work status data is used to reduce the response delay of scheduling adjustments.

[0015] In a preferred example of the present application, the determination of the first load prediction value of the resident physician before admission and the second load prediction value before discharge specifically includes: Analyze the initial attribute data of residents, including qualification level, historical rotation trajectory, professional adaptation label, and available time window. Combined with the mandatory department list, weight coefficient, and maximum period limit in the cross-department rotation rules, calculate the theoretical workload benchmark before admission. According to the minimum training duration of a single department, the assessment index threshold and the professional adaptation label in the standardized training constraints, combined with the historical rotation trajectory characteristics, the actual load deviation value before leaving the department is predicted.

[0016] By adopting the above technical solutions, the accuracy of load prediction before entering the department is improved through matching analysis of historical rotation trajectory characteristics and professional adaptation labels; combined with the minimum training time constraint of a single department, the proportion of invalid scheduling plans (rejected due to excessive load) is reduced; the dynamic matching of professional adaptation labels with historical data improves the rotation adaptability score of residents.

[0017] In a preferred example of the present application, the generation of a scheduling comparison rule base for dynamically calibrating the rotation plan specifically includes: Based on the department's maximum cycle limit and the upper limit of daily working hours, a theoretical scheduling time matrix is ​​constructed; By analyzing the matching degree between historical rotation trajectory characteristics and professional adaptation labels, a load characteristic similarity model is established; Calculate the dynamic ratio of the theoretical load benchmark before admission to the department and the actual load deviation value before leaving the department to generate the load sensitivity coefficient; Combining the theoretical scheduling time matrix with the load sensitivity coefficient, a scheduling comparison rule library including time compensation rules and department replacement rules is constructed; Alternatively, the establishment of a dynamic regulation model for resident physician scheduling may specifically include: The load adjustment threshold in the scheduling parameter benchmark set is coupled with the load sensitivity coefficient to generate the scheduling flexibility interval; A Markov chain model is used to predict the rule triggering probability of the scheduling control rule base and establish a rule priority sorting mechanism; According to the scheduling flexibility interval and rule priority sorting mechanism, a scheduling dynamic control model including a multi-objective optimization function is constructed.

[0018] By adopting the above technical solution, the rule triggering probability is predicted through the Markov chain model (accuracy ≥ 85%), which is conducive to shortening and dynamically controlling the rule base update cycle in real time; the scheduling flexibility range can cover at least 90% of sudden load scenarios; through the optimization of time compensation rules, it is conducive to shortening the overtime hours of residents and saving labor costs.

[0019] In the second aspect, the invention objective of this application is achieved by adopting the following technical solutions: An intelligent scheduling system based on standardized resident physician training, comprising: Data collection and analysis module, used to obtain residents' training progress, each department's receptive capacity and standardized training requirements; A scheduling plan generation module, configured to generate an initial scheduling plan using a genetic algorithm based on the information output by the data acquisition and analysis module; A scheduling optimization module is used to optimize the initial scheduling plan through a linear programming model to achieve dynamic rotation scheduling; A rotation rule matching module is used to match the initial scheduling plan with the cross-department rotation rules and the preset rotation template, generate a rotation department matching table, and associate the matching table with the training progress of the resident physician; The authority management module is used to set hierarchical authority control rules for user roles and data security management mechanisms; The training and assessment linkage module is used to associate the rotation scheduling results with the assessment plan of the online examination system according to the rotation department matching table and authority control rules, and dynamically adjust the content and form of the standardized training examination according to the rotation stage of the resident physician.

[0020] In a third aspect, the invention objective of this application is achieved by adopting the following technical solutions: A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned intelligent scheduling method based on standardized training of resident physicians.

[0021] Fourthly, the invention objectives of this application are achieved by adopting the following technical solutions: A computer program product includes a computer program / instruction, which, when executed by a processor, implements the steps of the intelligent scheduling method based on standardized training of resident physicians as described above.

[0022] In summary, this application includes at least one of the following beneficial technical effects: 1. Balanced rotations within departments and flexible scheduling that can adapt to changes in the national syllabus (Standardized Resident Training Syllabus) facilitate optimal resource allocation. By setting up permission control rules and a data security management system (such as transaction processing to prevent scheduling interruptions) for user accounts with different roles, a dynamic, compliant, and personalized intelligent scheduling method is provided. 2. Using a Markov chain model to predict rule trigger probabilities (with an accuracy rate ≥ 85%) helps shorten and dynamically control the rule base update cycle in real time; the flexible scheduling range can cover at least 90% of sudden load scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a flow chart of an intelligent scheduling method based on standardized resident physician training in one embodiment of the present application; Figure 2 This is a flowchart after step S1 in the intelligent scheduling method based on standardized training of resident physicians in one embodiment of the present application. DETAILED DESCRIPTION

[0024] The present application is further described in detail below with reference to the accompanying drawings.

[0025] In one embodiment, if Figure 1 As shown, the present application discloses an intelligent scheduling method based on standardized resident training, which specifically includes the following steps: S1: Based on the training progress of residents, the department's acceptance capacity and standardized training requirements, a genetic algorithm is used to generate the initial scheduling plan, and the scheduling results are optimized through a linear programming model to achieve dynamic rotation scheduling.

[0026] In this embodiment, the training progress includes the rotation time of the resident physician in each department and the remaining training time (obtained from the historical scheduling database); the department's acceptance capacity includes the maximum number of physicians that each department can accept per day and the current number of vacant beds (collected in real time from the HIS system); standardized training requirements include the upper limit of daily working hours (such as 8 hours), the minimum training time for a single department (such as ≥4 weeks for cardiovascular medicine), and the assessment indicator threshold (such as case discussion participation rate ≥90%).

