Hospital registration reservation data processing method

Abstract

The invention relates to the technical field of medical data processing, and discloses a hospital registration reservation data processing method, and the method comprises the following steps: collecting multi-mode hospital registration reservation demand data and real-time number source inventory data, and generating a registration demand-number source inventory hybrid matrix; performing multi-dimensional registration demand feature extraction and number source dynamic matching degree analysis based on the registration demand-number source inventory hybrid matrix, and generating a registration demand feature vector set and number source matching deviation coefficient matrix; when hospital registration appointment data processing is carried out, through multi-dimensional feature extraction and dynamic matching degree analysis, the feature deviation between the registration demand and the number source inventory is sensed in real time, errors caused by inaccurate feature extraction in the matching process can be avoided, the accuracy and fairness of a number source distribution scheme are ensured, and the matching errors are reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical data processing, in particular to a hospital registration appointment data processing method. BACKGROUND

[0002] When going to a hospital, a patient needs to go through the processes of registration, diagnosis, charging and medicine taking. With the gradual increase of population, the long queue for registration and long waiting time in large hospitals cause the congestion of the hospital. With the development of information network technology, many hospitals have proposed network appointment registration services. The appointment registration is a convenient service for medical institutions to optimize the medical process, which is mainly realized through telephone, network or on-site channels.

[0003] At present, in the process of hospital registration appointment data processing, multi-source heterogeneous data and complex scheduling requirements are involved. When the registration demand is matched with the number source in real time, the matching algorithm cannot dynamically perceive the multi-dimensional feature deviation between the number source inventory and the patient demand. When there is a feature extraction deviation and a matching degree calculation error in the matching process, it may cause number source allocation conflict and matching precision decline, and cannot guarantee the accuracy and fairness of the registration appointment.

[0004] Therefore, the present application provides a hospital registration appointment data processing method to solve the above problems. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a hospital registration appointment data processing method, which solves the problems in the background art.

[0006] To achieve the above purpose, the present application provides the following technical scheme: a hospital registration appointment data processing method, the method comprising the following steps:

[0007] S1, collect multi-modal hospital registration appointment demand data and real-time number source inventory data, and generate a registration demand-number source inventory hybrid matrix;

[0008] S2, based on the registration demand-number source inventory hybrid matrix, perform multi-dimensional registration demand feature extraction and number source dynamic matching degree analysis, and generate a registration demand feature vector set and a number source matching deviation coefficient matrix;

[0009] S3, according to the registration demand feature vector set and the number source matching deviation coefficient matrix, perform intelligent triage processing of the registration demand, and generate an initial screening triage queue and an urgency calibration parameter set;

[0010] S4, based on the initial screening triage queue and the urgency calibration parameter set, perform number source adaptive allocation and conflict detection, and generate a number source pre-allocation scheme and a conflict detection report;

[0011] S5, when the conflict detection report shows no conflict, directly perform S7 step;

[0012] S6, when the conflict detection report shows that there is a conflict, intelligent resolution of the number source conflict and priority redistribution processing is performed, and a number source redistribution scheme and a conflict resolution log are generated;

[0013] S7, based on the number source redistribution scheme, registration reservation confirmation and patient notification processing are performed, and registration confirmation data and reservation success notification set are generated;

[0014] S8, according to the registration confirmation data, the number source state real-time update and the balance monitoring are performed, and the number source state update log and the balance early warning signal are generated;

[0015] S9, based on the balance early warning signal, the number pool dynamic expansion and elastic scheduling processing are performed, and the number pool expansion scheme and the scheduling execution instruction are generated;

[0016] S10, the registration reservation whole process monitoring data is constructed and the registration efficiency and satisfaction are evaluated, and the registration efficiency evaluation report and the user satisfaction index are generated.

[0017] Preferably, the S1 multi-modal hospital registration reservation demand data and real-time number source inventory data collection includes the following steps:

[0018] S11, collecting patient registration demand data through hospital registration system interface and third party reservation platform data interface, including patient basic information, medical history label, reservation department, time preference, emergency level identification, generating patient registration demand data matrix;

[0019] S12, real-time extraction of department number source inventory data from hospital number source management system, including number source type, appointmentable time slice, number source quantity, number source state, generating number source inventory data matrix;

[0020] S13, aligning and fusing the patient registration demand data matrix and the number source inventory data matrix according to department-time dimension, constructing a registration demand-number source inventory hybrid matrix, wherein the mathematical representation of the hybrid matrix is:

[0021] ;

[0022] Wherein the registration demand-number source inventory hybrid matrix, is the patient registration demand data matrix, is the number source inventory data matrix.

