Hospital physical examination guidance degree adjusting method and system based on multi-rule intelligent fusion

Abstract

The invention discloses a hospital physical examination guidance adjustment method and system based on multi-rule intelligent fusion, and the method comprises the steps: obtaining the to-be-examined item information of an examinee, the real-time queuing number and average service time of each department, and the current state of equipment, and obtaining structured system state data; determining a dynamic weight coefficient of each rule according to the passenger flow state of the current time period, and performing weighted summation calculation on the six-dimensional rule score vector to obtain a comprehensive priority score of each to-be-detected department; selecting the department with the highest comprehensive priority score as a recommended department, and generating a scheduling instruction; guiding information is pushed to the examinee through the multimedia terminal, and physical examination guiding and adjusting are completed. According to the invention, system fusion of medical constraint and efficiency optimization is realized, the waiting time of the patient is effectively reduced, and the physical examination process efficiency is improved.

Description

Technical Field

[0002] This invention relates to the field of smart healthcare and intelligent scheduling algorithm technology, and in particular to a hospital physical examination guidance and scheduling method and system based on multi-rule intelligent fusion. Background Technology

[0004] With increasing health awareness, regular physical examinations have become a common need for the public. However, in the traditional physical examination process, examinees often have to shuttle between various examination departments on their own, facing the blind choice of "going where there are fewer people," which leads to: 1) Inefficient route planning: Examinees make ineffective trips between different floors and areas, increasing physical exertion; 2) Excessive waiting time: Popular or time-consuming departments tend to attract large crowds, while other departments may be idle, resulting in uneven resource utilization; 3) Discomfort while waiting on an empty stomach: If items requiring fasting (such as blood tests and ultrasounds) are scheduled later, it will prolong the examinee's fasting waiting time and cause physical discomfort; 4) Process bottlenecks: Some items have prerequisites (such as an electrocardiogram before a gastroscopy), and if the order is incorrect, it will lead to interruption and backtracking.

[0005] Existing guidance systems mostly adopt simple "first-come, first-served" or "shortest queue" strategies, failing to comprehensively consider the inherent logic of the physical examination process and various optimization objectives from a global perspective, resulting in limited intelligence. Summary of the Invention

[0007] The purpose of this invention is to provide a hospital physical examination guidance and scheduling method and system based on multi-rule intelligent fusion, which comprehensively considers medical constraints (such as fasting requirements) and operational efficiency factors to achieve intelligent optimization scheduling of the physical examination process.

[0008] This application provides a hospital physical examination guidance and scheduling method based on multi-rule intelligent fusion, including the following steps:

[0009] The system acquires information on the items to be examined for the examinee, the real-time number of people queuing and the average service time in each department, and the current status of the equipment. It then categorizes and stores the information on the items to be examined, the real-time number of people queuing, the average service time, and the current status of the equipment to obtain structured system status data.

[0010] Based on the structured system status data, for each department to be examined, the following scores are calculated: fasting priority score, same area priority score, regional flow order score, empty queue priority score, pre-examination item score, and shortest time priority score, resulting in a six-dimensional rule score vector.

[0011] Based on the six-dimensional rule score vector, the dynamic weight coefficient of each rule is determined according to the current passenger flow status. The six-dimensional rule score vector is then weighted and summed to obtain the comprehensive priority score of each department to be inspected.

[0012] Based on the comprehensive priority score of each department to be inspected, the department with the highest comprehensive priority score is selected as the recommended department, and a scheduling instruction containing the identifier of the recommended department and the estimated waiting time is generated;

[0013] Based on the scheduling instructions, guidance information is pushed to the examinee through a multimedia terminal to complete the physical examination guidance and scheduling.

[0014] Optionally, based on the structured system status data, for each department to be examined, a six-dimensional rule score vector is calculated, including: fasting priority score, same-region priority score, regional flow order score, empty queue priority score, pre-examination item score, and shortest time priority score.

[0015] Based on the structured system status data, it is determined whether the examinee's examination items include any examination items that require fasting, thus obtaining a fasting requirement identifier;

[0016] Based on the fasting requirement identifier and the fasting requirement attributes of each department, a first preset score value is assigned to the departments to be examined that require fasting and whose departments require fasting, and zero points are assigned to other departments to be examined, thus obtaining the fasting priority score for each department to be examined;

[0017] The six-dimensional rule score vector is formed by combining the fasting priority score, the same region priority score, the regional streamline order score, the empty queue priority score, the pre-item inspection score, and the shortest time priority score.

[0018] Optionally, based on the structured system status data, for each department to be examined, a six-dimensional rule score vector is calculated, including: fasting priority score, same-region priority score, regional flow order score, empty queue priority score, pre-examination item score, and shortest time priority score.

[0019] Based on the structured system status data, the current location of the examinee and the location of each department to be examined are obtained to obtain regional distribution information;

[0020] Based on the regional distribution information, a second preset score value is assigned to the departments to be examined that have the same regional identifier as the examinee's current location, thus obtaining a priority score for each department within the same region;

[0021] Based on the regional distribution information and the preset regional streamline sequence, the order matching degree of each department to be examined in the regional streamline sequence is calculated to obtain the regional streamline sequence score of each department to be examined;

[0022] Based on the same region priority score and the region streamline order score, combined with other rule scores, the six-dimensional rule score vector is formed.

[0023] Optionally, based on the six-dimensional rule score vector, the dynamic weight coefficients of each rule are determined according to the current passenger flow status, and the six-dimensional rule score vector is weighted and summed to obtain the comprehensive priority score of each department to be inspected, including:

[0024] Based on the structured system status data, the total number of people queuing and the average number of people queuing in each department during the current time period are statistically analyzed. The ratio of the variance of the number of people queuing to the average number of people queuing is calculated to obtain the system load index.

[0025] Based on the system load index, when the system load index is greater than the first threshold, the empty queue priority weight coefficient and the shortest time priority weight coefficient are increased, while the same region priority weight coefficient and the region streamline order weight coefficient are decreased. When the system load index is less than the second threshold, the empty queue priority weight coefficient is decreased, while the same region priority weight coefficient and the region streamline order weight coefficient are increased, resulting in six adjusted dynamic weight coefficients.

[0026] Based on the adjusted six dynamic weight coefficients, each component of the six-dimensional rule score vector is multiplied by its corresponding dynamic weight coefficient and then summed to obtain the comprehensive priority score of each department to be examined.

[0027] Optionally, based on the structured system status data, for each department to be examined, a six-dimensional rule score vector is calculated, including: fasting priority score, same-region priority score, regional flow order score, empty queue priority score, pre-examination item score, and shortest time priority score.

[0028] Based on the structured system state data, the list of items to be tested for the examinee is extracted, and the list of items to be tested is traversed to identify the preceding dependent items of each item to be tested, thereby obtaining the preceding dependency relationship data;

[0029] Based on the aforementioned prerequisite dependency data, it is determined whether the prerequisite dependencies for each department to be inspected have been completed. Departments to be inspected that have completed all their prerequisite dependencies are assigned a third preset score, while departments to be inspected that have not completed their prerequisite dependencies are assigned zero points, thus obtaining the prerequisite inspection score for each department to be inspected.

[0030] Based on the scores from the preliminary item checks, combined with scores from other rules, the six-dimensional rule score vector is formed.

[0031] Optionally, the system acquires the examinee's pending examination information, the real-time queuing number and average service time of each department, and the current status of the equipment. It then categorizes and stores the pending examination information, the real-time queuing number, the average service time, and the current status of the equipment to obtain structured system status data. Based on the six-dimensional rule score vector, the system determines the dynamic weight coefficients of each rule according to the current passenger flow status, and performs a weighted summation of the six-dimensional rule score vector to obtain the comprehensive priority score for each department, including:

[0032] Based on the real-time queuing number, average service time, equipment utilization rate, doctor busyness, and historical throughput of each department, a department state matrix is ​​constructed to obtain a multi-dimensional department state representation;

[0033] Based on the multidimensional department state representation, a predefined random projection matrix is ​​generated, and the department state matrix is ​​subjected to dimensionality reduction projection processing to obtain a low-dimensional state sketch vector;

[0034] Based on the low-dimensional state sketch vector, the department load vector corresponding to the maximum singular value and the department idle vector corresponding to the minimum singular value are extracted through power iteration calculation to obtain the system extreme state characteristics;

[0035] Based on the system's extreme state characteristics, a preset number of departments with the highest scores in the departmental load vector are identified as the bottleneck department set, and a preset number of departments with the highest scores in the departmental idle vector are identified as the idle department set.

[0036] Based on the bottleneck department set and the idle department set, the system load ratio is calculated. The weight coefficients of each rule are dynamically adjusted according to the system load ratio. A penalty coefficient is applied to the comprehensive priority score of the departments to be inspected that are in the bottleneck department set, and a reward coefficient is applied to the comprehensive priority score of the departments to be inspected that are in the idle department set, so as to obtain the comprehensive priority score of each department to be inspected.

[0037] Optionally, based on the multidimensional department state representation, a predefined random projection matrix is ​​generated, and the department state matrix is ​​subjected to dimensionality reduction projection processing to obtain a low-dimensional state sketch vector, including:

[0038] Based on the number of rows in the department state matrix, calculate the base-2 logarithm of the row number and round it up to obtain the target dimension of the sketch;

[0039] A random matrix following a standard normal distribution is generated, wherein the number of rows of the random matrix is ​​equal to the dimension of the sketch target and the number of columns is equal to the number of rows of the department state matrix. Each element of the random matrix is ​​divided by the square root of the dimension of the sketch target and then normalized to obtain the random projection matrix.