[0027] Furthermore, in step S1, the genetic algorithm is used to generate an initial scheduling plan, and the scheduling result is optimized through a linear programming model, specifically including: S11: Receive resident training progress information, department admission capacity, and historical scheduling data, and construct scheduling constraints.

[0028] Specifically, the department's admission capacity refers to the maximum admission capacity of each department; training progress information refers to the rotation time and remaining training time of residents in each department (for example, the internal medicine department has rotated for 2 weeks and has 4 weeks remaining); the department's admission capacity refers to the maximum number of physicians that each department can accept per day (such as "the cardiology department can accept a maximum of 3 people per day") and the current number of vacant beds (such as real-time data "5 beds remaining in the cardiology department"); historical scheduling data: scheduling records of each department in the past 12 months (such as "2 physicians in the gastroenterology department took leave last week"); hard rules that must be met in scheduling constraints (such as "daily working hours ≤ 12 hours" and "mandatory department rotation time ≥ 4 weeks").

[0029] Specifically, the constraints are modeled as follows: Department load constraints are as follows: ∑(number of physicians_department i × number of rotation days) ≤ department capacity i × number of cycle days. (For example, the cardiology department can accept a maximum of 3 patients x 7 days = 21 person-days this week, but has actually scheduled 2 patients x 3 days = 6 person-days, leaving 15 person-days of capacity.)

[0030] Working time constraints, such as: daily working hours of doctors ∈ [6, 12] hours (the total working time is calculated by the daily shift period in the schedule, such as "08:00-17:00" is 9 hours).

[0031] Rotation integrity constraints such as: coverage rate of required departments = (number of completed required departments / total number of required departments) × 100% ≥ 100%.

[0032] S12: Genetic algorithm is used to generate multiple candidate scheduling schemes, and the initial scheduling scheme that satisfies the load balance of departments is selected through crossover and mutation operations.

[0033] Specifically, the chromosome encoding of the genetic algorithm refers to converting the scheduling plan into a gene sequence (for example, "gene [3, 1, 2]" means that the physician rotates through departments 3→1→2 in sequence); the fitness function refers to the mathematical formula for evaluating the quality of the scheduling plan (for example, Fitness = department load balance × 0.6 - overtime hours × 0.4); the crossover operation refers to exchanging some gene fragments of the chromosome to generate a new plan (such as PMX crossover); and the mutation operation refers to randomly adjusting the order of chromosome genes (for example, exchanging the positions of two departments).

[0034] Specifically, 100 chromosomes are randomly generated (e.g., Python code: np.random.permutation(5) generates a random arrangement of 5 departments), where the length of each chromosome is equal to the total number of departments (e.g., cardiology, surgery, and emergency departments, a total of 3 departments); fitness is calculated based on department load balance and overtime penalty, where the calculation formula for department load balance is: 1-(max(department load)-min(department load)) / avg(department load), for example: department A has a load of 80, and department B has a load of 7 0, C load 75 → balance degree = 1-(80-70) / 75≈0.867; the calculation formula for overtime penalty is: overtime = ∑(physician's daily working hours - 8 hours) × overtime rate, where the rate can be set as: 1.5 times for 8-10 hours, and 2 times for 10-12 hours; use the roulette wheel selection method: chromosomes with high fitness have a higher probability of being selected (for example, the chromosome with the highest fitness has a 30% selection probability), and set the fitness termination condition, such as no fitness improvement for 50 consecutive generations or reaching the maximum number of iterations (for example, 1000 generations).

[0035] S13: Use a linear programming model to perform secondary optimization on the initial scheduling plan to minimize the overtime hours of residents and output a scheduling result that meets the requirements of standardized training.

[0036] Specifically, the objective function refers to the mathematical expression that needs to be minimized in linear programming (such as minΣ(overtime hours of physician i)), slack variables are set to serve as auxiliary variables for handling inequality constraints (such as converting the "≤" constraint into an equality constraint), and the basis variables are the principal element column variables in the simplex method of linear programming.

[0037] S2: During the scheduling process, based on the initial scheduling plan, combined with the cross-department rotation rules and the preset rotation template, a rotation department matching table is generated, and the rotation department matching table is associated with the training progress of the resident physician.

[0038] In this embodiment, the cross-department rotation rules include a list of mandatory departments (such as internal medicine, surgery, and emergency department), optional department weights (such as a weight of 0.3 for the radiology department and a weight of 0.2 for the anesthesiology department), and the longest department rotation period (if the same department is ≤ 8 weeks). The preset rotation templates include a standard path template and a flexible adjustment window. The standard path template is such as internal medicine (4 weeks) → surgery (6 weeks) → emergency department (2 weeks); the flexible adjustment window allows fine-tuning of the department order within a range of ±2 weeks.

[0039] Initial plan verification refers to comparing the scheduling plan generated by the genetic algorithm with the rules one by one. If the mandatory department constraint is violated (such as the emergency department is not included), the missing department will be inserted and the subsequent department order will be adjusted; if the department rotation duration is insufficient (such as only 3 weeks for surgery), it will be extended to the minimum requirement and the subsequent optional department duration will be compressed; the adjusted scheduling plan will be mapped to the preset template to generate a standardized rotation path; and the adjustment records that deviate from the template will be recorded (such as the emergency department rotation duration is increased to 4 weeks), and the start / end dates of the physician in each department will be marked in the rotation department matching table. When the scheduling plan changes, the time node in the matching table will be updated synchronously, and the downstream system notification will be triggered (such as sending a scheduling change email to the physician).

[0040] Furthermore, the intelligent scheduling method based on standardized resident training also includes: S10: Dynamically generate a phased competency assessment report based on the residents’ mandatory viewing progress and correct answer rate in the medical video learning system.