[0023] Preferably, the S2 multi-dimensional registration demand feature extraction and number source dynamic matching degree analysis includes the following steps:

[0024] S21. Perform multi-dimensional feature extraction on the hybrid matrix of registration demand and appointment availability, and use a convolutional autoencoder to extract patient demand features and appointment availability features to generate demand feature vectors and appointment availability feature vectors.

[0025] S22. Calculate the cosine similarity and Euclidean distance between the demand feature vector and the source feature vector, and construct the matching degree matrix.

[0026] S23. Based on the matching degree matrix, perform deviation analysis to generate a source number matching deviation coefficient matrix, where the formula for calculating the deviation coefficient is:

[0027] ;

[0028] in For the source matching deviation coefficient matrix, For the sample size, For the first A patient demand feature vector For the first Each source feature vector This indicates the sample index.

[0029] Preferably, the intelligent triage process for registration requests in S3 includes the following steps:

[0030] S31. Input the demand feature vector and the number source matching deviation coefficient matrix into the pre-trained triage neural network model, and output the initial screening triage queue.

[0031] S32. Extract urgency features based on patient medical history tags and real-time physiological data, calculate urgency weights using an attention mechanism, and generate an urgency calibration parameter set.

[0032] S33. Merge the initial screening triage queue with the urgency calibration parameter set to generate a weighted triage queue.

[0033] Preferably, the adaptive allocation and conflict detection of the signal source in S4 includes the following steps:

[0034] S41. Based on the weighted triage queue, adaptive allocation of appointment slots is performed. Greedy algorithm and genetic algorithm are used to perform multi-objective optimization matching of appointment slots and patients to generate appointment slot pre-allocation scheme.

[0035] S42. Perform conflict detection on the appointment pre-allocation scheme, including time conflicts, department conflicts, and doctor load conflicts, and generate a conflict detection report.

[0036] Preferably, the intelligent resolution and priority reallocation of number source conflicts in S6 includes the following steps:

[0037] S61. When the conflict detection report shows that a conflict exists, extract the conflict type and conflict node, and generate a conflict feature vector.

[0038] S62. Use a conflict resolution model based on reinforcement learning to process the conflict feature vectors and output a set of conflict resolution strategies.

[0039] S63. Based on the conflict resolution strategy set, prioritize the pre-allocation scheme of the number source and reassign it, and generate the number source reassignment scheme and conflict resolution log.

[0040] Preferably, the registration appointment confirmation and patient notification processing in S7 includes the following steps:

[0041] S71. Send the appointment reallocation plan generated after conflict resolution to the hospital's registration system for appointment confirmation and generate registration confirmation data.

[0042] S72. Send appointment success notifications to patients through multiple communication channels and generate a notification set.

[0043] Preferably, the real-time update of the number source status and the monitoring of remaining capacity in S8 includes the following steps:

[0044] S81. Monitor the changes in the status of available appointments in the registration confirmation data in real time and update the available appointment inventory matrix.

[0045] S82. Based on the updated inventory matrix, perform inventory level detection. If the inventory level is lower than the threshold, generate an inventory level warning signal.

[0046] Preferably, the dynamic expansion and elastic scheduling processing of the S9 medium-sized pool includes the following steps:

[0047] S91. Based on the reserve warning signal, initiate the dynamic expansion strategy of the number pool, use a linear programming model to calculate the optimal expansion scheme, and generate the number pool expansion scheme.

[0048] S92. Use the elastic scheduling algorithm to schedule idle appointment slots and cross-department appointment slots, and generate scheduling execution instructions.

[0049] Preferably, the registration efficiency and satisfaction evaluation in S10 includes the following steps:

[0050] S101. Collect data from the entire registration process, including triage time, matching accuracy, conflict resolution efficiency, and patient feedback, and construct a monitoring data matrix.

[0051] S102. Use the data envelopment analysis model to evaluate registration efficiency and generate a registration efficiency evaluation report.

[0052] S103. Process patient feedback text through a sentiment analysis model to generate a user satisfaction index.

[0053] Compared with the prior art, the present invention provides a method for processing hospital registration and appointment data, which has the following beneficial effects:

[0054] 1. In this invention, when processing hospital registration and appointment data, multi-dimensional feature extraction and dynamic matching degree analysis are used to perceive the feature deviation between registration demand and appointment availability in real time. This can avoid errors caused by inaccurate feature extraction during the matching process, ensure the accuracy and fairness of appointment allocation schemes, and reduce matching errors.