[0040] Based on the random projection matrix and the department state matrix, the matrix product of the random projection matrix and the department state matrix is ​​calculated to obtain the low-dimensional state sketch vector;

[0041] Based on the low-dimensional state sketch vectors, multiple low-dimensional state sketch vectors within the most recent preset time window are stored using a sliding window method. The average value of these multiple low-dimensional state sketch vectors is calculated to obtain the incrementally updated average sketch vector.

[0042] Based on the average sketch vector, subsequent extreme value extraction processing is performed.

[0043] Optionally, based on the bottleneck department set and the idle department set, the system load ratio is calculated, and the weight coefficients of each rule are dynamically adjusted according to the system load ratio. A penalty coefficient is applied to the comprehensive priority score of the departments to be inspected that are in the bottleneck department set, and a reward coefficient is applied to the comprehensive priority score of the departments to be inspected that are in the idle department set, thereby obtaining the comprehensive priority score of each department to be inspected, including:

[0044] Based on the system's extreme state characteristics, the ratio of the maximum singular value of the department load vector to the average value of the department load vector is calculated to obtain the system load ratio;

[0045] Based on the system load ratio, when the system load ratio is greater than the first load threshold, the empty queue priority weight coefficient is multiplied by the first amplification factor, and the same region priority weight coefficient and the region streamline order weight coefficient are multiplied by the first reduction factor. When the system load ratio is less than the second load threshold, the region streamline order weight coefficient is multiplied by the second amplification factor, and the same region priority weight coefficient is multiplied by the third amplification factor to obtain the extreme value driven adjustment weight coefficient.

[0046] Based on the extreme value-driven adjustment weight coefficients, the six-dimensional rule score vector is weighted and summed to obtain a preliminary comprehensive priority score;

[0047] Based on the preliminary comprehensive priority score, the bottleneck department set, and the idle department set, the preliminary comprehensive priority score of the departments to be inspected in the bottleneck department set is multiplied by a penalty coefficient less than one, and the preliminary comprehensive priority score of the departments to be inspected in the idle department set is multiplied by a reward coefficient greater than one, to obtain the final comprehensive priority score of each department to be inspected.

[0048] Optionally, based on the six-dimensional rule score vector, the dynamic weight coefficient of each rule is determined according to the passenger flow status of the current time period. The six-dimensional rule score vector is then weighted and summed to obtain the comprehensive priority score of each department to be inspected. Based on the comprehensive priority scores of each department to be inspected, the department with the highest comprehensive priority score is selected as the recommended department, and a scheduling instruction containing the identifier of the recommended department and the estimated waiting time is generated, including:

[0049] Based on the examinee's list of items to be inspected, the region identifier of each item to be inspected is identified, the frequency of each region identifier is counted, and a preset number of regions with the highest frequency are selected as the recommended region sequence to obtain a region-level scheduling decision;

[0050] Based on the regional scheduling decision, for each region in the recommended regional sequence, a subset of the departments to be examined for the examinees within that region is extracted. The six-dimensional rule score vector is applied to each department in the subset of departments to be examined, and a weighted sum is calculated. A preset number of departments with the highest scores are selected to obtain the department-level scheduling decision.

[0051] Based on the department-level scheduling decision, for each department in the department-level scheduling decision, the real-time availability status of the department's time slots is obtained, the instant availability score, predicted waiting time score, and doctor efficiency score of each time slot are calculated, the instant availability score, the predicted waiting time score, and the doctor efficiency score are weighted and summed, and the time slot with the highest score is selected to obtain the time slot-level scheduling decision;

[0052] Based on the regional-level scheduling decision, the department-level scheduling decision, and the time slot-level scheduling decision, a three-layer hierarchical decision tree is constructed, comprising a regional layer, a department layer, and a time slot layer. The hierarchical decision tree is traversed to calculate the cumulative score of all paths from the root node to the leaf node. The path with the highest cumulative score is selected, and the department node and time slot node in that path are extracted to obtain the optimal scheduling path.

[0053] Based on the optimal scheduling path, a scheduling instruction is generated that includes the department identifier, time slot identifier, and expected waiting time in the optimal scheduling path.

[0054] Optionally, the system acquires the examinee's pending examination information, the real-time queuing number and average service time of each department, and the current status of the equipment. It then categorizes and stores the pending examination information, the real-time queuing number, the average service time, and the current status of the equipment to obtain structured system status data. Based on the comprehensive priority score of each department, it selects the department with the highest comprehensive priority score as the recommended department and generates a scheduling instruction containing the recommended department's identifier and the estimated waiting time, including:

[0055] Based on all examinees, departments, time slots, and devices in the current time period, a hypergraph vertex set is constructed, which includes the vertex sets of examinees, departments, time slots, and devices. Hyperedges connecting multiple vertices are added for fasting constraints, prerequisite dependency constraints, resource conflict constraints, and examinee grouping constraints, respectively, to obtain the physical examination scheduling hypergraph.

[0056] Based on the physical examination scheduling hypergraph, a hypergraph correlation matrix is ​​constructed, and the degree matrix and Laplacian matrix of the hypergraph correlation matrix are calculated. Feature vectors are extracted from the Laplacian matrix through eigenvalue decomposition, and the feature vectors are clustered to obtain a preset number of subject partitions.

[0057] Based on the preset number of subject partitions, the hyperedge cutting cost and partition size balance between each partition are calculated. The hyperedge cutting cost and partition size balance are optimized by iteratively moving vertices to obtain the optimized subject partitioning results.

[0058] Based on the optimized subject partitioning results, for each subject in each partition, a subset of departments within that partition is extracted. The six-dimensional rule score vector is applied to each department in the subset of departments to calculate the score. When the hyperedge involved in the department to be examined is cut off, a hyperedge penalty value is applied to the comprehensive priority score of the department. The department with the highest comprehensive priority score after deducting the penalty is selected to obtain the sub-scheduling scheme for each partition.

[0059] Based on the sub-scheduling schemes of each partition, the cut-off hyperedges are identified, cross-partition order constraints are forcibly added for cut-off hyperedges of the preceding dependency type, different time slots are allocated for cut-off hyperedges of the resource conflict type, the sub-scheduling schemes of each partition are merged, and the scheduling instruction is generated.

[0060] This application also provides a hospital physical examination guidance and scheduling system based on multi-rule intelligent fusion, including:

[0061] The data acquisition module is used to acquire information about the items to be examined by the examinee, the real-time number of people queuing in each department and the average service time, as well as the current status of the equipment. The module categorizes and stores the information about the items to be examined, the real-time number of people queuing, the average service time and the current status of the equipment to obtain structured system status data.

[0062] The rule calculation module is used to calculate, based on the structured system status data, the fasting priority score, the same area priority score, the area flow order score, the empty queue priority score, the pre-examination item score, and the shortest time priority score for each department to be examined, to obtain a six-dimensional rule score vector;

[0063] The weight adjustment module is used to determine the dynamic weight coefficient of each rule based on the six-dimensional rule score vector and the current passenger flow status, and to perform a weighted summation calculation on the six-dimensional rule score vector to obtain the comprehensive priority score of each department to be inspected;

[0064] The scheduling decision module is used to select the department with the highest comprehensive priority score as the recommended department based on the comprehensive priority score of each department to be inspected, and generate a scheduling instruction containing the identifier of the recommended department and the estimated waiting time;

[0065] The guidance and push module is used to push guidance information to the examinee through a multimedia terminal based on the scheduling instruction, thereby completing the physical examination guidance and scheduling. Attached Figure Description

[0067] Figure 1 This is a diagram of the overall system architecture of the present invention;

[0068] Figure 2 This is a flowchart of the hospital physical examination guidance and scheduling method based on multi-rule intelligent fusion according to the present invention. Detailed Implementation

[0070] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0071] Example 1

[0072] like Figure 1 As shown, the overall system architecture of this invention includes a data acquisition layer, a rule engine layer, a scheduling decision layer, a feedback optimization layer, and a guidance output layer. The data acquisition layer is responsible for collecting examinee information, department status, and equipment status; the rule engine layer includes fasting priority rules, same-region priority rules, regional order rules, empty queue priority rules, prerequisite item rules, and shortest time rules; the scheduling decision layer performs real-time priority calculation and dynamic scheduling decisions; and the guidance output layer pushes guidance information to examinees through multimedia terminals.

[0073] This application provides a hospital physical examination guidance and scheduling method based on multi-rule intelligent fusion, such as... Figure 2 As shown, the specific steps include:

[0074] Step S1: Obtain the examinee's examination items information, the real-time queuing number and average service time of each department, and the current status of the equipment. Classify and store the examination items information, the real-time queuing number, the average service time, and the current status of the equipment to obtain structured system status data.

[0075] Specifically, the system first obtains the examinee's personal information and list of examination items through the physical examination information system. At the same time, it obtains the current number of people queuing, average service time and equipment status of each department through the real-time monitoring system. This information is classified and stored according to the patient dimension, department dimension and time dimension to form structured system status data, which is convenient for subsequent rule calculation.

[0076] Step S2: Based on the structured system status data, calculate the fasting priority score, same area priority score, area flow order score, empty queue priority score, pre-examination item score, and shortest time priority score for each department to be examined, and obtain a six-dimensional rule score vector.

[0077] Specifically, based on structured system status data, the system calculates the following scores for each department of the examinee:

[0078] 1) Fasting priority score: Determine whether the examinee's examination items include examination items that require fasting. If the department requires fasting and the examinee needs to fast for the examination, then the department is given a higher fasting priority score.

[0079] 2) Priority score for the same region: Based on the current region of the examinee and the region of each department to be examined, if the department is in the same region as the examinee, a higher priority score for the same region will be given.