[0041] In this embodiment, the mandatory viewing progress refers to the course viewing progress that residents must complete in the medical video learning system (such as "100% completion rate of cardiovascular emergency courses"); the correct answer rate refers to the accuracy rate of the physician's answers to in-class tests or chapter tests in the video courses (such as "92% correct answer rate for cardiopulmonary resuscitation operation questions"); the periodic ability assessment report refers to the physician's ability assessment file generated based on learning data, which includes scores on dimensions such as theoretical mastery and skill proficiency.

[0042] Specifically, the theoretical mastery scoring formula is: Theoretical score = (forced viewing progress × 0.6) + (answer accuracy × 0.4), for example: progress 100%, accuracy 80% → score 0.6 × 1 + 0.4 × 0.8 = 0.92.

[0043] Skill proficiency scores are: Scoring is based on the degree of completion of key operational steps in the video (e.g., the “chest compression depth compliance rate” in a cardiopulmonary resuscitation video accounts for 50% of the weight).

[0044] Comprehensive score = theoretical score × 0.7 + skill proficiency × 0.3; the scores for each dimension include theory / skill / total score; pre-define and set the matching degree with the assessment indicators of the rotation department (such as "theoretical score ≥ 0.8 is required for surgical rotation, and the current standard is 0.92") and risk warning mechanism (such as "skill score < 0.6 requires strengthened practical training").

[0045] S20: Match the resident physician's competency assessment results with the assessment indicators of the current rotation department, and automatically adjust the weights of teaching case discussions and practical tasks in subsequent rotation plans.

[0046] In this embodiment, teaching case discussion refers to a case analysis meeting in which residents participate (e.g., a cardiovascular case discussion held once a week); practical task weight refers to the time allocation ratio of different tasks in the rotation plan (e.g., "theoretical training: practical training = 3:7"); and dynamic adjustment rules are algorithmic logic that automatically modifies task weights based on competency assessment results.

[0047] Specifically, obtaining the current department requirements involves extracting the current department's assessment indicators from the rotation template (e.g., "Neurology Department members must complete ≥5 cerebrovascular case discussions"); setting thresholds, such as defining the mapping relationship between ability scores and task weights (e.g., "When the theoretical score is <0.7, the practical task weight is increased by 10%"). The comprehensive score is then compared with the department's assessment indicators one by one (e.g., "Case discussion participation score must be ≥0.8, currently 0.65"), marking assessment dimensions that do not meet the requirements (e.g., "Skill operation score is 0.58, below the threshold of 0.6"), and setting rules and triggering conditions for triggering task weight adjustments. For example, if the skill score is <0.6: the practical task weight is increased by 0.15, and the theoretical task weight is decreased by 0.10; if the case discussion participation score is <0.7: the discussion task weight is increased by 0.20. After adjustment, the total weight must still meet 100%.

[0048] S3: Set up permission control rules and data security management mechanisms for hierarchical management of user roles.

[0049] In this embodiment, user roles (or user accounts) for intelligent scheduling of standardized resident physician training are managed in a hierarchical manner. Role definitions include: system administrator: has all permissions (scheduling rule modification, data export), department head: can view the department's scheduling details and submit scheduling suggestions, resident physician: can only view personal scheduling information, and permission allocation management is implemented through the RBAC (role-based access control) model, that is, the permission granularity is refined to the field level (for example, only the "rotation department" field is allowed to be viewed). The TLS 1.3 protocol is used to encrypt the transmission of doctor-patient data, and sensitive information (such as the physician's ID number) is stored using AES-256 encryption.

[0050] S4: Based on the rotation department matching table and authority control rules, the rotation scheduling results are linked to the assessment plan of the online examination system, and the content and format of the standardized training examination are dynamically adjusted according to the rotation stage of the resident.

[0051] In this embodiment, a dynamic association management mechanism is set up for the scheduling and examination linkage mechanism. The rotation stage data of the scheduling system is pushed to the online examination system in real time through the RESTful API (data interface), and the examination stages are automatically divided according to the rotation department matching table (such as the internal medicine rotation period → internal medicine theory examination). The dynamic adjustment of the examination content includes the examination plan and format adaptation. The examination plan includes basic assessment and dynamic assessment. Basic assessment refers to fixed content (such as medical ethics), and dynamic assessment refers to customization according to the rotation department (such as adding "arrhythmia case analysis" after the cardiovascular medicine rotation). The format adaptation is also carried out in stages: the low-level stage (such as the early stage of the rotation) is mainly multiple-choice questions; the high-level stage (such as the late stage of the rotation) adds diversified question types such as simulated diagnosis and treatment, operation video scoring, etc.

[0052] Specifically, the triggering conditions for the automated triggering process are as follows: Condition 1: When a physician enters a new rotation department, the assessment module associated with that department will be automatically unlocked; Condition 2: When a physician's practice time in a certain department meets the requirements, the department's final examination will be triggered.

[0053] The test results are fed back to the scheduling system in real time, affecting the subsequent scheduling weights (for example, if a physician fails the operating room assessment, the subsequent operating room rotation time will be increased by 2 weeks).

[0054] Furthermore, a teaching and learning interactive evaluation system based on multi-dimensional evaluation indicators is constructed, including: S100: Calculate the phased competence evaluation indicators through weighted calculation and generate the phased competence evaluation matrix for residents. The phased competence evaluation indicators include WPBA behavior records, 360-degree evaluation feedback, and participation in teaching case discussions.

