Deep learning-based outpatient service scheduling method and system

Abstract

The invention discloses an outpatient service scheduling method and system based on deep learning. And obtaining outpatient service information of the user. And obtaining a first fusion feature vector based on the user outpatient service information. And through a first prediction network, based on the user outpatient service information and the first fusion feature vector, detecting a time change rule. According to the method, consulting room positions are detected by different users at different time points in a plurality of historical days, and the queuing condition is found by adopting big data, so that a first prediction network capable of estimating the queuing number is found. Detecting a time point of arriving at the position of each user detection consulting room, and finding a queuing number and a to-be-processed transaction number corresponding to the time point through a first prediction network; through multiple times of judgment, deep learning and a greedy algorithm are adopted to jointly judge the shortest time point when a detection report is obtained after queuing and processing and the user detection consulting room position recommended to the user for sequential detection.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an outpatient scheduling method and system based on deep learning. Background Art

[0002] In traditional hospital outpatient care, patients undergo tests such as blood tests and CT scans after their outpatient visit. After obtaining this data, they must seek a second doctor's diagnosis for a more accurate diagnosis. However, due to varying queues in blood collection rooms and CT scan rooms, patients must visit multiple testing clinics for multiple tests. This makes it difficult to flexibly adjust the system based on actual patient needs and medical resource usage. Hospital managers struggle to accurately monitor the availability of each testing clinic in real time, leading to crowded areas in some clinics and nearly empty spaces in others. This results in long wait times for patients seeking a second diagnosis, resulting in a poor medical experience. This inefficient use of hospital resources makes it difficult to meet the growing demand for medical services. Therefore, there is an urgent need for a method and system that can intelligently plan routes to testing clinics based on real-time data. Summary of the Invention

[0003] The purpose of the present invention is to provide an outpatient scheduling method and system based on deep learning to solve the above-mentioned problems existing in the prior art.

[0004] In a first aspect, an embodiment of the present invention provides an outpatient scheduling method based on deep learning, comprising: Obtaining user clinic information; the user clinic information includes the user's current location and m user detection clinic locations; the user detection clinic location indicates the location of the clinic where the user needs to undergo examination; Based on the user clinic information, a first fused feature vector is obtained; the first fused feature vector represents the characteristics of the location and time of the user's arrival at the m user detection clinic locations; Through the first prediction network, based on the user clinic information and the first fusion feature vector, the time change pattern is detected to obtain a recommended user clinic vector and a planned time length; the recommended user clinic vector represents the user clinic location recommended to the user for sequential detection; the planned time length represents the shortest waiting time after arriving at the user clinic location in the recommended user clinic vector.

[0005] Optionally, obtaining a first fusion feature vector based on the user outpatient information includes: Based on the current location of the user and the locations of the detection clinics of multiple users, a first connectivity matrix, a second connectivity vector and a third connectivity vector are obtained; Based on the first connectivity matrix, the second connectivity vector and the third connectivity vector, features are extracted to obtain a first fused feature vector; the first fused feature vector represents features that fuse geographic location information, processing status and queuing status.

[0006] Optionally, obtaining the first connectivity matrix, the second connectivity vector, and the third connectivity vector based on the current location of the user and the locations of the detection clinics of multiple users includes: Obtain the number of pending transactions, processing speed, queue number, and queue speed; the number of pending transactions represents the number of pending transactions in a user detection clinic location; the processing speed represents the speed of processing transactions in a user detection clinic location; the queue number represents the number of people queuing in a user detection clinic location; and the queue speed represents the speed of queuing in a user detection clinic location; Filling the distance between the user's current location and the locations of the m user detection clinics into a first connectivity matrix; the rows of the first connectivity matrix include the user's current location and the locations of the m user detection clinics; the columns of the first connectivity matrix include the user's current location and the locations of the m user detection clinics; The number of pending transactions at the m user detection clinic locations is divided by the corresponding processing speed to obtain a second connectivity vector; The number of queues at the m user detection clinic locations is divided by the corresponding queue speed to obtain a third connectivity vector.