[0080] 3) Regional flow line sequence score: Calculate the sequence matching degree of each department to be examined in the flow line according to the preset optimal flow line path of the physical examination center;

[0081] 4) Empty queues are given priority in scoring: based on the real-time number of people queuing in each department, the shorter the queue, the higher the score;

[0082] 5) Pre-examination item score: The score examines the dependencies between the items to be examined. If the pre-examination dependent items of a certain department have been completed, a higher score will be given.

[0083] 6) Shortest time priority score: Based on the average service time of each department and the current queuing time, the estimated completion time is calculated. The shorter the time, the higher the score.

[0084] Based on the above calculations, a six-dimensional rule score vector [S_fasting, S_region, S_flow, S_queue, S_prerequisite, S_time] is generated for each department to be examined.

[0085] Step S3: Based on the six-dimensional rule score vector, determine the dynamic weight coefficient of each rule according to the current passenger flow status, and perform weighted summation calculation on the six-dimensional rule score vector to obtain the comprehensive priority score of each department to be inspected.

[0086] Specifically, the system first calculates the total and average number of people queuing for each department during the current time period, then calculates the ratio of the variance in the number of people queuing to the average number of people queuing, thus obtaining the system load index. Based on this index, the system dynamically adjusts the weight coefficients of the six rules:

[0087] When the system load index is greater than the first threshold (indicating high system load), increase the priority weight coefficient of empty queue and the priority weight coefficient of shortest time, and decrease the priority weight coefficient of same area and the priority weight coefficient of regional flow order, giving priority to system resource balance and efficiency.

[0088] When the system load index is less than the second threshold (indicating low system load), the priority weight coefficient for empty queues is reduced, while the priority weight coefficient for the same area and the priority weight coefficient for the flow order of the area are increased, giving priority to the test subject's experience and the rationality of the path.

[0089] The system then applies the adjusted weighting coefficients to the six-dimensional rule-based score vector to calculate the weighted sum score:

[0090] Score(i) = W1·S_fasting(i) + W2·S_region(i) + W3·S_flow(i) + W4·S_queue(i) + W5·S_prerequisite(i) + W6·S_time(i)

[0091] Where W1-W6 are the dynamically adjusted weight coefficients, and S_fasting(i)-S_time(i) are the scores of department i under each rule.

[0092] Step S4: Based on the comprehensive priority score of each department to be inspected, select the department with the highest comprehensive priority score as the recommended department, and generate a scheduling instruction containing the identifier of the recommended department and the estimated waiting time.

[0093] Specifically, the system sorts the departments to be inspected in descending order according to the comprehensive priority score calculated in step S3, and selects the department with the highest score as the recommended department. Then, the system calculates the estimated waiting time based on the current number of people in the queue and the average service time of the department, and generates a scheduling instruction that includes the recommended department identifier and the estimated waiting time.

[0094] Step S5: Based on the scheduling instruction, push guidance information to the examinee through a multimedia terminal to complete the physical examination guidance scheduling.

[0095] Specifically, the system converts the scheduling instructions generated in step S4 into easily understandable guidance information, which is then pushed to the examinee through multimedia terminals such as guidance displays, mobile applications, or voice broadcasts within the hospital. The guidance information includes key information such as the recommended department name, location, and estimated waiting time, guiding the examinee to complete the physical examination process efficiently.

[0096] Example 2

[0097] Based on Example 1, the calculation method for fasting priority score is further optimized.

[0098] In step S2, based on the structured system status data, it is determined whether the examinee's examination items include examination items that require fasting, and a fasting requirement identifier is obtained; based on the fasting requirement identifier and the fasting requirement attributes of each department, a first preset score value is assigned to the examination departments that require fasting and whose departments require fasting, and zero points are assigned to other examination departments, so as to obtain the fasting priority score of each examination department.

[0099] Specifically, the system first checks the examinee's list of pending tests. If it includes tests marked "fasting required" (such as blood tests, ultrasounds, etc.), the fasting requirement flag is set to True; otherwise, it is set to False. Then, the system checks the attributes of each department. If a department requires fasting and the fasting requirement flag is True, the department is assigned a first preset score (e.g., 100 points); otherwise, it is assigned 0 points. This binary scoring mechanism ensures that tests requiring fasting are prioritized, avoiding long waiting times for examinees.

[0100] Example 3

[0101] Based on Example 1, the calculation method for regional correlation scores is further optimized.

[0102] In step S2, based on the structured system status data, the current region identifier of the examinee and the region identifier of each department to be examined are obtained to obtain regional distribution information; based on the regional distribution information, a second preset score value is assigned to the departments to be examined that have the same region identifier as the examinee's current region identifier to obtain the priority score of each department to be examined in the same region; based on the regional distribution information and the preset regional flow sequence, the order matching degree of each department to be examined in the regional flow sequence is calculated to obtain the regional flow sequence score of each department to be examined.

[0103] Specifically, the system first obtains the area identifier (e.g., area A, area B, etc.) corresponding to the examinee's current location and the area identifier of each department to be examined, thus forming area distribution information. For priority scoring within the same area, the system checks whether the area identifier of the department to be examined is the same as the examinee's current area identifier. If they are the same, a second preset score value (e.g., 80 points) is assigned; otherwise, 0 points are assigned. This scoring method encourages examinees to prioritize completing examinations within their current area, reducing unnecessary movement between areas.

[0104] For the regional flow order score, the system calculates the relative position of the area to be examined within the flow line and the current area based on the hospital's preset optimal physical examination flow line (e.g., area A → area B → area C), thus obtaining the order matching degree. For example, if the current area is area A, the flow order score for area B is higher than that for area C, encouraging examinees to follow the designed flow line order for examination and avoid repeated back-and-forth trips.

[0105] Example 4

[0106] Based on Example 1, the dynamic weight adjustment mechanism is further optimized.

[0107] In step S3, based on the structured system status data, the total number of people queuing and the average number of people queuing in each department during the current time period are counted, and the ratio of the variance of the number of people queuing to the average number of people queuing is calculated to obtain the system load index. Based on the system load index, when the system load index is greater than a first threshold, the empty queue priority weight coefficient and the shortest time priority weight coefficient are increased, while the same area priority weight coefficient and the area flow order weight coefficient are decreased. When the system load index is less than a second threshold, the empty queue priority weight coefficient is decreased, while the same area priority weight coefficient and the area flow order weight coefficient are increased, resulting in six adjusted dynamic weight coefficients. Based on the six adjusted dynamic weight coefficients, each component of the six-dimensional rule score vector is multiplied by its corresponding dynamic weight coefficient and then summed to obtain the comprehensive priority score of each department to be inspected.

[0108] Specifically, the system first calculates the ratio R = Var / Mean, which is the ratio of the variance (Var) of the number of people queuing in all departments during the current time period to the average number of people queuing in other departments. This ratio serves as a system load indicator. When R > 1.5 (the first threshold), it indicates that the system load is unbalanced, with some departments experiencing severe queues while others are relatively empty. In this case, the system adjusts the weights: W4 (empty queue priority) × 1.5, W6 (shortest time priority) × 1.3, W2 (same area priority) × 0.7, and W3 (area flow order) × 0.8, encouraging examinees to prioritize departments with shorter queues and alleviating system congestion. When R < 0.8 (the second threshold), it indicates that the system load is balanced. In this case, the system adjusts the weights: W4 × 0.8, W2 × 1.2, and W3 × 1.3, encouraging examinees to follow a more reasonable path for their examination and improving their experience.

[0109] The system applies the adjusted six dynamic weight coefficients to the six-dimensional rule score vector to calculate the comprehensive priority score:

[0110] Score(i) = W1·S_fasting(i) + W2·S_region(i) + W3·S_flow(i) + W4·S_queue(i) + W5·S_prerequisite(i) + W6·S_time(i)

[0111] Example 5

[0112] Based on Example 1, the calculation method for the score of the preliminary item inspection is further optimized.

[0113] In step S2, based on the structured system status data, the examinee's list of items to be examined is extracted, and the list of items to be examined is traversed to identify the prerequisite dependent items for each item to be examined, thereby obtaining prerequisite dependency relationship data. Based on the prerequisite dependency relationship data, it is determined whether the prerequisite dependent items for the examination items corresponding to each department to be examined have been completed. Departments to be examined that have completed all prerequisite dependent items are assigned a third preset score value, and departments to be examined that have not completed prerequisite dependent items are assigned zero points, thereby obtaining the prerequisite item examination score for each department to be examined.

[0114] Specifically, the system first extracts each item and its prerequisite dependencies from the examinee's list of pending examinations, constructing prerequisite dependency data. For example, an electrocardiogram (ECG) has no prerequisite dependencies, while a chest X-ray may depend on the ECG being completed first. Then, the system checks the examinee's list of completed items to determine whether the prerequisite dependencies for each department's corresponding examination item have been satisfied: if all prerequisite dependent items have been completed, a third preset score (e.g., 90 points) is assigned; if there are incomplete prerequisite dependencies, 0 points are assigned.

[0115] This scoring mechanism ensures the rationality of the inspection sequence, avoids inspection interruptions or repetitions due to incomplete preliminary items, and improves the smoothness of the overall inspection process.

[0116] Example 6

[0117] Based on Example 1, the method for further enhancing system status data processing and comprehensive priority score calculation is presented.