[0055] 2. In this invention, when processing hospital registration and appointment data, a real-time conflict detection and intelligent resolution mechanism is used to dynamically identify time conflicts, departmental conflicts, and resource overload issues in appointment allocation. An adaptive algorithm is used to generate resolution strategies in real time, enabling the system to quickly correct allocation anomalies, avoid situations where appointment allocation schemes are not feasible, and ensure the smoothness of the registration process.

[0056] 3. In this invention, when processing hospital registration and appointment data, a multi-objective collaborative optimization and elastic scheduling method is used to monitor the status of the appointment pool and changes in patient demand in real time, dynamically adjust the appointment expansion strategy and scheduling instructions, achieve efficient utilization of appointment resources and balance of multiple objectives, improve appointment utilization rate and patient satisfaction, and enhance the overall service efficiency and quality of the hospital registration system. Attached Figure Description

[0057] Figure 1 This is a flowchart of a hospital registration and appointment data processing method according to the present invention. Detailed Implementation

[0058] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] For specific implementation examples, please refer to: Figure 1 A method for processing hospital registration and appointment data, the method includes the following steps:

[0060] S1. Collect multimodal hospital registration and appointment demand data and real-time appointment inventory data to generate a hybrid matrix of registration demand and appointment inventory;

[0061] S2. Based on the hybrid matrix of registration demand and appointment availability, perform multi-dimensional feature extraction of registration demand and dynamic matching degree analysis of appointment availability to generate a feature vector set of registration demand and a matching deviation coefficient matrix of appointment availability.

[0062] S3. Based on the feature vector set of registration demand and the deviation coefficient matrix of appointment source matching, intelligent triage of registration demand is performed to generate a preliminary screening triage queue and an urgency calibration parameter set.

[0063] S4. Based on the initial screening and triage queue and the urgency calibration parameter set, perform adaptive allocation of appointment slots and conflict detection, and generate appointment slot pre-allocation scheme and conflict detection report;

[0064] S5. When the conflict detection report shows no conflict, proceed directly to step S7;

[0065] S6. When the conflict detection report shows that there is a conflict, perform intelligent conflict resolution and priority reallocation of the number source, and generate a number source reallocation scheme and conflict resolution log.

[0066] S7. Based on the appointment allocation scheme, perform appointment confirmation and patient notification processing, and generate appointment confirmation data and appointment success notification sets.

[0067] S8. Based on the registration confirmation data, update the status of appointment slots and monitor the availability in real time, and generate appointment slot status update logs and availability warning signals.

[0068] S9. Based on the reserve warning signal, perform dynamic expansion and elastic scheduling of the number pool to generate a number pool expansion plan and scheduling execution instructions.

[0069] S10. Build monitoring data for the entire registration and appointment process and conduct registration efficiency and satisfaction assessments to generate a registration efficiency assessment report and a user satisfaction index.

[0070] The collection of multimodal hospital registration and appointment demand data and real-time appointment availability data in S1 includes the following steps:

[0071] S11. Collect patient registration demand data through the hospital registration system interface and the data interface of the third-party appointment platform, including patient basic information, medical history tags, appointment department, time preference, urgency level indicator, and generate a patient registration demand data matrix;

[0072] S12. Extract real-time appointment inventory data for each department from the hospital appointment management system, including appointment type, available appointment time slots, number of appointments, and appointment status, and generate an appointment inventory data matrix.

[0073] S13. Align and merge the patient registration demand data matrix and the appointment availability inventory data matrix according to the department-time dimension to construct a hybrid matrix of registration demand and appointment availability inventory. The mathematical representation of the hybrid matrix is ​​as follows:

[0074] ;

[0075] in Registration demand - appointment availability inventory hybrid matrix A data matrix of patient registration needs. This is a matrix of inventory data for available license plates.

[0076] The multi-dimensional registration demand feature extraction and dynamic matching degree analysis of appointment sources in S2 include the following steps:

[0077] S21. Perform multi-dimensional feature extraction on the hybrid matrix of registration demand and appointment availability. Use a convolutional autoencoder to extract patient demand features and appointment availability features, generating demand feature vectors and appointment availability feature vectors. This includes the following steps:

[0078] S211. Construct a convolutional autoencoder network structure, including an input layer, a convolutional layer, a pooling layer, an encoding layer, a decoding layer, and an output layer. The input layer receives a mixed matrix of registration demand and number source inventory as input data.

[0079] S212. The input data is feature-mapped through convolutional layers, and the convolutional output is processed using the ReLU activation function to generate a feature map.

[0080] S213. Downsample the feature map in the pooling layer, and use max pooling to reduce the data dimensionality while retaining key features;

[0081] S214. In the coding layer, the downsampled features are compressed into a low-dimensional vector to generate the required feature vector and the source feature vector.