[0007] Optionally, the training method of the first prediction network includes: Acquire a training queue data set for multiple days; the training queue data set includes the number of training queues for multiple testing clinic locations at n queuing time points; the training queue number represents the number of people queuing at one testing clinic location at one time point; n represents 24 hours divided into n equal parts; the testing clinic locations represent the locations of all clinics in the hospital; Based on the detection clinic location and the number of training queues, a training fusion feature vector is obtained; Construct a queue detection vector based on 1 testing clinic location, n / 12 queue time points, and the corresponding number of training queues; the number of elements in the queue detection vector is 1+n / 12+n / 12; Inputting the queue detection vector and the training fusion feature vector into a first prediction network to obtain a predicted queue number; the predicted queue number represents the number of people queuing at the next queue time point at the same testing clinic location; One testing clinic location corresponds to one first prediction network; The training queue data at the next queue time point is compared with the predicted queue quantity to obtain the loss, and back propagation is performed to train the first prediction network.

[0008] Optionally, the training queue data at the next queue time point is compared with the predicted queue quantity to obtain a loss, and the first convolutional neural network, the first deep neural network, the second deep neural network, and the third deep neural network are trained simultaneously.

[0009] Optionally, the detecting time variation pattern by the first prediction network based on the user clinic information and the first fusion feature vector to obtain the recommended user clinic vector and the planned time length includes: Obtain user speed, current time point, and multiple current queue numbers; the user speed represents the user's walking speed; the current time point represents the time point corresponding to the user's current location; the current queue number represents the number of people queuing at a user testing clinic location at the current time point; Obtaining, through the first prediction network, m first waiting time lengths, m first detection time points, and m first predicted queue sizes based on the user speed, the current time point, the m current queue sizes, and the m queue speeds; Obtaining, through the first prediction network, m-1 second waiting time lengths, m-1 second detection time points, and m-1 second predicted queue sizes based on the user speed, the first detection time point, m-1 first predicted queue sizes, and m-1 queue speeds; Traverse all user detection clinic locations to obtain 1*2*3…(m-1)*m user routes and corresponding processing time lengths; the user route represents a route passing through all user detection clinic locations; the processing time length represents the sum of the first waiting time length, the second waiting time length, ..., the mth waiting time length on a user route; The processing time length that is shorter than other processing time lengths is used as the planning time length, and the user detection clinic positions corresponding to the user routes corresponding to the planning time lengths are sequentially used as elements in the recommended user clinic vector.

[0010] Optionally, obtaining, by the first prediction network, m first waiting time lengths, m first detection time points, and m first predicted queue numbers based on the user speed, the current time point, m current queue numbers, and m queue speeds, includes: Divide the distance from the user's current position in the first connectivity matrix to the locations of the m user's detection clinics by the user's speed to obtain m first time lengths; the first time length represents the time length for the user's current position to reach the location of one user's clinic; Based on the user's detection clinic location, the current time point, and the corresponding current queue number, m first predicted permutation numbers are obtained; Divide the first predicted queue number by the queue speed to obtain the first queue time length; Add the current time point to the first time length and the first arrangement time length to obtain the first detection time point; m user clinic locations correspond to m first detection time points; The number of pending transactions at the first detection point is obtained and divided by the corresponding processing speed to obtain the first waiting time length; m first detection time points correspond to m first waiting time lengths.

[0011] Optionally, obtaining m first predicted permutation numbers based on the user's detected clinic location, the current time point, and the corresponding current queue number includes: The current queue detection vector is constructed by combining the user's detection clinic location, current time point and the corresponding current queue number; According to the current queue detection vector, a first predicted queue number is obtained through a trained first prediction network; the first predicted queue number represents the number of queues at a time point after a first time length from the current time point; m user clinic locations correspond to m first predicted arrangement quantities.

[0012] Optionally, extracting features based on the first connectivity matrix, the second connectivity vector, and the third connectivity vector to obtain a first fused feature vector includes: Inputting the first connectivity matrix into the first convolutional neural network, extracting features, and obtaining a first feature map; Inputting the first feature map into a first deep neural network to extract features and obtain a first feature vector; Inputting the first connectivity matrix into a second deep neural network, extracting features, and obtaining a second eigenvector; Inputting the first connectivity matrix into a third deep neural network to extract features and obtain a third eigenvector; The first eigenvector, the second eigenvector and the third eigenvector have the same number of elements; The first eigenvector, the second eigenvector, and the third eigenvector are averaged by their corresponding subscripts to obtain a first fused eigenvector.