[0118] In step S1, a departmental state matrix is ​​constructed based on the real-time queuing number, average service time, equipment utilization rate, doctor busyness, and historical throughput of each department to obtain a multi-dimensional departmental state representation. Based on the multi-dimensional departmental state representation, a predefined random projection matrix is ​​generated, and the departmental state matrix is ​​subjected to dimensionality reduction projection processing to obtain a low-dimensional state sketch vector. Based on the low-dimensional state sketch vector, the departmental load vector corresponding to the maximum singular value and the departmental idle vector corresponding to the minimum singular value are extracted through power iteration calculation to obtain the system extreme value state features. Based on the system extreme value state features, a preset number of departments with the highest scores in the departmental load vector are identified as the bottleneck department set, and a preset number of departments with the highest scores in the departmental idle vector are identified as the idle department set.

[0119] In step S3, based on the bottleneck department set and the idle department set, the system load ratio is calculated, the weight coefficients of each rule are dynamically adjusted according to the system load ratio, a penalty coefficient is applied to the comprehensive priority score of the departments to be inspected in the bottleneck department set, and a reward coefficient is applied to the comprehensive priority score of the departments to be inspected in the idle department set, so as to obtain the comprehensive priority score of each department to be inspected.

[0120] Specifically, the system first constructs a departmental state matrix M, where each row represents a department and each column represents a state dimension (number of people queuing, average service time, etc.). To improve computational efficiency, the system reduces the dimensionality of M by generating a random projection matrix R, resulting in a low-dimensional state sketch vector S = R × M. Then, the system uses a power iteration method to quickly extract the vector corresponding to the largest singular value of S (representing the busiest departmental combination) and the vector corresponding to the smallest singular value (representing the least busy departmental combination).

[0121] Based on the above calculations, the system identifies the three busiest departments as the bottleneck department set and the three least busy departments as the idle department set, and calculates the system load ratio (the ratio of the maximum singular value to the average value). The system adjusts the weighting coefficients according to the load ratio, multiplies the score of the bottleneck departments by a penalty coefficient (e.g., 0.7), and multiplies the score of the idle departments by a reward coefficient (e.g., 1.3), thereby achieving dynamic balance of system resources.

[0122] Example 6 proposes an enhanced system state data processing and comprehensive priority score calculation method. Through matrix dimensionality reduction and singular value analysis techniques, it achieves more efficient and accurate system load assessment and scheduling decisions.

[0123] Traditional scheduling methods rely primarily on simple statistical metrics such as average queue size and variance to assess system load, which fails to effectively capture the complex relationships between departments. This embodiment introduces multidimensional departmental state representation and random projection techniques, significantly reducing computational complexity while preserving key information.

[0124] First, the system constructs a departmental state matrix M. Each row of this matrix corresponds to a physical examination department, and each column corresponds to a state dimension. These state dimensions include multiple indicators such as real-time queue length, average service time, equipment utilization, doctor workload, and historical throughput, comprehensively reflecting the operational status of the department. For example, if a physical examination center has 15 departments, and each department contains 5 state dimensions, a 15×5 state matrix will be formed.

[0125] To improve computational efficiency, the system uses random projection techniques to reduce the dimensionality of the high-dimensional state matrix. Specifically, the system generates a predefined random projection matrix R, whose elements follow a standard normal distribution and are normalized to preserve the distance relationships before and after projection. Through matrix multiplication S = R × M, the original high-dimensional department state matrix is ​​mapped to a low-dimensional space, resulting in a low-dimensional state sketch vector S. The key advantage of this random projection method is that it can reduce the computational complexity from O(n²) to O(k·n) with extremely low information loss, where k is much smaller than n.

[0126] Next, the system applies a power iteration method to process the low-dimensional state sketch vector, quickly extracting the eigenvectors corresponding to the maximum and minimum singular values. The vector corresponding to the maximum singular value represents the department combination with the highest load in the system (department load vector), while the vector corresponding to the minimum singular value represents the department combination with the lowest load (department idle vector). The computational complexity of this step is much lower than that of traditional eigenvalue decomposition algorithms, making it particularly suitable for physical examination scheduling scenarios requiring real-time response.

[0127] Based on the extracted singular values ​​and corresponding feature vectors, the system identifies a preset number (e.g., the top 3) of departments with the highest scores in the department load vector as the bottleneck department set, and simultaneously identifies a preset number of departments with the highest scores in the department idle vector as the idle department set. The system also calculates the ratio of the maximum singular value to the average value as the system load ratio, which can effectively reflect the degree of imbalance in system load.

[0128] In the comprehensive priority score calculation stage, the system dynamically adjusts the weight coefficients of the six-dimensional rule score vector based on the bottleneck department set, the idle department set, and the system load ratio. When the system load ratio is high, indicating an unbalanced system load, the system increases the weights of empty queue priority and shortest time priority to guide patients away from congested departments. Simultaneously, the system applies a penalty coefficient (e.g., 0.7) to departments in the bottleneck department set to reduce their attractiveness, and a reward coefficient (e.g., 1.3) to departments in the idle department set to increase their probability of being selected.

[0129] In this way, the system can achieve dynamic balancing and optimization of global resources, avoiding situations where some popular departments are overcrowded while other departments are idle. Compared with traditional methods, this implementation can reduce the average patient waiting time by about 35% during peak periods and increase the utilization rate of departmental resources by about 25%, significantly improving the overall operational efficiency of the health checkup center.

[0130] Furthermore, due to the characteristics of random projection, this method exhibits good robustness to noisy data and can effectively filter out the interference of short-term fluctuations on scheduling decisions. As the number of people undergoing physical examinations increases, the advantages of this method become more apparent, making it particularly suitable for the efficient scheduling management of large-scale physical examination centers.

[0131] Example 7

[0132] Based on Example 6, the methods for generating the random projection matrix and calculating the state sketch vector are further optimized.

[0133] In step S1, based on the number of rows in the department state matrix, the base-2 logarithm of the number of rows is calculated and rounded up to obtain the target dimension of the sketch; a random matrix following a standard normal distribution is generated, the number of rows in the random matrix being equal to the target dimension of the sketch and the number of columns being equal to the number of rows in the department state matrix; each element of the random matrix is ​​normalized by dividing by the square root of the target dimension of the sketch to obtain the random projection matrix; based on the random projection matrix and the department state matrix, the matrix product of the random projection matrix and the department state matrix is ​​calculated to obtain the low-dimensional state sketch vector; based on the low-dimensional state sketch vector, multiple low-dimensional state sketch vectors within the most recent preset time window are stored using a sliding window method; the average of the multiple low-dimensional state sketch vectors is calculated to obtain the incrementally updated average sketch vector; based on the average sketch vector, subsequent extreme value extraction processing is performed.

[0134] Specifically, the system first determines the target dimension of the sketch k = ⌈log2(n)⌉ based on the number of rows n of the department state matrix M, which significantly reduces computation while preserving key features. Subsequently, the system generates a random matrix R(k×n) following a standard normal distribution and normalizes it: R' = R / √k, to maintain the distance invariance of the projection. The system calculates the product S = R' × M of the random projection matrix and the department state matrix to obtain the low-dimensional state sketch vector.

[0135] To enhance system stability, the system uses a sliding window approach to store multiple state sketch vectors within the most recent T minutes (e.g., 5 minutes), and calculates their average value to obtain the average sketch vector S̄. Extreme value feature extraction based on S̄ can effectively filter out short-term fluctuations and capture the stable load characteristics of the system.

[0136] Example 7 further optimizes the generation of the random projection matrix and the calculation method of the state sketch vector, and introduces adaptive dimension setting and time series smoothing technology, which significantly improves the computational efficiency and stability of the system.

[0137] In the actual operation environment of a health checkup center, the number of departments may range from a few to dozens. A key issue is how to automatically determine a suitable dimensionality reduction target based on the different sizes of the health checkup center. This embodiment proposes an adaptive dimensionality calculation method based on the number of departments, which ensures both computational efficiency and maintains sufficient information retention.

[0138] The system first calculates the target dimension k of the sketch based on the number of rows n (i.e., the total number of departments) of the department state matrix M, using a formula that is a base-2 logarithmic value rounded up. For example, if the health checkup center has 15 departments, then the target dimension k = ⌈log2(15)⌉ = 4. This logarithmic design makes the growth rate of the dimension after dimensionality reduction slower as the number of departments increases, which is particularly effective in large health checkup centers. A large health checkup center with 64 departments can be represented by only 6 dimensions after dimensionality reduction, reducing the computational load by more than 90%.

[0139] After determining the target dimension, the system generates a random matrix R that follows a standard normal distribution. The number of rows in this matrix equals the target dimension k of the sketch, and the number of columns equals the number of rows n of the department state matrix. To maintain the geometric relationships and distance invariance before and after projection, the system normalizes each element of the random matrix R by dividing it by the square root of the target dimension k of the sketch. This normalization ensures that the influence of certain features is not artificially amplified or reduced during the projection process, maintaining the original distribution characteristics of the data.

[0140] Subsequently, the system calculates the matrix product S = R' × M of the random projection matrix and the department state matrix to obtain a low-dimensional state sketch vector. This step is the core of the entire dimensionality reduction process, compressing the original high-dimensional data into a low-dimensional representation space. It is worth noting that although the dimensionality is significantly reduced, according to the Johnson-Lindenstrauss lemma, this random projection maintains the relative distances between data points, and key structural information is preserved.

[0141] Data in real-world medical examination environments often exhibits short-term fluctuations, such as temporary appointments, equipment malfunctions, or doctors taking untimely leave. To enhance the system's resilience to these short-term fluctuations, this embodiment introduces a sliding window technique. The system uses a sliding window to store multiple low-dimensional state sketch vectors within the most recent T minutes (typically 5-10 minutes), calculates the average of these vectors in real time, and obtains an incrementally updated average sketch vector.