[0055] In this example, WPBA behavior records refer to workplace-based assessments, which are based on direct observation of residents' clinical performance (such as history taking and physical examinations). 360-degree feedback is a comprehensive evaluation from senior physicians, nurses, patients, and other parties (such as communication skills and teamwork). Participation in teaching case discussions refers to the frequency of residents' speaking in case discussions and their contribution to problem solving.

[0056] Specifically, WPBA data captures residents' ward rounds and operation records through clinical management systems (such as electronic medical record systems); 360-degree feedback is obtained by aggregating online evaluation form data; case discussion data includes records of meeting sign-ins and speech content keywords (such as "actively proposing differential diagnosis plans" extracted through NLP); each indicator of the stage-by-stage ability assessment index is assigned a corresponding weight coefficient, such as WPBA behavior records (40%) directly reflect clinical ability, 360-degree evaluation (30%) reflects soft skills; case discussion participation (30%) reflects learning initiative; the ability score = (WPBA score × 0.4) + (360-degree feedback score × 0.3) + (discussion participation score × 0.3).

[0057] S200: Set dynamic association rules between the stage-by-stage capability assessment results and the rotation plan. When the dimension score of a stage-by-stage capability assessment indicator is lower than the preset score threshold, generate a targeted reinforcement task package and push it in a targeted manner.

[0058] In this embodiment, dynamic association rules refer to the mapping relationship between trigger conditions based on evaluation results and subsequent tasks (such as "skill score <70 points → automatically push 10 simulated operation training cases"); reinforcement task packages are learning or training tasks customized to address ability shortcomings (such as "Cardiopulmonary Resuscitation Intensive Training Package: Contains 5 VR simulation cases + 2 instructor guidance").

[0059] Specifically, the targeted reinforcement task package includes the following: Knowledge-based: videos on knowledge points (e.g., “Shock Diagnosis and Treatment Process”); Skills: simulated operation tasks (such as "tracheal intubation VR training"); Practical: Participation in case discussion tasks (such as "Giving a lecture on an acute myocardial infarction case next week").

[0060] S300: Collect the learning behavior data of residents in the medical video learning system (such as the duration of key frame stay and the accuracy rate of interactive answering), combine it with the department work log of the rotation scheduling, and construct a multi-source evaluation data set.

[0061] In this embodiment, the key frame dwell time refers to the length of time the resident physician watches key operation segments (such as cardiopulmonary resuscitation compressions) in the medical video; the department work log refers to the resident physician's actual work record in the department (such as the type of surgery participated in and the length of time on duty).

[0062] Specifically, the video viewing data is converted into behavioral features (such as "attention to key operations = key frame dwell time / total viewing time"); the department log data is quantified into workload indicators (such as "average number of operating tables participated in per day") and stored in a time series database.

[0063] S400: Using a Markov decision process model, the probability of rotation path deviation is predicted based on a multi-source evaluation data set to dynamically adjust the subsequent department rotation sequence and training focus.

[0064] In this embodiment, the Markov decision process (MDP) refers to a mathematical model that predicts future behavior through state transition probability and is applicable to dynamic decision-making scenarios; the rotation path deviation probability refers to the probability that the actual rotation path of the resident deviates from the ideal template.

[0065] Specifically, in the Markov decision process model, the current state includes features such as ability score, task completion, and department preference; the action space refers to possible rotation adjustment options, such as "next department to choose neurosurgery / endocrinology / continue cardiology", and the Q-learning algorithm is used to solve the optimal action sequence (such as "next department to choose neurosurgery, and subsequent rotation order to adjust"); an adjustment recommendation report is generated, including the recommended department and the reason for the adjustment (such as "based on video learning data, the shortcomings of neurosurgery operation capabilities need to be strengthened").

[0066] S500: Associating learning behavior data with the knowledge graph of the medical video learning system, determining the knowledge points that the residents have not mastered, and triggering the push of targeted learning tasks based on the unmastered knowledge points.

[0067] In this embodiment, the knowledge graph refers to a structured medical knowledge network consisting of concept nodes (e.g., "shock") and relationship edges (e.g., "shock → treatment → fluid resuscitation"). Unmastered knowledge points refer to weak links identified through learning behavior data analysis (e.g., "missing three key steps in cardiopulmonary resuscitation"). Unmastered knowledge points are identified based on learning behavior analysis. For example, segments where the duration of a key frame in a video is less than the average are marked as "unmastered areas"; knowledge points associated with incorrectly answered questions are marked as "weak points."

[0068] Specifically, the knowledge graph contains knowledge point labels for medical video courses (such as "Cardiopulmonary Resuscitation → Step 1: Confirm environmental safety") and test point labels for online question banks (such as "Shock → Diagnostic Criteria for Hypovolemic Shock"), and uses the Neo4j graph database to store node relationships.

[0069] In one embodiment, if Figure 2 As shown, after step S1, the intelligent scheduling method based on standardized resident training further includes: S011: Obtain the initial attribute data of the resident physician and the standardized training constraints, and determine the first load prediction value of the resident physician before entering the department and the second load prediction value before leaving the department based on the initial attribute data of the resident physician, the cross-department rotation rules and the standardized training constraints.

[0070] In this embodiment, the load forecast before admission to the department is the first load forecast value; the load forecast before discharge from the department is the second load forecast value; the initial attribute data refers to the personal characteristic data of the resident physician, including qualification level (such as "Master's degree + 3 years of experience"), historical rotation record (such as "rotated in cardiology 3 times"), professional adaptation label (such as "excelled in surgical operations"), and available time period (such as "only available for scheduling during daytime on weekdays"); cross-department rotation rules refer to mandatory department rotation constraints, such as "mandatory department list (internal medicine, surgery, emergency department)" "optional department weight (radiology department weight 0.3)" and "longest department rotation period (same department ≤ 8 weeks)"; standardized training constraints are based on the standardized training requirements for resident physicians corresponding to the national syllabus, such as "daily working hours ≤ 12 hours" and "minimum training time for a single department (such as neurosurgery ≥ 6 weeks)".