[0013] In a second aspect, an embodiment of the present invention provides an outpatient scheduling system based on deep learning, comprising: An acquisition module is used to acquire user clinic information; the user clinic information includes the user's current location and m user detection clinic locations; the user detection clinic location indicates the location of the clinic where the user needs to undergo examination; A feature module is configured to obtain a first fused feature vector based on the user's clinic information; the first fused feature vector represents the characteristics of the location and time of the user's arrival at the locations of the m user detection clinics; The detection module is used to detect the time change pattern based on the user clinic information and the first fusion feature vector through the first prediction network, and obtain the recommended user clinic vector and the planned time length; the recommended user clinic vector represents the user clinic location recommended to the user for sequential detection; the planned time length represents the shortest waiting time after arriving at the user clinic location in the recommended user clinic vector.

[0014] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects: The embodiments of the present invention also provide an outpatient scheduling method and system based on deep learning.

[0015] In the present invention, the time it takes for a user to undergo testing is estimated based on the number of pending transactions, the number of queues, the processing speed, and the queuing speed at different user testing clinic locations at different time points. By using the different user testing clinic locations at different time points in multiple historical days, big data is used to find the queuing situation, thereby finding a first prediction network that can estimate the number of queues. The time point at which each user's testing clinic location is detected is used, and then the number of queues and the number of pending transactions corresponding to this time point are found. By making multiple judgments, deep learning and greedy algorithms are used to jointly determine the shortest time point for obtaining a test report after queuing and processing. This can achieve the technical effect of saving waiting time for testing. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flow chart of an outpatient scheduling method based on deep learning provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The present invention will be described in detail below with reference to the accompanying drawings.

[0018] Example 1

[0019] like Figure 1 As shown, an embodiment of the present invention provides an outpatient scheduling method based on deep learning, the method comprising: S101: Obtain user clinic information; the user clinic information includes the user's current location and m user detection clinic locations; the user detection clinic location represents the location of the clinic where the user needs to undergo examination.

[0020] In this embodiment, the user detection clinic location includes the location of the laboratory department, the location of the imaging department, and the location of the functional examination department.

[0021] Among them, the laboratory is used to perform blood routine, urine routine, stool routine, etc. The imaging department is used to perform X-ray examinations, CT, magnetic resonance imaging, ultrasound, etc. The functional examination department is used to perform electrocardiograms, electromyograms, etc.

[0022] S102: Based on the user clinic information, a first fused feature vector is obtained; the first fused feature vector represents the characteristics of the location and time of the user's arrival at the m user detection clinic locations.

[0023] S103: Through the first prediction network, based on the user clinic information and the first fusion feature vector, the time variation pattern is detected to obtain a recommended user clinic vector and a planned time length; the recommended user clinic vector represents the user clinic location recommended to the user for sequential detection; the planned time length represents the shortest waiting time after arriving at the user clinic location in the recommended user clinic vector.

[0024] Optionally, obtaining a first fusion feature vector based on the user outpatient information includes: Based on the current position of the user and the positions of the detection clinics of multiple users, a first connectivity matrix, a second connectivity vector and a third connectivity vector are obtained.

[0025] Based on the first connectivity matrix, the second connectivity vector and the third connectivity vector, features are extracted to obtain a first fused feature vector; the first fused feature vector represents features that fuse geographic location information, processing status and queuing status.

[0026] Optionally, obtaining the first connectivity matrix, the second connectivity vector, and the third connectivity vector based on the current location of the user and the locations of the detection clinics of multiple users includes: Obtain the number of pending transactions, processing speed, queue number and queue speed; the number of pending transactions represents the number of pending transactions in one user detection clinic location; the processing speed represents the speed of processing transactions in one user detection clinic location; the number of queues represents the number of people queuing in one user detection clinic location; the queue speed represents the speed of queuing in one user detection clinic location.