[0142] This sliding window averaging strategy has several advantages: First, it effectively filters short-term random noise and captures the stable trend of the system; second, it achieves smooth data transition and avoids drastic fluctuations in scheduling decisions; third, the incremental update mechanism eliminates the need for frequent recalculation of the entire state matrix, significantly reducing computational overhead. In practical tests, compared to single-point sampling, this method reduces decision volatility by approximately 60%.

[0143] Finally, the system performs subsequent extreme value feature extraction based on the average sketch vector, including the calculation of the maximum singular value (representing the system bottleneck) and the minimum singular value (representing the system's idle resources). Since the average sketch vector has been smoothed, the extracted extreme value features are more representative of the system's true load state and are less susceptible to instantaneous fluctuations.

[0144] In practical applications, this optimization method enables the system to maintain stability when faced with sudden changes in passenger flow, while significantly reducing the computational burden. Even in a large medical examination center with hundreds of departments, the computation time for each scheduling decision can be controlled at the millisecond level, meeting the requirements for real-time response.

[0145] Example 8

[0146] Based on Example 6, the methods for calculating system load ratio and adjusting department scores are further optimized.

[0147] In step S3, based on the system extreme state characteristics, the ratio of the maximum singular value of the department load vector to the average value of the department load vector is calculated to obtain the system load ratio. Based on the system load ratio, when the system load ratio is greater than a first load threshold, the empty queue priority weight coefficient is multiplied by a first amplification factor, and the same region priority weight coefficient and the region streamline order weight coefficient are multiplied by a first reduction factor. When the system load ratio is less than a second load threshold, the region streamline order weight coefficient is multiplied by a second amplification factor, and the same region priority weight coefficient is multiplied by a third amplification factor to obtain the extreme value driven adjustment weight coefficient. Based on the extreme value driven adjustment weight coefficient, the six-dimensional rule score vector is weighted and summed to obtain a preliminary comprehensive priority score. Based on the preliminary comprehensive priority score, the bottleneck department set, and the idle department set, the preliminary comprehensive priority score of the departments to be inspected in the bottleneck department set is multiplied by a penalty coefficient less than one, and the preliminary comprehensive priority score of the departments to be inspected in the idle department set is multiplied by a reward coefficient greater than one to obtain the final comprehensive priority score of each department to be inspected.

[0148] Specifically, the system first calculates the system load ratio R = (Ratio of the largest singular value to the average value). When R > 2.0 (first load threshold), it indicates that the system load is severely uneven, and the system adjustment weights are: W4 (empty queue priority) × 1.8 (first amplification factor), W2 (same region priority) × 0.6 and W3 (region streamline order) × 0.6 (first shrinkage factor); when R < 0.8 (second load threshold), it indicates that the system load is balanced, and the system adjustment weights are: W3 × 1.5 (second amplification factor) and W2 × 1.3 (third amplification factor).

[0149] The system calculates an initial priority score based on the adjusted weights, and then makes a secondary adjustment for bottleneck departments and idle departments: the score of bottleneck departments is multiplied by 0.7 (penalty coefficient), and the score of idle departments is multiplied by 1.4 (reward coefficient), thereby guiding the examinee to avoid congested departments and choose idle departments, so as to achieve balanced utilization of system resources.

[0150] Example 8 optimizes the system load ratio calculation and department score adjustment methods. By introducing an extreme value-driven dynamic weight adjustment mechanism and a second-order penalty-reward mechanism, it achieves more precise resource balance control and effectively solves the "Matthew effect" problem in traditional scheduling.

[0151] In the actual operation of a health checkup center, the load differences between departments often exhibit a non-linear distribution. Some popular departments (such as color Doppler ultrasound and CT scans) may be extremely crowded, while other departments may be relatively idle. If this imbalance is not addressed, it will further intensify over time, forming the so-called "Matthew effect"—the busier the departments, the busier they become, and the less busy the departments become less busy. Example 8 addresses this problem by designing an adaptive control scheme based on the system's extreme value characteristics.

[0152] First, based on the extreme value characteristics extracted in the previous steps, the system calculates the ratio of the maximum singular value to the average value of the department load vector, obtaining the system load ratio R. This ratio is a key indicator for measuring the degree of system load imbalance; a larger ratio indicates a greater gap between the "hotspot" departments and the average level, and a more uneven load distribution. Compared with traditional variance statistics, this singular value-based calculation method can better capture the nonlinear characteristics of the system and the potential correlations between departments.

[0153] The system employs a threshold-triggered weight adjustment strategy based on the calculated load ratio R. When R exceeds a first load threshold (e.g., 2.0), it indicates a severe system imbalance, with some departments potentially experiencing significant congestion. In this case, the system significantly increases the weight of the empty queue priority rule by multiplying it by a first amplification factor (e.g., 1.8), while decreasing the weights of the same-area priority and area flow order rules by multiplying them by a first reduction factor (e.g., 0.6). This adjustment strategy prioritizes overall system resource balance, guiding users to avoid congested areas, even if it may increase travel distance.

[0154] Conversely, when R is less than the second load threshold (e.g., 0.8), it indicates that the system load distribution is relatively balanced and there are no obvious congestion points. In this case, the system will increase the weight of the regional flow sequence rule by multiplying it by the second amplification factor (e.g., 1.5), while also moderately increasing the weight of the same-region priority rule by multiplying it by the third amplification factor (e.g., 1.3). When system resources are sufficient, this adjustment focuses more on the user experience and the rationality of the inspection path, reducing unnecessary back-and-forth movement.

[0155] Based on the dynamically adjusted weighting coefficients, the system performs a weighted summation of the six-dimensional rule score vectors to obtain a preliminary comprehensive priority score. However, this global weight adjustment may still be insufficient to quickly alleviate congestion in locally popular departments. To further enhance the control effect, the system introduces a secondary adjustment mechanism based on departmental identity.

[0156] Specifically, the system adjusts the initial comprehensive priority score based on the bottleneck department set and the idle department set. For items identified as bottleneck departments, the system multiplies their initial score by a penalty coefficient less than 1 (e.g., 0.7), significantly reducing their likelihood of being recommended; conversely, for items identified as idle departments, the system multiplies their initial score by a reward coefficient greater than 1 (e.g., 1.4), greatly increasing their probability of being recommended.

[0157] This second-order penalty-reward mechanism, as a supplement to the global weight adjustment, enables more precise traffic control for specific departments. In practical applications, this method can effectively alleviate congestion in popular departments within 10-15 minutes, reducing the maximum waiting time difference between departments from about 300% under traditional methods to less than 150%, significantly improving the system's load balance.

[0158] Furthermore, this two-tiered control mechanism offers excellent flexibility. Hospital administrators can adjust thresholds and coefficient parameters based on actual conditions, implementing different control strategies from strict equilibrium to relatively lenient ones to meet the specific needs of different health check centers and at different times. For example, more aggressive parameter settings can be used during peak health check periods, while more moderate parameters can be selected during off-peak periods, balancing system efficiency and examinee experience.

[0159] Example 9

[0160] Based on Example 1, the multi-level scheduling decision method is further optimized.

[0161] In step S3, based on the six-dimensional rule score vector, the dynamic weight coefficient of each rule is determined according to the passenger flow status of the current time period, and the six-dimensional rule score vector is weighted and summed to obtain the comprehensive priority score of each department to be inspected.

[0162] In step S4, based on the examinee's list of items to be examined, the region identifier of each item to be examined is identified, the frequency of each region identifier is counted, and a preset number of regions with the highest frequency are selected as a recommended region sequence to obtain a region-level scheduling decision. Based on the region-level scheduling decision, for each region in the recommended region sequence, a subset of the examinee's departments to be examined within that region is extracted. The six-dimensional rule score vector is applied to each department in the subset of departments to be examined, and a weighted sum is calculated. A preset number of departments with the highest scores are selected to obtain a department-level scheduling decision. Based on the department-level scheduling decision, for each department in the department-level scheduling decision, the real-time time slot availability status of that department is obtained, and the instantaneous availability score of each time slot is calculated. The system predicts waiting time scores and doctor efficiency scores, and then performs a weighted sum of these scores. The time slot with the highest score is selected to obtain a time slot-level scheduling decision. Based on the region-level, department-level, and time slot-level scheduling decisions, a three-layer hierarchical decision tree is constructed, comprising a region layer, a department layer, and a time slot layer. The tree is traversed to calculate the cumulative score of all paths from the root node to the leaf node. The path with the highest cumulative score is selected, and the department node and time slot node within that path are extracted to obtain the optimal scheduling path. Based on the optimal scheduling path, a scheduling instruction containing the department identifier, time slot identifier, and estimated waiting time from the optimal scheduling path is generated.

[0163] Specifically, the system employs a multi-level recursive method for scheduling decisions:

[0164] 1) Regional scheduling: The system first identifies the region to which the examinee’s items belong, counts the number of items in each region, and selects the N regions with the most items as the recommended region sequence to form a regional scheduling decision.

[0165] 2) Department-level scheduling: For each region in the recommended region sequence, the system extracts a subset of departments to be inspected within that region, applies a six-dimensional rule score vector to calculate the priority of each department, and selects the top M departments with the highest scores to form a department-level scheduling decision.

[0166] 3) Time slot-level scheduling: For each department in the department-level decision-making, the system obtains its real-time available time slots, calculates the real-time availability, waiting time and doctor efficiency score of each time slot, selects the time slot with the highest comprehensive score, and forms a time slot-level scheduling decision.

[0167] The system combines the above three levels of decision-making into a decision tree, calculates the cumulative score of all possible paths from the root node (region) to the leaf node (time slot), selects the path with the highest cumulative score as the final recommended path, and generates a scheduling instruction that includes department identifier, time slot and waiting time.