[0071] Specifically, we obtained physician qualification levels and historical rotation records from the human resources system, encoded the qualification levels into numerical values ​​(e.g., "PhD + 5 years of experience" = 5 points, "Bachelor's degree + 1 year of experience" = 2 points). We also converted historical rotation records into department preference vectors (e.g., [cardiology: 0.8, surgery: 0.6, emergency department: 0.4]). The load forecasting model before admission was a linear regression model based on historical data, as follows: Predicted load = β0 + β1 × qualification level + β2 × historical rotation complexity + ε; where β0 is the intercept term, β1 and β2 are regression coefficients, and ε is the random error term; the load prediction model is based on qualification level (continuous value), historical rotation complexity (such as the number of rotation departments × average department load), the number of required departments in the cross-department rotation rules, and outputs the theoretical load benchmark before entering the department (such as "expected working hours per week are 8.5 hours").

[0072] The load prediction model before leaving the department selects an LSTM model based on time series. The input sequence includes: daily working hours in historical rotations, department load fluctuation index (such as the standard deviation of emergency department load) and training intensity near the end of the department (such as assessment intensity). The output data is the actual load deviation value before leaving the department (such as "expected to exceed the minimum training time of a single department by 20%").

[0073] Specifically, in step S011, determining the first load prediction value of the resident physician before admission to the department and the second load prediction value before discharge from the department specifically includes: S0111: Analyze the qualification level, historical rotation trajectory, professional adaptation label and available time window in the initial attribute data of the resident physician, and calculate the theoretical load benchmark before entering the department by combining the required department list, weight coefficient and maximum period limit in the cross-department rotation rules.

[0074] In this embodiment, the qualification level refers to the combination of the resident's educational background and work experience; the historical rotation trajectory refers to the work performance data of the past rotation departments; the professional adaptation label refers to the skill tendency label generated based on historical data; and the maximum cycle limit refers to the maximum time of continuous rotation in the same department.

[0075] Specifically, the academic qualifications and years of work experience are converted into numerical scores to perform field mapping of the qualification level, calculate the historical average working hours of departments and the coverage rate of required departments, convert the available time period into the number of available working days, and construct a theoretical compliance benchmark formula, such as theoretical load benchmark = (qualification level score × 0.3) + (historical average working hours × 0.4) + (number of required departments × 0.2) - (available time window × 0.1).

[0076] S0112: Based on the minimum training duration of a single department, assessment indicator thresholds, and professional adaptation labels in the standardized training constraints, combined with historical rotation trajectory characteristics, the actual load deviation value before leaving the department is predicted.

[0077] In this embodiment, the minimum training duration for a single department refers to the minimum number of rotation days for a single department required by policy; the assessment indicator threshold refers to the ability standard that must be achieved; and the historical trajectory characteristics refer to the load fluctuation pattern in past rotations.

[0078] Specifically, a random forest regression model is used based on the load fluctuation amplitude in historical rotations (such as standard deviation), the gap between the remaining rotation days and the minimum training time, and the matching degree of professional adaptation labels (such as "emergency department adaptation degree 85% → load risk + 10%"), and the output load deviation value is (minimum training time - rotation time) × risk coefficient + (1-assessment compliance rate) × penalty coefficient + historical volatility × 0.2.

[0079] S012: Generate a scheduling comparison rule library for dynamically calibrating the rotation plan based on the standardized training constraint conditions, the first load forecast value, and the second load forecast value.

[0080] In this embodiment, the load sensitivity coefficient refers to a sensitivity indicator that measures the difference between the predicted load before entering the department and the actual load before leaving the department (for example, a coefficient greater than 1 indicates a high risk of load fluctuation); the time compensation rule refers to a rule that automatically extends the working hours when the load exceeds the limit; and the department replacement rule refers to a rule that automatically adjusts the department rotation sequence when the load is insufficient.

[0081] Specifically, the theoretical scheduling time matrix is: theoretical scheduling time = ∑ (daily working hours) × department rotation cycle; the constraint condition is: if the theoretical scheduling time is greater than the upper limit of daily working hours, it will be reduced proportionally (for example, total working hours from 120 hours to 112 hours); a load characteristic similarity model is constructed, based on the load fluctuation characteristics in historical rotation trajectories (such as "cardiology department load variance 0.3"), and the similarity calculation formula is: similarity = 1-(current load variance - historical average variance) / historical average variance; among them, department combinations with high similarity (>0.8) are matched first, and low-similarity combinations are marked as high-risk (such as the "emergency department + pediatrics" combination).

[0082] Specifically, step S012 includes: S0121: Construct a theoretical scheduling matrix based on the department's maximum cycle limit and the upper limit of daily working hours.

[0083] In this embodiment, the maximum cycle limit of a department refers to the maximum time of continuous rotation of the same department; the theoretical scheduling time matrix is ​​a two-dimensional table with departments as rows and time as columns, which records the planned rotation time of each department in different time periods.

[0084] S0122: Establish a load feature similarity model by analyzing the matching degree between historical rotation trajectory characteristics and professional adaptation labels.

[0085] In this embodiment, the historical rotation trajectory feature refers to the load distribution feature of the resident physician's past rotation departments, such as "cardiology department load variance 0.3"; the professional adaptation label is a skill tendency label generated based on historical data; and the load feature similarity is used to measure the similarity of the load features of different physicians or departments.

[0086] S0123: Calculate the dynamic ratio of the theoretical load benchmark before entering the department and the actual load deviation value before leaving the department to generate the load sensitivity coefficient.