[0082] S215. The low-dimensional vector is upsampled and deconvolved through the decoding layer to reconstruct the output data, and the mean squared error loss function is used to optimize the network parameters to ensure the accuracy of feature extraction.

[0083] S22. Calculate the cosine similarity and Euclidean distance between the demand feature vector and the source feature vector, and construct the matching degree matrix. This includes the following steps:

[0084] S221. Normalize the demand feature vector and the signal source feature vector so that the vector magnitude is 1.

[0085] S222. Calculate cosine similarity:

[0086] ;

[0087] in Represents the cosine similarity score. Represents the demand feature vector. The symbol represents the source feature vector, and the dot sign represents the dot product operation. Indicates the magnitude of the vector;

[0088] S223. Calculate the Euclidean distance:

[0089] ;

[0090] in Indicates Euclidean distance;

[0091] S224, Cosine similarity Euclidean distance Combined into a matching degree matrix:

[0092] ;

[0093] in Represents the matching degree matrix;

[0094] S23. Based on the matching degree matrix, perform deviation analysis to generate a source number matching deviation coefficient matrix, where the formula for calculating the deviation coefficient is:

[0095] ;

[0096] in For the source matching deviation coefficient matrix, For the sample size, For the first A patient demand feature vector For the first Each source feature vector This indicates the sample index.

[0097] The intelligent triage process for registration requests in S3 includes the following steps:

[0098] S31. Input the demand feature vector and the number source matching deviation coefficient matrix into the pre-trained triage neural network model to output the initial screening triage queue. This includes the following steps:

[0099] S311. Collect historical registration data as a training set, including patient demand feature vector, appointment source matching deviation coefficient matrix and corresponding triage result labels;

[0100] S312. Construct a triage neural network model. The model structure includes an input layer, a hidden layer, and an output layer. The input layer receives the demand feature vector and the number source matching deviation coefficient matrix. The hidden layer uses a fully connected layer and a Dropout layer to prevent overfitting. The output layer outputs the initial screening triage queue.

[0101] S313. The model is trained using the backpropagation algorithm, with the weights optimized using the cross-entropy loss function. The learning rate is 0.001, and the number of iterations is 1000. The training of the model using the backpropagation algorithm includes the following steps:

[0102] S3131. Initialize the weight parameters of the triage neural network model. Use the Xavier initialization method to set the initial weights and initialize the bias term to zero.

[0103] S3132, The output of the forward propagation calculation model, based on the input demand feature vector and the number source matching deviation coefficient matrix, is used to calculate and predict the triage queue through the hidden layer.

[0104] S3133. Calculate the cross-entropy loss function, the formula is:

[0105] ;

[0106] in Indicates the loss value. Indicates the true triage label, This indicates the predicted triage probability. Indicates the sample index;

[0107] S3134. Calculate the gradient using the backpropagation algorithm, and update the weights using the gradient descent method. The weight update formula is:

[0108] ;

[0109] in Represents the weight parameters. Indicates the learning rate. Represents the gradient of the loss function;

[0110] S314. Verify model accuracy by using K-fold cross-validation to ensure the triage queue accuracy is higher than 95%, including the following steps:

[0111] S3141. Randomly divide the historical registration training dataset into... There are 3 subsets, with K = 5, to ensure that the data in each subset is evenly distributed;

[0112] S3142. Use each subset as the validation set in turn, and the rest... -1 subsets are used as the training set. The triage neural network model is trained using the training set, and the accuracy is calculated on the validation set.

[0113] S3143, Calculate all The average accuracy of the validation trials is calculated using the following formula:

[0114] ;

[0115] in This represents the average accuracy rate. Indicates the number of folds. Indicates the first The accuracy of the second verification;

[0116] S3144. Ensure the average accuracy is higher than 95%. If it is lower than the threshold, adjust the model parameters and retrain until the accuracy requirement is met.

[0117] S32. Extract urgency features based on patient medical history tags and real-time physiological data, calculate urgency weights using an attention mechanism, and generate an urgency calibration parameter set. This includes the following steps:

[0118] S321. Extract patient medical history tags and real-time physiological data to construct an urgency feature vector;

[0119] S322. In the attention mechanism, calculate the attention scores of the query vector, key vector, and value vector;

[0120] S323. Normalize the attention scores using the Softmax function to generate attention weights;

[0121] S324. Sum the attention weights and the value vectors in a weighted manner to output the urgency calibration parameter set, where:

[0122] ;

[0123] in This represents the set of urgency calibration parameters. Indicates attention weights, Represents the urgency feature vector. This represents the summation operation. Indicates multiplication operation;

[0124] S33. Merge the initial screening triage queue with the urgency calibration parameter set to generate a weighted triage queue.