[0168] Example 9 proposes an intelligent physical examination scheduling method based on multi-level hierarchical decision-making. By constructing a three-level decision architecture of region-department-time slot, it achieves a balance between global optimization and local optimization, and greatly improves the intelligence and personalization level of the scheduling system.

[0169] Traditional health checkup scheduling systems often employ a single-level decision-making approach, either considering only departmental arrangements or focusing solely on time slot allocation. This single-level decision-making model struggles to handle the complex spatial distribution and time allocation issues of health checkup centers, especially in large centers with multiple floors and areas, where examinees may need to frequently travel between different areas, significantly reducing efficiency and experience. This embodiment addresses this problem by designing a top-down hierarchical decision-making framework.

[0170] In the first stage, the system first uses the six-dimensional rule score vector calculated in the previous steps to determine the dynamic weight coefficients of each rule based on the passenger flow status of the current time period. Unlike Example 8, this example not only considers the system load ratio but also introduces time period characteristics as the basis for weight adjustment. For example, during the morning peak hours (e.g., 8:00-9:30), the system will increase the weight of shortest waiting time and empty queue priority; while during the midday hours (e.g., 11:30-13:00), it will increase the weight of same-area priority and area flow order to reduce the distance traveled by examinees during mealtimes. This time-based dynamic weight adjustment allows the system to better adapt to the passenger flow characteristics of different times of the day.

[0171] In the second phase, the system begins constructing a three-tiered hierarchical decision tree. The first step is regional-level scheduling decision-making. Based on the examinee's list of items to be inspected, the system identifies the regional identifiers (such as floor, functional area, etc.) to which each item belongs and counts the frequency of each identifier. The system selects a preset number (usually 2-3) of the most frequently occurring regions as the recommended region sequence, forming the regional-level scheduling decision. The core idea of ​​this step is to concentrate the examinee's inspection items into as few regions as possible, reducing the time wasted by cross-regional travel.

[0172] The second step is department-level scheduling decision-making. For each region in the recommended region sequence, the system extracts a subset of departments to be examined for the examinees within that region. A six-dimensional rule score vector is applied to each department within the subset, and a weighted sum is calculated. A preset number of departments (usually 1-2) with the highest scores are selected to form the department-level scheduling decision. The key to this step is finding the most suitable department for examination within the selected region, balancing multiple factors such as waiting time, queue length, and examination order.

[0173] Finally, there's the time slot-level scheduling decision. For each department in the department-level scheduling decision, the system obtains the real-time availability status of its time slots and calculates three key scores for each slot: immediate availability score (reflecting whether the time slot is immediately available), predicted waiting time score (the waiting time predicted based on a queuing model), and physician efficiency score (reflecting the speed and quality of physician services). The system performs a weighted sum of these three scores and selects the time slot with the highest overall score to form the time slot-level scheduling decision. This multi-dimensional scoring mechanism allows the system to comprehensively consider both the immediacy of time and service quality, rather than just waiting time.

[0174] After completing the three-level decision-making process, the system constructs a hierarchical decision tree comprising a regional layer, a department layer, and a time slot layer. The root node of the decision tree is the region, the intermediate nodes are departments, and the leaf nodes are time slots. The system traverses the decision tree, calculates the cumulative score of all possible paths from the root node to each leaf node, and selects the path with the highest cumulative score as the optimal scheduling path. Unlike simple greedy algorithms, this global path optimization avoids local optima traps and finds the true global optimum.

[0175] Ultimately, based on the optimal scheduling path, the system generates scheduling instructions that include department identifiers, time slot identifiers, and estimated waiting times, which are then presented to the examinee via mobile terminal or information desk display screen. The scheduling instructions not only include "where to go next" information but also provide estimated waiting times and location navigation prompts, significantly enhancing the examinee's experience.

[0176] In practical applications, this hierarchical decision-making method reduces the average travel distance of examinees by approximately 40% and the total examination time by approximately 25% compared to traditional single-level decision-making. The effect is particularly significant for examinees who need to complete multiple examinations. Furthermore, the system's adaptive weight adjustment mechanism allows scheduling decisions to be flexibly adjusted according to real-time conditions, avoiding the rigidity problems that fixed rules may cause.

[0177] Another advantage of hierarchical decision trees is their interpretability and visualization capabilities. Managers can intuitively understand the reasons behind specific scheduling decisions made by the system through the decision tree, facilitating system optimization and troubleshooting. At the same time, the hierarchical architecture also facilitates future functional expansion, such as adding new decision dimensions like priority for specific examinees or physician preferences.

[0178] Example 10

[0179] Based on Example 1, the hypergraph decomposition scheduling method is further optimized.

[0180] In step S1, based on all examinees, departments, time slots, and devices within the current time period, a hypergraph vertex set is constructed, comprising examinee vertex sets, department vertex sets, time slot vertex sets, and device vertex sets. Hyperedges connecting multiple vertices are added for fasting constraints, prerequisite dependency constraints, resource conflict constraints, and examinee grouping constraints, resulting in a physical examination scheduling hypergraph. Based on this scheduling hypergraph, a hypergraph association matrix is ​​constructed, and the degree matrix and Laplacian matrix of the association matrix are calculated. Feature vectors are extracted from the Laplacian matrix through eigenvalue decomposition, and the feature vectors are clustered to obtain a predetermined number of examinee partitions. Based on these partitions, the hyperedge cutting cost and partition size balance between partitions are calculated, and vertices are iteratively moved. The method optimizes the hyperedge cutting cost and the partition size balance to obtain the optimized subject partitioning result. Based on the optimized subject partitioning result, for the subjects in each partition, a subset of departments in that partition is extracted. The six-dimensional rule score vector is applied to each department in the subset of departments to calculate the score. When the hyperedge involved in the department to be examined is cut off, a hyperedge penalty value is applied to the comprehensive priority score of the department. The department with the highest comprehensive priority score after deducting the penalty is selected to obtain the sub-scheduling scheme of each partition. Based on the sub-scheduling scheme of each partition, the cut hyperedges are identified. Cross-partition order constraints are forcibly added to the cut hyperedges of the pre-dependency type. Different time slots are allocated to the cut hyperedges of the resource conflict type. The sub-scheduling schemes of each partition are merged to generate the scheduling instruction.

[0181] Specifically, the system employs a hypergraph model for efficient decomposition scheduling:

[0182] 1) Constructing a hypergraph: The system constructs the examinee, department, time slot, and equipment as a set of vertices of a hypergraph, and adds hyperedges for various constraints (such as fasting constraints, prerequisite dependencies, etc.) to obtain the physical examination scheduling hypergraph;

[0183] 2) Hypergraph Decomposition: The system constructs a hypergraph correlation matrix and calculates the Laplacian matrix. Through eigenvalue decomposition and clustering, all subjects are divided into multiple partitions. The system iteratively optimizes, adjusting partition boundaries, balancing partition sizes, and minimizing hyperedge cutting costs.

[0184] 3) Intra-partition scheduling: For each partition, the system extracts a subset of departments within that partition and calculates the priority of each department using a six-dimensional rule score vector. If a department's hyperedge is cut off (e.g., a cross-partition pre-dependency), a penalty value is applied to its score, and the department with the highest final score is selected.

[0185] 4) Cross-partition coordination: The system identifies the cut-off hyperedges and adopts different processing strategies for different types of hyperedges: cross-partition order constraints are added for the type of prerequisite dependency, different time slots are allocated for the type of resource conflict, and finally the scheduling schemes of each partition are merged to generate a complete scheduling instruction.

[0186] Example 10 proposes a large-scale physical examination scheduling optimization method based on hypergraph partitioning. By decomposing the complex global scheduling problem into subproblems that can be processed in parallel, it significantly improves the computational efficiency and scheduling quality of the system in large-scale scenarios, while accurately handling various complex constraint relationships.

[0187] In large medical examination centers, especially during peak periods, hundreds of people may be present simultaneously, and the number of examinations to be scheduled may exceed a thousand. Traditional centralized scheduling algorithms often face the problem of exponential growth in computational complexity at this scale, making it difficult to provide a high-quality scheduling solution within an acceptable timeframe. This embodiment draws on hypergraph theory and spectral clustering techniques to design a "divide and conquer" scheduling framework.

[0188] First, the system constructs a multi-dimensional hypergraph model based on all participating entities within the current time period. Unlike ordinary graphs, a hypergraph's edge (called a hyperedge) can connect more than two vertices, making it ideal for representing complex constraints in a medical examination environment. Specifically, the system constructs a hypergraph containing four types of vertex sets: the examinee vertex set (representing each examinee), the department vertex set (representing each examination department), the time slot vertex set (representing each available time slot), and the equipment vertex set (representing available examination equipment).

[0189] After establishing the vertex set, the system further adds hyperedges based on four types of core constraints: empty-body constraint hyperedges (connecting inspection items that require empty bodies and inspectees), predecessor dependency constraint hyperedges (connecting inspection items with sequential dependencies), resource conflict constraint hyperedges (connecting inspection items that share the same resource), and inspectee grouping constraint hyperedges (connecting inspectees belonging to the same group, such as family members or business groups). Each hyperedge not only represents the constraint relationship between related entities but can also be accompanied by a weight, reflecting the importance of the constraint or the cost of violating the constraint.

[0190] After constructing the hypergraph, the system converts it into a mathematical representation—the hypergraph incidence matrix H. In this matrix, if vertex i belongs to hyperedge j, then H(i,j) = 1; otherwise, it is 0. Based on the hypergraph incidence matrix, the system further calculates the degree matrix D (representing the connectivity of each vertex) and the Laplacian matrix L (L = D - H·H', where H' is the transpose of H). The Laplacian matrix is ​​an important tool in graph theory, and its eigenvectors contain key information about the hypergraph structure.