[0087] In this embodiment, the theoretical load benchmark before entering the department is the load value before entering the department predicted based on qualifications and historical data (such as "expected working hours per week are 8.5 hours"); the actual load deviation value before leaving the department refers to the difference between the actual load and the minimum training requirements (such as "the actual load exceeds the minimum requirement by 20%"); the load sensitivity coefficient is used to measure the sensitivity index of the load prediction deviation risk (such as a coefficient > 1 indicates high risk), and the risk level is divided based on the load sensitivity coefficient: low risk: coefficient ≤ 10%, no adjustment required; medium risk: 10% < coefficient ≤ 30%, triggering a warning; high risk: coefficient > 30%, forced adjustment, and the sensitivity coefficient is bound to the scheduling rules.

[0088] Load sensitivity coefficient = (load deviation value before leaving the department / theoretical load benchmark before entering the department) × 100%; S0124: Combine the theoretical scheduling time matrix and the load sensitivity coefficient to construct a scheduling comparison rule library that includes time compensation rules and department replacement rules.

[0089] In this embodiment, the time compensation rule refers to the rule for automatically extending working hours when the load exceeds the limit (such as "overtime of 1 hour → compensation of 0.5 hours the next day"); the department replacement rule refers to the rule for automatically adjusting the department rotation sequence when the load is insufficient (such as "the load of the cardiology department does not meet the standard → replaced by the surgery rotation"); the scheduling comparison rule library contains a database of multiple rules and their triggering conditions.

[0090] Specifically, the rule types include time compensation rules and department replacement rules. The triggering condition for the time compensation rule is a load deviation of >15%, and the compensation strategy is: the next day's working hours + 0.5 hours (upper limit 2 hours / week); the triggering condition for the department replacement rule is: load deviation >20%; the replacement strategy is to select a department with the same weight (such as replacing the emergency department with the surgery department).

[0091] The priority of the rules in the scheduling comparison rule library is high-risk first: such as "forced department replacement when load deviation is greater than 30%"; low-risk prompts: such as "generate an early warning notification when load deviation is 5%-10%".

[0092] S013: Construct a scheduling parameter benchmark set that includes load adjustment thresholds, combine the scheduling parameter benchmark set with the scheduling comparison rule base, and establish a dynamic regulation model for resident scheduling.

[0093] In this embodiment, the load regulation threshold is the allowable load fluctuation range (e.g., ±15%); the scheduling flexibility interval is the scheduling parameter space that can be dynamically adjusted within the threshold range (e.g., the daily working hours can be ±1 hour), and the scheduling flexibility interval = load regulation threshold × (1 + load sensitivity coefficient); the Markov chain model is a probability transfer model used to predict the probability of rule triggering.

[0094] Specifically, a scheduling comparison rule base is constructed, and the rule types include: Time compensation rules: 1 hour overtime → 0.5 hours of compensation the next day (up to 2 hours / week); Department replacement rule: Load deviation > 20% → Replace with a department of the same weight; Priority sorting rules: required departments > optional departments > flexibly adjusted departments.

[0095] The Markov chain model defines states as follows: State 1: Normal load (deviation ≤ 5%); State 2: Slight overload (5% < deviation ≤ 15%); and State 3: Severe overload (deviation > 15%). The frequency of transitions between states is calculated based on historical data (e.g., a 30% probability of transitioning from state 1 to state 2). Rules are then prioritized, with high-risk rules (e.g., severe overload) receiving the highest priority (weight 0.6) and low-risk rules (e.g., mild overload) receiving the lowest priority (weight 0.2).

[0096] Construct a multi-objective optimization function, such as minΣ(rule triggering cost)+λ×(load balance loss); where λ is the weight coefficient, balancing cost and balance; the constraints include adjustment constraints within the elastic range and the required department rotation time reaching the standard.

[0097] In this embodiment, a dynamic control model for resident physician scheduling is established, specifically including: S0131: Couple the load adjustment threshold in the scheduling parameter benchmark set with the load sensitivity coefficient to generate a scheduling elasticity interval.

[0098] In this embodiment, the load regulation threshold is the allowable load fluctuation range (e.g., ±15%), which is determined by standardized training requirements and labor laws and regulations. The load sensitivity coefficient is a sensitivity indicator used to reflect the risk of load forecast deviation. The scheduling flexibility range is an operational range (e.g., ±10% of the threshold) that is dynamically adjusted based on the load regulation threshold and the sensitivity coefficient.

[0099] Specifically, the scheduling flexibility range = load adjustment threshold × (1 ± load sensitivity coefficient × weight); weight adjustment includes: dynamic adjustment according to risk level (high-risk scenario weight = 1.2, low-risk = 0.8).

[0100] S0132: Use the Markov chain model to predict the rule triggering probability of the scheduling control rule base and establish a rule priority sorting mechanism.

[0101] In this embodiment, the Markov chain model is a probabilistic model that predicts future events through state transition probabilities and is applicable to dynamic decision-making scenarios. The rule trigger probability refers to the probability that a scheduling rule will be triggered in a specific scenario (for example, the probability of "triggering the department replacement rule when the load deviation is greater than 20%" is 75%). The rule priority sorting determines the execution order based on the trigger probability and rule importance, with high-risk rules taking precedence over low-risk rules.

[0102] Specifically, based on the rotation data of 2,000 physicians over the past year, the frequency of inter-state transitions was calculated; S0133: Based on the scheduling flexibility interval and rule priority sorting mechanism, a scheduling dynamic control model including multi-objective optimization function is constructed.