[0125] S4 source adaptive allocation and conflict detection includes the following steps:

[0126] S41. Based on the weighted triage queue, adaptive allocation of appointment slots is performed. A greedy algorithm and a genetic algorithm are used for multi-objective optimization matching of appointment slots and patients to generate a pre-allocation scheme. This includes the following steps:

[0127] S411. Initialize the patient-appointment matching population, where each individual represents a matching scheme, and the objective function is to maximize the matching degree and minimize the conflict.

[0128] S412. In the greedy algorithm stage, the current best match is selected locally first, and the instant matching score between each appointment source and the patient is calculated.

[0129] S413. In the genetic algorithm stage, selection, crossover, and mutation operations are performed;

[0130] Fitness calculation: The selection operation is based on fitness scores, and the fitness calculation formula is as follows:

[0131] ;

[0132] in This represents the individual's fitness score. Indicates the matching score. Indicates the conflict score. and These are the weighting coefficients;

[0133] Selection Operation: Using the roulette wheel selection method, the individual selection probability formula is:

[0134] ;

[0135] in Indicates the first The probability of an individual being selected. Indicates population size, An index representing an individual in the population. Indicates the summation index;

[0136] Crossover operation: Single-point crossover is used, and the crossover point is randomly selected to exchange matching pairs;

[0137] Mutation operation: Randomly adjust the matching pairs, and set the mutation probability to 0.01;

[0138] S414. Iterate and optimize until convergence, output the signal source pre-assignment scheme, and record the Pareto optimal solution. The convergence condition formula is:

[0139]

[0140] in Indicates the first The best fitness in the next iteration This is the convergence threshold;

[0141] S42. Perform conflict detection on the appointment pre-allocation scheme, including time conflicts, department conflicts, and doctor load conflicts, and generate a conflict detection report.

[0142] The intelligent conflict resolution and priority reallocation process for source numbers in S6 includes the following steps:

[0143] S61. When the conflict detection report shows that a conflict exists, extract the conflict type and conflict node, and generate a conflict feature vector.

[0144] S62. Process the conflict feature vectors using a reinforcement learning-based conflict resolution model to output a conflict resolution strategy set, specifically including the following steps:

[0145] S621. Define the reinforcement learning environment: the state space is the conflict feature vector, the action space is the conflict resolution strategy, and the reward function is based on the conflict resolution efficiency and patient satisfaction.

[0146] S622. The model is trained using the Q-learning algorithm, and the Q-value update formula is:

[0147] ;

[0148] in Represents a state-action value function. Indicates state, Let α represent the action and α represent the learning rate. γ represents the immediate reward, γ represents the discount factor, and s' represents the next state. Indicates the next action. This represents the maximum value operation;

[0149] S623. Explore the action space using an ε-greedy strategy, iterate training until the Q-value stabilizes, and output a set of conflict resolution strategies.

[0150] S63. Based on the conflict resolution strategy set, prioritize the pre-allocation scheme of the number source and reassign it, and generate the number source reassignment scheme and conflict resolution log.

[0151] The registration and appointment confirmation and patient notification process in S7 includes the following steps:

[0152] S71. Send the appointment reallocation plan generated after conflict resolution to the hospital's registration system for appointment confirmation and generate registration confirmation data.

[0153] S72. Send appointment success notifications to patients through multiple communication channels and generate a notification set.

[0154] Real-time updates and remaining capacity monitoring of S8's inventory status include the following steps:

[0155] S81. Monitor the changes in the status of available appointments in the registration confirmation data in real time and update the available appointment inventory matrix.

[0156] S82. Based on the updated inventory matrix, perform inventory level detection. If the inventory level is lower than the threshold, generate an inventory level warning signal.

[0157] The dynamic expansion and elastic scheduling of the S9 medium-sized pool includes the following steps:

[0158] S91. Based on the reserve warning signal, initiate the dynamic expansion strategy of the number pool, calculate the optimal expansion scheme using a linear programming model, and generate the number pool expansion scheme, specifically including the following steps:

[0159] S911. Construct a linear programming objective function: minimize the cost of expanding the registration pool, with constraints including surplus warning signals, upper limit of the number of registration sources, and departmental load balancing.