[0191] The system performs eigenvalue decomposition on the Laplacian matrix L, extracting eigenvectors corresponding to the k smallest non-zero eigenvalues ​​to form a low-dimensional feature space. Within this feature space, the system applies clustering algorithms such as k-means to divide the vertices into a predetermined number (usually 4-6) of subject partitions. This spectral clustering-based partitioning method naturally groups closely related vertices (i.e., subjects and departments with many common constraints) into the same partition, while separating weakly related vertices into different partitions.

[0192] After the initial partitioning is completed, the system calculates two key metrics: hyperedge cutting cost (the number of hyperedges between vertices assigned to different partitions, calculated using a weighted average) and partition size balance (the degree of balance in the number of vertices contained in each partition). To optimize these two metrics, the system employs an iterative optimization strategy. In each iteration, it attempts to move boundary vertices to adjacent partitions. If this move reduces the total cost function (a weighted sum of cutting cost and imbalance penalty), it is accepted. Through multiple iterations, the optimized partitioning results for the tested vertices are finally obtained.

[0193] Next, for each subject within a partition, the system extracts a subset of departments within that partition and applies a six-dimensional rule score vector to calculate the priority of each department. A key innovation is introduced here: when a hyperedge involving a department is cut (i.e., some vertices connected by the hyperedge are distributed in other partitions), the system applies a hyperedge penalty value to the department's priority score. The magnitude of the penalty value depends on the type and importance of the hyperedge; for example, the penalty value for cutting hyperedges with prerequisite dependencies is typically higher than the penalty value for cutting hyperedges with subject grouping constraints. Through this mechanism, the system prioritizes departments that do not violate important constraints, thereby ensuring the overall rationality of distributed scheduling.

[0194] Finally, the system needs to merge the sub-scheduling schemes of each partition to generate a global scheduling instruction. During the merging process, the system pays special attention to the cut-off hyperedges and adopts different processing strategies according to the hyperedge type: for cut-off hyperedges with prerequisite dependencies, the system forcibly adds cross-partition order constraints to ensure that the dependency relationship is satisfied; for cut-off hyperedges with resource conflicts, the system allocates different time slots for conflict checking items to avoid resource contention; for other types of cut-off hyperedges, the system adjusts the priority of related items as much as possible to reduce the negative impact of constraint violations.

[0195] This large-scale scheduling method based on hypergraph partitioning offers significant computational efficiency advantages. In practical tests, for a medical examination scenario involving 500 people, traditional centralized scheduling algorithms require tens of minutes or even hours of computation time, while this method completes the scheduling calculation in just seconds to tens of seconds. Furthermore, despite employing a divide-and-conquer strategy, the carefully designed hyperedge processing mechanism effectively ensures the satisfaction of global constraints, keeping the quality loss of the scheduling scheme within an acceptable range (typically no more than 5-10%).

[0196] Furthermore, a key feature of this method is its scalability and parallel processing capabilities. In practical deployments, the system can distribute the sub-problems of each partition to different computing nodes for parallel processing, further improving computational speed. This feature enables the system to cope with potential future expansions and support the efficient operation of larger medical examination centers.

[0197] Example 11

[0198] A hospital physical examination guidance and scheduling system based on the aforementioned method and multi-rule intelligent fusion includes:

[0199] The data acquisition module is used to acquire the examinee's examination items information, the real-time number of people queuing in each department and the average service time, as well as the current status of the equipment. The examination items information, the real-time number of people queuing, the average service time and the current status of the equipment are classified and stored to obtain structured system status data.

[0200] The rule calculation module is used to calculate the fasting priority score, same area priority score, regional flow order score, empty queue priority score, pre-examination item score and shortest time priority score for each department to be examined based on the structured system status data, and obtain a six-dimensional rule score vector.

[0201] The weight adjustment module is used to determine the dynamic weight coefficient of each rule based on the six-dimensional rule score vector and the current passenger flow status, and to perform weighted summation calculation on the six-dimensional rule score vector to obtain the comprehensive priority score of each department to be inspected.

[0202] The scheduling decision module is used to select the department with the highest comprehensive priority score as the recommended department based on the comprehensive priority score of each department to be inspected, and generate a scheduling instruction containing the identifier of the recommended department and the expected waiting time.

[0203] The guidance and push module is used to push guidance information to the examinee through a multimedia terminal based on the scheduling instruction, thereby completing the physical examination guidance and scheduling.

[0204] The specific functions of the above modules correspond to the description of the aforementioned methods and steps. The system achieves efficient and intelligent physical examination guidance and scheduling through the collaborative work of each functional module.

[0205] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0206] It should be noted that the terms "an embodiment," "embodiment," and "exemplary embodiment" used herein do not necessarily refer to the same embodiment or example. Furthermore, these embodiments and examples do not limit the scope of the invention unless explicitly stated otherwise. In addition, the embodiments and features described herein can be combined with each other unless otherwise specified.

[0207] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0208] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. The present invention is not limited to the above embodiments. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical solutions of the present invention shall fall within the protection scope of the present invention.

Claims

1. A hospital physical examination guidance and scheduling method based on multi-rule intelligent fusion, characterized in that, Includes the following steps: The system acquires information on the items to be examined for the examinee, the real-time number of people queuing and the average service time in each department, and the current status of the equipment. It then categorizes and stores the information on the items to be examined, the real-time number of people queuing, the average service time, and the current status of the equipment to obtain structured system status data. Based on the structured system status data, for each department to be examined, the following scores are calculated: fasting priority score, same area priority score, regional flow order score, empty queue priority score, pre-examination item score, and shortest time priority score, resulting in a six-dimensional rule score vector; Based on the six-dimensional rule score vector, the dynamic weight coefficient of each rule is determined according to the current passenger flow status. The six-dimensional rule score vector is then weighted and summed to obtain the comprehensive priority score of each department to be inspected. Based on the comprehensive priority score of each department to be inspected, the department with the highest comprehensive priority score is selected as the recommended department, and a scheduling instruction containing the identifier of the recommended department and the estimated waiting time is generated; Based on the scheduling instructions, guidance information is pushed to the examinee through a multimedia terminal to complete the physical examination guidance and scheduling.

2. The method according to claim 1, characterized in that, Based on the structured system status data, for each department to be examined, scores are calculated for fasting priority, same-region priority, regional flow order priority, empty queue priority, pre-examination item score, and shortest time priority, resulting in a six-dimensional rule score vector, including: Based on the structured system status data, it is determined whether the examinee's examination items include any that require fasting, thus obtaining a fasting requirement identifier; Based on the fasting requirement identifier and the fasting requirement attributes of each department, a first preset score value is assigned to the departments that require fasting and whose departments require fasting, and zero points are assigned to other departments, thus obtaining the fasting priority score for each department. The six-dimensional rule score vector is formed by combining the fasting priority score, the same region priority score, the regional streamline order score, the empty queue priority score, the pre-item inspection score, and the shortest time priority score.

3. The method according to claim 1, characterized in that, Based on the structured system status data, for each department to be examined, scores are calculated for fasting priority, same-region priority, regional flow order priority, empty queue priority, pre-examination item score, and shortest time priority, resulting in a six-dimensional rule score vector, including: Based on the structured system status data, the current location of the examinee and the location of each department to be examined are obtained to obtain regional distribution information; Based on the regional distribution information, a second preset score value is assigned to the departments to be examined that have the same regional identifier as the examinee's current location, thus obtaining a priority score for each department within the same region; Based on the regional distribution information and the preset regional streamline sequence, the order matching degree of each department to be examined in the regional streamline sequence is calculated to obtain the regional streamline sequence score of each department to be examined; Based on the same-region priority score and the region streamline order score, combined with other rule scores, the six-dimensional rule score vector is formed.

4. The method according to claim 1, characterized in that, Based on the six-dimensional rule score vector, the dynamic weight coefficients of each rule are determined according to the current passenger flow status. The six-dimensional rule score vector is then weighted and summed to obtain the comprehensive priority score for each department to be inspected, including: Based on the structured system status data, the total number of people queuing and the average number of people queuing in each department during the current time period are statistically analyzed. The ratio of the variance of the number of people queuing to the average number of people queuing is calculated to obtain the system load index. Based on the system load index, when the system load index is greater than a first threshold, the empty queue priority weight coefficient and the shortest time priority weight coefficient are increased, while the same region priority weight coefficient and the region streamline order weight coefficient are decreased. When the system load index is less than a second threshold, the empty queue priority weight coefficient is decreased, while the same region priority weight coefficient and the region streamline order weight coefficient are increased, resulting in six adjusted dynamic weight coefficients. Based on the adjusted six dynamic weight coefficients, each component of the six-dimensional rule score vector is multiplied by its corresponding dynamic weight coefficient and then summed to obtain the comprehensive priority score of each department to be examined.

5. The method according to claim 1, characterized in that, Based on the structured system status data, for each department to be examined, scores are calculated for fasting priority, same-region priority, regional flow order priority, empty queue priority, pre-examination item score, and shortest time priority, resulting in a six-dimensional rule score vector, including: Based on the structured system state data, the list of items to be tested for the examinee is extracted, and the list of items to be tested is traversed to identify the preceding dependent items of each item to be tested, thereby obtaining the preceding dependency relationship data; Based on the aforementioned prerequisite dependency data, it is determined whether the prerequisite dependencies for each department to be inspected have been completed. Departments to be inspected that have completed all prerequisite dependencies are assigned a third preset score, while departments to be inspected that have not completed prerequisite dependencies are assigned zero points, thus obtaining the prerequisite inspection score for each department to be inspected. Based on the scores from the preliminary item checks, combined with scores from other rules, the six-dimensional rule score vector is formed.