[0027] The queue number can be obtained by taking a number. The queue speed refers to the speed at which the user's number is called.

[0028] For example, the pending transaction at the user detection clinic corresponding to the blood sampling room is a blood test. The number of pending transactions indicates the number of samples that have not yet undergone a blood test. The processing speed indicates the speed at which the blood test is performed.

[0029] The number of queues changes with time.

[0030] The distance between the user's current location and the locations of the m user detection clinics is filled in a first connectivity matrix; the rows of the first connectivity matrix include the user's current location and the m user detection clinic locations; the columns of the first connectivity matrix include the user's current location and the m user detection clinic locations.

[0031] Wherein, the partial structure of the first connectivity matrix is ​​shown in Table 1:

[0032] The unit of the values ​​in the first connectivity matrix is ​​meter.

[0033] The values ​​in the first connectivity matrix represent the distances from the user's current location to the locations of the m user detection clinics and the distances between any two of the m user detection clinics.

[0034] The distance is not a straight-line distance, but the length of the corresponding route.

[0035] The number of pending transactions for the m user detection clinic locations is divided by the corresponding processing speed to obtain a second connectivity vector.

[0036] The unit of the value in the second connectivity vector is minute.

[0037] The number of queues at the m user detection clinic locations is divided by the corresponding queue speed to obtain a third connectivity vector.

[0038] The unit of the value in the second connectivity vector is meter / minute.

[0039] Optionally, the training method of the first prediction network includes: A training queue data set of multiple days is obtained; the training queue data set includes the number of training queues at multiple testing clinic locations at n queuing time points; the training queue number represents the number of people queuing at one testing clinic location at one user testing clinic location at one time point; n represents dividing 24 hours into n equal parts; the testing clinic location represents the location of all clinics in the hospital.

[0040] Wherein, n is a positive integer. In this embodiment, n is 72, which means that 24 hours are divided into 72 parts, each part represents 20 minutes.

[0041] Among them, the user detection clinic location is the clinic that the user needs to go to for detection, and the detection clinic location is all the clinics in the hospital. The meanings of the two are different.

[0042] Based on the detection clinic location and the number of training queues, a training fusion feature vector is obtained.

[0043] The method for obtaining the training fusion feature vector is the same as the method for obtaining the first fusion feature vector.

[0044] A queue detection vector is constructed by combining 1 detection clinic location, n / 12 queue time points, and the corresponding number of training queues; the number of elements in the queue detection vector is 1+n / 12+n / 12.

[0045] Among them, the first element of the queuing detection vector is the detection clinic location, the second element is the queuing time point, the third element is the training queue number corresponding to the queuing time point corresponding to the second element, the fourth element is the next queuing time point of the queuing time point corresponding to the second element, the fifth element is the training queue number corresponding to the queuing time point corresponding to the fourth element,…, the n / 12+n / 12th element is the next queuing time point corresponding to the queuing time point corresponding to the n / 12+n / 12-2th element, and the 1+n / 12+n / 12th element is the training queue number corresponding to the queuing time point corresponding to the n / 12+n / 12th element.

[0046] Among them, the queuing time point and the corresponding next queuing time point are adjacent time points.

[0047] Wherein, n is greater than 12.

[0048] The queue detection vector and the training fusion feature vector are input into the first prediction network to obtain the predicted queue number; the predicted queue number represents the number of people queuing at the next queue time point at the same detection clinic location.

[0049] Overall, hospital queues show regular patterns, such as higher numbers on weekends than on weekdays and higher numbers in the afternoon than in the morning. Furthermore, the number of queues varies by testing room location. For example, the routine blood test room has more queues than the CT room. Therefore, we use the testing room location, queue time, and corresponding training queue numbers to identify regularities and determine the queue size for the same testing room location at the next queue time.

[0050] One testing clinic location corresponds to one first prediction network; The training queue data at the next queue time point is compared with the predicted queue quantity to obtain the loss, and back propagation is performed to train the first prediction network.

[0051] In this embodiment, the cross entropy loss function is used to calculate the loss.