[0160] S912. Define the decision variable as the number of appointment slots expanded in each department, and the objective function is:

[0161] ;

[0162] in This represents the minimization operation. This represents the expansion cost coefficient. Represent decision variables;

[0163] S913. Use the simplex method to solve the linear programming problem, generate a solution for expanding the number pool, and verify the feasibility of the solution, including the following steps:

[0164] S9131. Construct a simplex tableau: Transform the linear programming objective function and constraints into a standard form. The constraints include the lower limit of the number of appointment slots corresponding to the margin warning signal, the upper limit of the number of appointment slots, and the departmental load balancing limit.

[0165] The constraints are formalized as follows:

[0166] ;

[0167] in Represents the constraint coefficient. This represents the constant on the right-hand side of the constraint condition. As decision variables, Indicates the constraint index. Indicates the index of the decision variable;

[0168] S9132. Introduce slack variables to transform inequality constraints into equality constraints, generate an initial simplex tableau, and initialize basic variables as slack variables;

[0169] S9133, Iterative optimization process:

[0170] Calculate the test number:

[0171] ;

[0172] in The test number represents the number of non-basic variables. The coefficients of the objective function, Indicates the index of the basic variable. Indicates the index of a non-basic variable;

[0173] Select the variable corresponding to the largest positive test number as the input variable. When all test numbers are non-regular, the iteration terminates. Determine the basic variables according to the minimum ratio rule and generate a new basic feasible solution.

[0174] S9134. Verify the feasibility of the solution: Check whether the final solution satisfies all constraints and ensure that the expansion scheme of the No. 1 pool meets the department load balancing requirements and minimizes the objective function value.

[0175] S92. The idle appointment slots and cross-department appointment slots are scheduled using an elastic scheduling algorithm, generating scheduling execution instructions. This includes the following steps:

[0176] S921. Monitor the status of available appointment slots and cross-departmental appointment slots, and construct a scheduling graph model, where nodes represent appointment slots and edges represent scheduling paths;

[0177] S922. Use Dijkstra's algorithm to calculate the shortest scheduling path, with the weights being scheduling time and resource consumption;

[0178] S923: Dynamically adjust scheduling instructions, prioritize the allocation of idle numbers, and ensure a balanced surplus.

[0179] The S10 assessment of registration efficiency and satisfaction includes the following steps:

[0180] S101. Collect data from the entire registration process, including triage time, matching accuracy, conflict resolution efficiency, and patient feedback, and construct a monitoring data matrix.

[0181] S102. Use the Data Envelopment Analysis (DEA) model to evaluate registration efficiency and generate a registration efficiency evaluation report, which includes the following steps:

[0182] S1021. Define input and output variables, and construct a data envelopment analysis model. Input variables include triage time and matching accuracy, and output variable is registration efficiency.

[0183] S1022. The efficiency score is calculated using the CCR model, and the formula is as follows:

[0184] ;

[0185] S1023. Solve for the efficiency value using linear programming, generate a registration efficiency evaluation report, and identify invalid cells;

[0186] S103. Process patient feedback text using a sentiment analysis model to generate a user satisfaction index, specifically including the following steps:

[0187] S1031. Collect patient feedback text data and perform preprocessing, including word segmentation, stop word removal, and stemming.

[0188] S1032. Extract feature vectors from the preprocessed text and construct a sentiment dictionary;

[0189] S1033. Train the sentiment analysis model through a classifier and output the user satisfaction index. The accuracy threshold is set to a predetermined value based on the model performance.

[0190] The operational steps of this hospital registration and appointment data processing method are as follows:

[0191] Step 1: Multimodal Data Acquisition and Hybrid Matrix Construction

[0192] Patient registration demand data is collected through the hospital's registration system interface and third-party appointment platform data interface, including basic patient information, medical history tags, appointment department, time preference, and urgency level indicators, generating a patient registration demand data matrix. At the same time, appointment availability data for each department is extracted in real time from the hospital's appointment management system, including appointment type, available time slots, number of appointments, and appointment status, generating an appointment availability data matrix. The two types of data matrices are aligned and merged according to the department-time dimension to construct a hybrid matrix of registration demand and appointment availability, laying the data foundation for subsequent feature extraction.

[0193] Step 2: Multi-dimensional feature extraction and dynamic matching degree analysis

[0194] Based on the hybrid matrix of registration demand and appointment availability, a convolutional autoencoder is used to extract multi-dimensional features, generating demand feature vectors and appointment feature vectors respectively. By calculating cosine similarity and Euclidean distance, a matching degree matrix is ​​constructed, and deviation analysis is performed based on the matching degree matrix to generate an appointment matching deviation coefficient matrix. This process realizes a quantitative evaluation of the dynamic matching relationship between registration demand and appointment availability, providing data support for intelligent triage.