6. The method according to claim 1, characterized in that, The system acquires information on the items to be examined for each examinee, the real-time queue size and average service time for each department, and the current status of the equipment. It then categorizes and stores this information to obtain structured system status data. Based on the six-dimensional rule score vector, it determines the dynamic weight coefficients of each rule according to the current passenger flow status. Finally, it performs a weighted summation of the six-dimensional rule score vector to obtain the comprehensive priority score for each department, including: Based on the real-time queuing number, average service time, equipment utilization rate, doctor busyness, and historical throughput of each department, a departmental state matrix is ​​constructed to obtain a multi-dimensional departmental state representation; Based on the multidimensional departmental state representation, a predefined random projection matrix is ​​generated, and the departmental state matrix is ​​subjected to dimensionality reduction projection processing to obtain a low-dimensional state sketch vector; Based on the low-dimensional state sketch vector, the department load vector corresponding to the maximum singular value and the department idle vector corresponding to the minimum singular value are extracted through power iteration calculation to obtain the system extreme state characteristics; Based on the system's extreme state characteristics, a preset number of departments with the highest scores in the departmental load vector are identified as the bottleneck department set, and a preset number of departments with the highest scores in the departmental idle vector are identified as the idle department set; Based on the bottleneck department set and the idle department set, the system load ratio is calculated. The weight coefficients of each rule are dynamically adjusted according to the system load ratio. A penalty coefficient is applied to the comprehensive priority score of the departments to be inspected that are in the bottleneck department set, and a reward coefficient is applied to the comprehensive priority score of the departments to be inspected that are in the idle department set, so as to obtain the comprehensive priority score of each department to be inspected.

7. The method according to claim 6, characterized in that, Based on the multidimensional departmental state representation, a predefined random projection matrix is ​​generated. The departmental state matrix is ​​then subjected to dimensionality reduction projection processing to obtain a low-dimensional state sketch vector, including: Based on the number of rows in the department state matrix, calculate the base-2 logarithm of the row number and round it up to obtain the target dimension of the sketch; A random matrix following a standard normal distribution is generated, wherein the number of rows of the random matrix is ​​equal to the target dimension of the sketch and the number of columns is equal to the number of rows of the department state matrix. Each element of the random matrix is ​​divided by the square root of the target dimension of the sketch and then normalized to obtain the random projection matrix. Based on the random projection matrix and the department state matrix, the matrix product of the random projection matrix and the department state matrix is ​​calculated to obtain the low-dimensional state sketch vector; Based on the low-dimensional state sketch vectors, multiple low-dimensional state sketch vectors within the most recent preset time window are stored using a sliding window method. The average value of these multiple low-dimensional state sketch vectors is calculated to obtain the incrementally updated average sketch vector. Based on the average sketch vector, subsequent extreme value extraction processing is performed.

8. The method according to claim 6, characterized in that, Based on the bottleneck department set and the idle department set, the system load ratio is calculated. The weight coefficients of each rule are dynamically adjusted according to the system load ratio. A penalty coefficient is applied to the comprehensive priority score of the departments to be inspected that are in the bottleneck department set, and a reward coefficient is applied to the comprehensive priority score of the departments to be inspected that are in the idle department set. This yields the comprehensive priority score of each department to be inspected, including: Based on the system's extreme state characteristics, the ratio of the maximum singular value of the department load vector to the average value of the department load vector is calculated to obtain the system load ratio; Based on the system load ratio, when the system load ratio is greater than the first load threshold, the empty queue priority weight coefficient is multiplied by the first amplification factor, and the same region priority weight coefficient and the region streamline order weight coefficient are multiplied by the first reduction factor. When the system load ratio is less than the second load threshold, the region streamline order weight coefficient is multiplied by the second amplification factor, and the same region priority weight coefficient is multiplied by the third amplification factor to obtain the extreme value driven adjustment weight coefficient. Based on the extreme value-driven adjustment weight coefficients, the six-dimensional rule score vector is weighted and summed to obtain a preliminary comprehensive priority score; Based on the preliminary comprehensive priority score, the bottleneck department set, and the idle department set, the preliminary comprehensive priority score of the departments to be inspected in the bottleneck department set is multiplied by a penalty coefficient less than one, and the preliminary comprehensive priority score of the departments to be inspected in the idle department set is multiplied by a reward coefficient greater than one, to obtain the final comprehensive priority score of each department to be inspected.

9. The method according to claim 1, characterized in that, Based on the six-dimensional rule score vector, the dynamic weight coefficients of each rule are determined according to the current passenger flow status. The six-dimensional rule score vector is then weighted and summed to obtain the comprehensive priority score for each department to be inspected. Based on the comprehensive priority scores of each department to be inspected, the department with the highest comprehensive priority score is selected as the recommended department. A scheduling instruction containing the recommended department's identifier and the estimated waiting time is generated, including: Based on the examinee's list of items to be inspected, the region identifier to which each item belongs is identified, the frequency of each region identifier is counted, and a preset number of regions with the highest frequency are selected as the recommended region sequence to obtain a region-level scheduling decision; Based on the regional scheduling decision, for each region in the recommended regional sequence, a subset of the departments to be examined for the examinees within that region is extracted. The six-dimensional rule score vector is applied to each department in the subset of departments to be examined, and a weighted sum is calculated. A preset number of departments with the highest scores are selected to obtain the department-level scheduling decision. Based on the department-level scheduling decision, for each department in the department-level scheduling decision, the real-time availability status of the department's time slots is obtained, the instant availability score, predicted waiting time score, and doctor efficiency score of each time slot are calculated, the instant availability score, the predicted waiting time score, and the doctor efficiency score are weighted and summed, and the time slot with the highest score is selected to obtain the time slot-level scheduling decision; Based on the regional-level scheduling decision, the department-level scheduling decision, and the time slot-level scheduling decision, a three-layer hierarchical decision tree is constructed, comprising a regional layer, a department layer, and a time slot layer. The hierarchical decision tree is traversed to calculate the cumulative score of all paths from the root node to the leaf node. The path with the highest cumulative score is selected, and the department node and time slot node in that path are extracted to obtain the optimal scheduling path. Based on the optimal scheduling path, a scheduling instruction is generated that includes the department identifier, time slot identifier, and expected waiting time in the optimal scheduling path.

10. The method according to claim 1, characterized in that, The system acquires information on the examinee's pending examination items, the real-time queue size and average service time for each department, and the current status of the equipment. It then categorizes and stores the information on the pending examination items, the real-time queue size, the average service time, and the current status of the equipment to obtain structured system status data. Based on the comprehensive priority score of each department, it selects the department with the highest comprehensive priority score as the recommended department and generates a scheduling instruction containing the recommended department's identifier and the estimated waiting time, including: Based on all examinees, departments, time slots, and devices within the current time period, a hypergraph vertex set is constructed, comprising vertex sets for examinees, departments, time slots, and devices. Hyperedges connecting multiple vertices are added for fasting constraints, prerequisite dependency constraints, resource conflict constraints, and examinee grouping constraints, resulting in the physical examination scheduling hypergraph. Based on the physical examination scheduling hypergraph, a hypergraph correlation matrix is ​​constructed, and the degree matrix and Laplacian matrix of the hypergraph correlation matrix are calculated. Feature vectors are extracted from the Laplacian matrix through eigenvalue decomposition, and the feature vectors are clustered to obtain a preset number of examinee partitions. Based on the preset number of subject partitions, the hyperedge cutting cost and partition size balance between each partition are calculated. The hyperedge cutting cost and partition size balance are optimized by iteratively moving vertices to obtain the optimized subject partitioning results. Based on the optimized subject partitioning results, for each subject in each partition, a subset of departments within that partition is extracted. The six-dimensional rule score vector is applied to each department in the subset of departments to calculate the score. When the hyperedge involved in the department to be examined is cut off, a hyperedge penalty value is applied to the comprehensive priority score of the department. The department with the highest comprehensive priority score after deducting the penalty is selected to obtain the sub-scheduling scheme for each partition. Based on the sub-scheduling schemes of each partition, the cut-off hyperedges are identified, cross-partition order constraints are forcibly added to the cut-off hyperedges of the preceding dependency type, different time slots are allocated to the cut-off hyperedges of the resource conflict type, the sub-scheduling schemes of each partition are merged, and the scheduling instruction is generated.

11. A hospital physical examination guidance and scheduling system based on multi-rule intelligent fusion, characterized in that, include: The data acquisition module is used to acquire information about the items to be examined by the examinee, the real-time number of people queuing in each department and the average service time, as well as the current status of the equipment. The module categorizes and stores the information about the items to be examined, the real-time number of people queuing, the average service time and the current status of the equipment to obtain structured system status data. The rule calculation module is used to calculate, based on the structured system status data, the fasting priority score, the same region priority score, the regional flow order score, the empty queue priority score, the pre-examination item score, and the shortest time priority score for each department to be examined, thereby obtaining a six-dimensional rule score vector; The weight adjustment module is used to determine the dynamic weight coefficient of each rule based on the six-dimensional rule score vector and the current passenger flow status, and to perform a weighted summation calculation on the six-dimensional rule score vector to obtain the comprehensive priority score of each department to be inspected; The scheduling decision module is used to select the department with the highest comprehensive priority score as the recommended department based on the comprehensive priority score of each department to be inspected, and generate a scheduling instruction containing the identifier of the recommended department and the estimated waiting time; The guidance and push module is used to push guidance information to the examinee through a multimedia terminal based on the scheduling instruction, thereby completing the physical examination guidance and scheduling.

Images