[0103] In this embodiment, the multi-objective optimization function is to optimize multiple conflicting objectives simultaneously, such as minimizing overtime hours and maximizing load balance; the constraints are the policies, regulations, and resource limits that the scheduling must meet.

[0104] For example, the constraints combine hard and soft constraints: Hard constraints include daily work hours ∈ [6, 12] hours and mandatory department rotation duration ≥ minimum requirements. Soft constraints include load balancing ≥ 0.8 (calculated by variance) and rule triggering ≤ 3 times per week.

[0105] Specifically, mixed integer programming and genetic algorithm were selected as the solution algorithms; mixed integer programming (MIP) is suitable for discrete decision variables (such as department selection); genetic algorithm is used to globally search for the optimal solution (crossover probability 0.8, mutation probability 0.1).

[0106] S014: Obtain the work status data of resident doctors in real time, input the work status data into the dynamic scheduling control model, and output department rotation sequence adjustment instructions.

[0107] In this embodiment, work status data refers to real-time collected physician work data, including current working hours and real-time department load data; department rotation sequence adjustment instructions are automatically generated scheduling change instructions. The input parameters of the scheduling dynamic control model include real-time working hours percentage, department load limit flag, and remaining scheduling flexibility range. The reasoning process includes: ① The Markov chain predicts the current state transition probability (e.g., the probability of state 1 → state 2 is 40%); ② The multi-objective optimization model generates candidate adjustment plans (e.g., “moving out of the emergency department can reduce load deviation by 25%”); ③ Select the optimal solution based on rule priority.

[0108] The instruction issuance operation includes: pushing to the scheduling system through the API interface (such as RESTful API call) and / or triggering SMS / email notifications to relevant doctors and department heads.

[0109] It should be understood that the serial numbers of the steps in the above embodiments do not imply the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0110] In one embodiment, an intelligent scheduling system based on standardized training of resident physicians is provided. The intelligent scheduling system based on standardized training of resident physicians corresponds to the intelligent scheduling method based on standardized training of resident physicians in the above embodiment.

[0111] The intelligent scheduling system based on standardized resident physician training includes a data collection and analysis module, a scheduling plan generation module, a scheduling optimization module, a rotation rule matching module, a rights management module, and a training and assessment linkage module. Detailed descriptions of each functional module are as follows: Data collection and analysis module, used to obtain residents' training progress, each department's receptive capacity and standardized training requirements; The scheduling plan generation module is used to generate the initial scheduling plan using genetic algorithm based on the information output by the data collection and analysis module; The shift scheduling optimization module is used to optimize the initial shift scheduling plan through a linear programming model to achieve dynamic rotation scheduling; The rotation rule matching module is used to match the initial scheduling plan with the cross-department rotation rules and the preset rotation template, generate a rotation department matching table, and associate the matching table with the training progress of the resident physician; The authority management module is used to set hierarchical authority control rules for user roles and data security management mechanisms; The training and assessment linkage module is used to link the rotation scheduling results with the assessment plan of the online examination system according to the rotation department matching table and authority control rules, and dynamically adjust the content and format of the standardized training examination according to the rotation stage of the resident physician.

[0112] For the specific limitations of the intelligent scheduling system based on standardized training of resident physicians, please refer to the limitations of the intelligent scheduling method based on standardized training of resident physicians in the above text, which will not be repeated here; each module in the above-mentioned intelligent scheduling system based on standardized training of resident physicians can be implemented in whole or in part through software, hardware and their combination; each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each of the above modules.

[0113] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented: S1: Based on the training progress of residents, the department's acceptance capacity and standardized training requirements, a genetic algorithm is used to generate the initial scheduling plan, and the scheduling results are optimized through a linear programming model to achieve dynamic rotation scheduling; S2: During the scheduling process, based on the initial scheduling plan, the cross-department rotation rules are matched with the preset rotation template to generate a rotation department matching table, which is then linked to the resident's training progress. S3: Set up permission control rules and data security management mechanisms for hierarchical management of user roles; S4: Based on the rotation department matching table and authority control rules, the rotation scheduling results are linked to the assessment plan of the online examination system, and the content and format of the standardized training examination are dynamically adjusted according to the rotation stage of the resident.

[0114] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described method embodiments. Any reference to memory, storage, database, or other media used in the various embodiments provided herein may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM).

[0115] In one embodiment, particularly according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product comprising a computer program / instructions that, when executed by a processor, implement the steps of the method described above. In such an embodiment, the computer program can be downloaded and installed from a network via a communication module and / or installed from removable media. When executed by a central processing unit (CPU), the computer program performs the various functions defined in the present invention.

[0116] Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0117] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments may still be modified, or some of the features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. An intelligent scheduling method based on standardized resident physician training, characterized by: include: Based on the training progress of residents, the department's acceptance capacity and standardized training requirements, a genetic algorithm is used to generate the initial scheduling plan, and the scheduling results are optimized through a linear programming model to achieve dynamic rotation scheduling; During the scheduling process, a rotation department matching table is generated based on the initial scheduling plan combined with the cross-department rotation rules and the preset rotation template, and the rotation department matching table is associated with the resident physician's training progress; Set up permission control rules and data security management mechanisms for hierarchical management of user roles; According to the rotation department matching table and the authority control rules, the rotation scheduling results are associated with the assessment plan of the online examination system, and the content and form of the standardized training examination are dynamically adjusted according to the rotation stage of the resident physician.

2. The intelligent scheduling method based on standardized resident physician training according to claim 1 is characterized in that: The genetic algorithm is used to generate the initial scheduling plan, and the linear programming model is used to optimize the scheduling results, specifically including: Receive resident training progress information, department admission capacity, and historical scheduling data, and establish scheduling constraints; Genetic algorithms are used to generate multiple candidate scheduling schemes, and the initial scheduling scheme that satisfies department load balance is selected through crossover and mutation operations; A linear programming model is used to perform secondary optimization on the initial scheduling plan to minimize the overtime hours of residents and output a scheduling result that meets the requirements of standardized training.