[0195] Step 3: Intelligent Triage and Urgency Level Assessment

[0196] The demand feature vector and the appointment source matching deviation coefficient matrix are input into the pre-trained triage neural network model, which outputs the initial screening triage queue. At the same time, based on the patient's medical history tags and real-time physiological data, urgency features are extracted, and urgency weights are calculated through an attention mechanism to generate an urgency calibration parameter set. The initial screening triage queue and the urgency calibration parameter set are merged to generate a weighted triage queue, thereby achieving scientific ranking of patients by priority.

[0197] Step 4: Adaptive Allocation of Numbers and Conflict Detection and Resolution

[0198] Based on the weighted triage queue, appointment slots are adaptively allocated. Greedy algorithm and genetic algorithm are used for multi-objective optimization matching of appointment slots and patients to generate appointment slot pre-allocation schemes. Conflict detection is performed on the pre-allocation schemes, including time conflicts, department conflicts, and doctor load conflicts, and a conflict detection report is generated. When a conflict is detected, the conflict feature vector is processed by a conflict resolution model based on reinforcement learning, outputting a set of conflict resolution strategies. Priority reallocation is then performed on the appointment slot pre-allocation schemes to generate the final appointment slot reallocation scheme.

[0199] Step 5: Monitoring and Optimizing License Plate Status and Flexible Scheduling

[0200] The system monitors changes in the status of appointment slots in the registration confirmation data in real time, updates the appointment slot inventory matrix, and initiates a dynamic expansion strategy for the appointment pool based on the surplus warning signal. It calculates the optimal expansion scheme through a linear programming model and combines it with an elastic scheduling algorithm to dynamically schedule idle appointment slots and cross-department appointment slots, generating scheduling execution instructions. This process realizes real-time optimization and configuration of appointment slot resources, ensuring maximum utilization of appointment slots.

[0201] Step Six: Full-Process Monitoring and Multi-Dimensional Evaluation

[0202] A data matrix for monitoring the entire registration and appointment process is constructed, collecting multi-dimensional data such as triage time, matching accuracy, conflict resolution efficiency, and patient feedback. A data envelopment analysis model is used to evaluate registration efficiency and generate a registration efficiency evaluation report. At the same time, a sentiment analysis model is used to process patient feedback text to generate a user satisfaction index. Through continuous monitoring and evaluation, a closed-loop optimization mechanism is formed to continuously improve the system's service level.

[0203] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0204] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for processing hospital registration and appointment data, characterized in that: The method includes the following steps: S1. Collect multimodal hospital registration and appointment demand data and real-time appointment inventory data to generate a hybrid matrix of registration demand and appointment inventory; S2. Based on the hybrid matrix of registration demand and appointment availability, perform multi-dimensional feature extraction of registration demand and dynamic matching degree analysis of appointment availability to generate a feature vector set of registration demand and a matching deviation coefficient matrix of appointment availability. S3. Based on the feature vector set of registration demand and the deviation coefficient matrix of appointment source matching, intelligent triage of registration demand is performed to generate a preliminary screening triage queue and an urgency calibration parameter set. S4. Based on the initial screening and triage queue and the urgency calibration parameter set, perform adaptive allocation of appointment slots and conflict detection, and generate appointment slot pre-allocation scheme and conflict detection report; S5. When the conflict detection report shows no conflict, proceed directly to step S7; S6. When the conflict detection report shows that there is a conflict, perform intelligent conflict resolution and priority reallocation of the number source, and generate a number source reallocation scheme and conflict resolution log. S7. Based on the appointment allocation scheme, perform appointment confirmation and patient notification processing, and generate appointment confirmation data and appointment success notification sets. S8. Based on the registration confirmation data, update the status of appointment slots and monitor the availability in real time, and generate appointment slot status update logs and availability warning signals. S9. Based on the reserve warning signal, perform dynamic expansion and elastic scheduling of the number pool to generate a number pool expansion plan and scheduling execution instructions. S10. Build monitoring data for the entire registration and appointment process and conduct registration efficiency and satisfaction assessments to generate a registration efficiency assessment report and a user satisfaction index.

2. The method for processing hospital registration and appointment data according to claim 1, characterized in that: The collection of multimodal hospital registration and appointment demand data and real-time appointment availability data in S1 includes the following steps: S11. Collect patient registration demand data through the hospital registration system interface and the data interface of the third-party appointment platform, including patient basic information, medical history tags, appointment department, time preference, urgency level indicator, and generate a patient registration demand data matrix; S12. Extract real-time appointment inventory data for each department from the hospital appointment management system, including appointment type, available appointment time slots, number of appointments, and appointment status, and generate an appointment inventory data matrix. S13. Align and merge the patient registration demand data matrix and the appointment availability inventory data matrix according to the department-time dimension to construct a hybrid matrix of registration demand and appointment availability inventory. The mathematical representation of the hybrid matrix is ​​as follows: ； in Registration demand - appointment availability inventory hybrid matrix A data matrix of patient registration needs. This is a matrix of inventory data for available license plates.