3. The intelligent scheduling method based on standardized resident physician training according to claim 1 is characterized in that: The method further comprises: Dynamically generate a phased competency assessment report based on residents' mandatory viewing progress and correct answer rate in the medical video learning system; Match the resident physician's competency assessment results with the assessment indicators of the current rotation department, and automatically adjust the weights of teaching case discussions and practical tasks in subsequent rotation plans.

4. The intelligent scheduling method based on standardized resident physician training according to claim 1 or 3, characterized in that: The method further comprises: Construct a teaching and learning interactive and mutual evaluation system based on multi-dimensional evaluation indicators, including: By weighted calculation of phased competency assessment indicators, a phased competency assessment matrix for residents is generated. The phased competency assessment indicators include WPBA behavior records, 360-degree evaluation feedback, and participation in teaching case discussions. Set dynamic association rules between phased capability assessment results and rotation plans. When the dimension score of a phased capability assessment indicator is lower than the preset score threshold, generate and push targeted reinforcement task packages. Collect residents' learning behavior data in the medical video learning system, combine it with the department's work logs of rotation scheduling, and construct a multi-source evaluation data set; Using a Markov decision process model, the probability of rotation path deviation is predicted based on the multi-source evaluation data set to dynamically adjust the subsequent department rotation sequence and training focus; The learning behavior data is associated with the knowledge graph of the medical video learning system to determine the knowledge points that the resident doctors have not mastered, and the push of targeted learning tasks is triggered based on the unmastered knowledge points.

5. The intelligent scheduling method based on standardized resident physician training according to claim 4 is characterized in that: After generating an initial shift scheduling plan using a genetic algorithm and optimizing the scheduling result using a linear programming model, the method further includes: Obtaining initial attribute data of the resident physician and standardized training constraints, and determining a first load prediction value of the resident physician before entering the department and a second load prediction value before leaving the department based on the initial attribute data of the resident physician, the cross-department rotation rules, and the standardized training constraints; Generating a scheduling comparison rule base for dynamically calibrating a rotation plan according to the standardized training constraint condition, the first load forecast value, and the second load forecast value; Constructing a scheduling parameter benchmark set including a load adjustment threshold, and combining the scheduling parameter benchmark set with the scheduling comparison rule base to establish a dynamic regulation model for resident scheduling; The work status data of the resident physicians is obtained in real time, the work status data is input into the dynamic scheduling control model, and the department rotation sequence adjustment instruction is output.

6. The intelligent scheduling method based on standardized resident physician training according to claim 5 is characterized in that: The determination of the first load prediction value of the resident physician before admission to the department and the second load prediction value before discharge from the department specifically includes: Analyze the initial attribute data of residents, including qualification level, historical rotation trajectory, professional adaptation label, and available time window. Combined with the mandatory department list, weight coefficient, and maximum period limit in the cross-department rotation rules, calculate the theoretical workload benchmark before admission. According to the minimum training duration of a single department, the assessment index threshold and the professional adaptation label in the standardized training constraints, combined with the historical rotation trajectory characteristics, the actual load deviation value before leaving the department is predicted.

7. The intelligent scheduling method based on standardized resident physician training according to claim 5 is characterized in that: The generation of a scheduling comparison rule base for dynamically calibrating the rotation plan specifically includes: Based on the department's maximum cycle limit and the upper limit of daily working hours, a theoretical scheduling time matrix is ​​constructed; By analyzing the matching degree between historical rotation trajectory characteristics and professional adaptation labels, a load characteristic similarity model is established; Calculate the dynamic ratio of the theoretical load benchmark before admission to the department and the actual load deviation value before leaving the department to generate the load sensitivity coefficient; Combining the theoretical scheduling time matrix with the load sensitivity coefficient, a scheduling comparison rule library including time compensation rules and department replacement rules is constructed; Alternatively, the establishment of a dynamic regulation model for resident physician scheduling may specifically include: The load adjustment threshold in the scheduling parameter benchmark set is coupled with the load sensitivity coefficient to generate the scheduling flexibility interval; A Markov chain model is used to predict the rule triggering probability of the scheduling control rule base and establish a rule priority sorting mechanism; According to the scheduling flexibility interval and rule priority sorting mechanism, a scheduling dynamic control model including a multi-objective optimization function is constructed.

8. The intelligent scheduling system based on standardized resident training is characterized by: The system comprises: Data collection and analysis module, used to obtain residents' training progress, each department's receptive capacity and standardized training requirements; A scheduling plan generation module, configured to generate an initial scheduling plan using a genetic algorithm based on the information output by the data acquisition and analysis module; A scheduling optimization module is used to optimize the initial scheduling plan through a linear programming model to achieve dynamic rotation scheduling; A rotation rule matching module is used to match the initial scheduling plan with the cross-department rotation rules and the preset rotation template, generate a rotation department matching table, and associate the matching table with the training progress of the resident physician; The authority management module is used to set hierarchical authority control rules for user roles and data security management mechanisms; The training and assessment linkage module is used to associate the rotation scheduling results with the assessment plan of the online examination system according to the rotation department matching table and authority control rules, and dynamically adjust the content and form of the standardized training examination according to the rotation stage of the resident physician.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the intelligent scheduling method based on standardized training of resident physicians as described in any one of claims 1 to 7 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the steps of the intelligent scheduling method based on standardized training of resident physicians as described in any one of claims 1 to 7 are implemented.

Images