3. The hospital registration and appointment data processing method according to claim 1, characterized in that: The multi-dimensional registration demand feature extraction and dynamic matching degree analysis of appointment sources in S2 include the following steps: S21. Perform multi-dimensional feature extraction on the hybrid matrix of registration demand and appointment availability, and use a convolutional autoencoder to extract patient demand features and appointment availability features to generate demand feature vectors and appointment availability feature vectors. S22. Calculate the cosine similarity and Euclidean distance between the demand feature vector and the source feature vector, and construct the matching degree matrix. S23. Based on the matching degree matrix, perform deviation analysis to generate a source number matching deviation coefficient matrix, where the formula for calculating the deviation coefficient is: ； in For the source matching deviation coefficient matrix, For the sample size, For the first A patient demand feature vector For the first Each source feature vector This indicates the sample index.

4. The hospital registration and appointment data processing method according to claim 1, characterized in that: The intelligent triage process for registration requests in S3 includes the following steps: S31. Input the demand feature vector and the number source matching deviation coefficient matrix into the pre-trained triage neural network model, and output the initial screening triage queue. S32. Extract urgency features based on patient medical history tags and real-time physiological data, calculate urgency weights using an attention mechanism, and generate an urgency calibration parameter set. S33. Merge the initial screening triage queue with the urgency calibration parameter set to generate a weighted triage queue.

5. The hospital registration and appointment data processing method according to claim 1, characterized in that: The adaptive allocation and conflict detection of signal sources in S4 includes the following steps: S41. Based on the weighted triage queue, adaptive allocation of appointment slots is performed. Greedy algorithm and genetic algorithm are used to perform multi-objective optimization matching of appointment slots and patients to generate appointment slot pre-allocation scheme. S42. Perform conflict detection on the appointment pre-allocation scheme, including time conflicts, department conflicts, and doctor load conflicts, and generate a conflict detection report.

6. The hospital registration and appointment data processing method according to claim 1, characterized in that: The intelligent conflict resolution and priority reallocation process for number sources in S6 includes the following steps: S61. When the conflict detection report shows that a conflict exists, extract the conflict type and conflict node, and generate a conflict feature vector. S62. Use a conflict resolution model based on reinforcement learning to process the conflict feature vectors and output a set of conflict resolution strategies. S63. Based on the conflict resolution strategy set, prioritize the pre-allocation scheme of the number source and reassign it, and generate the number source reassignment scheme and conflict resolution log.

7. The method for processing hospital registration and appointment data according to claim 1, characterized in that: The registration appointment confirmation and patient notification process in S7 includes the following steps: S71. Send the appointment reallocation plan generated after conflict resolution to the hospital's registration system for appointment confirmation and generate registration confirmation data. S72. Send appointment success notifications to patients through multiple communication channels and generate a notification set.

8. The hospital registration and appointment data processing method according to claim 1, characterized in that: The real-time update and remaining capacity monitoring of the S8 source status includes the following steps: S81. Monitor the changes in the status of available appointments in the registration confirmation data in real time and update the available appointment inventory matrix. S82. Based on the updated inventory matrix, perform inventory level detection. If the inventory level is lower than the threshold, generate an inventory level warning signal.

9. A method for processing hospital registration and appointment data according to claim 1, characterized in that: The dynamic expansion and elastic scheduling process of the S9 medium-sized pool includes the following steps: S91. Based on the reserve warning signal, initiate the dynamic expansion strategy of the number pool, use a linear programming model to calculate the optimal expansion scheme, and generate the number pool expansion scheme. S92. Use the elastic scheduling algorithm to schedule idle appointment slots and cross-department appointment slots, and generate scheduling execution instructions.

10. A method for processing hospital registration and appointment data according to claim 1, characterized in that: The evaluation of registration efficiency and satisfaction in S10 includes the following steps: S101. Collect data from the entire registration process, including triage time, matching accuracy, conflict resolution efficiency, and patient feedback, and construct a monitoring data matrix. S102. Use the data envelopment analysis model to evaluate registration efficiency and generate a registration efficiency evaluation report. S103. Process patient feedback text through a sentiment analysis model to generate a user satisfaction index.

Images