[0052] Optionally, the training queue data at the next queue time point is compared with the predicted queue quantity to obtain a loss, and the first convolutional neural network, the first deep neural network, the second deep neural network, and the third deep neural network are trained simultaneously.

[0053] Optionally, the detecting time variation pattern by the first prediction network based on the user clinic information and the first fusion feature vector to obtain the recommended user clinic vector and the planned time length includes: Obtain user speed, current time point, and multiple current queue numbers; the user speed represents the user's walking speed; the current time point represents the time point corresponding to the user's current position; the current queue number represents the number of people queuing at a user testing clinic location at the current time point.

[0054] Through the first prediction network, based on the user speed, the current time point, the m current queue numbers and the m queue speeds, m first waiting time lengths, m first detection time points and m first predicted queue numbers are obtained.

[0055] Among them, 1 first waiting time length corresponds to 1 first detection time point, 1 first predicted queue number, and 1 user detection clinic location.

[0056] Through the first prediction network, based on the user speed, the first detection time point, m-1 first predicted queue numbers and m-1 queue speeds, m-1 second waiting time lengths, m-1 second detection time points and m-1 second predicted queue numbers are obtained.

[0057] The m-1 first predicted queue numbers and m-1 queue speeds are the first predicted queue numbers and queue speeds of the m-1 user detection clinic locations excluding the user detection clinic location corresponding to the first detection time point.

[0058] Among them, 1 second waiting time length corresponds to 1 second detection time point, corresponds to 1 second predicted queue number, and corresponds to 1 user detection clinic location.

[0059] Among them, the method for obtaining m-1 second waiting time lengths, m-1 second detection time points and m-1 second predicted queue numbers is the same as the method for obtaining m first waiting time lengths, m first detection time points and m first predicted queue numbers.

[0060] Traverse all user detection clinic locations to obtain 1*2*3…(m-1)*m user routes and corresponding processing time lengths; the user route represents a route passing through all user detection clinic locations; the processing time length represents the sum of the first waiting time length, the second waiting time length, ..., the mth waiting time length on one user route.

[0061] The processing time length that is shorter than other processing time lengths is used as the planning time length, and the user detection clinic positions corresponding to the user routes corresponding to the planning time lengths are sequentially used as elements in the recommended user clinic vector.

[0062] Optionally, obtaining, by the first prediction network, m first waiting time lengths, m first detection time points, and m first predicted queue numbers based on the user speed, the current time point, m current queue numbers, and m queue speeds, includes: The distance from the user's current position in the first connectivity matrix to the m user detection clinic locations is divided by the user's speed to obtain m first time lengths; the first time length represents the time length for the user's current position to reach one user's clinic location.

[0063] Based on the user's detection clinic location, the current time point and the corresponding current queue number, m first predicted arrangement numbers are obtained.

[0064] Divide the first predicted queue number by the queue speed to obtain the first queue time length; Add the current time point to the first time length and the first arrangement time length to obtain the first detection time point; m user clinic locations correspond to m first detection time points; The number of pending transactions at the first detection point is obtained and divided by the corresponding processing speed to obtain the first waiting time length; m first detection time points correspond to m first waiting time lengths.

[0065] Optionally, obtaining m first predicted permutation numbers based on the user's detected clinic location, the current time point, and the corresponding current queue number includes: The current queue detection vector is constructed by combining the user's detection clinic location, current time point and the corresponding current queue number.

[0066] Because there is no data before the current time point, in this embodiment, if n = 72, then 72 / 24 = 3, so the number of elements in the current queue detection vector in this embodiment is 1 + 3 * 2 = 7 elements. Because the information before the current time point is unknown, the current time point is 8:00 AM, so the constructed current queue detection vector is [02, 0, 0, 0, 01, 24], indicating that the patient's test clinic location is 02, and at 8:00 AM on the day labeled 01, there are 24 people in the queue.

[0067] According to the current queue detection vector, a first predicted queue number is obtained through a trained first prediction network; the first predicted queue number represents the number of queues at a time point after a first time length from the current time point.

[0068] Among them, the current queue detection vector is input into the trained first prediction network to obtain the next prediction number; the next prediction number represents the number of queues at the next queue time point after the current time point. If the current time point is 8:00 in the morning, because n is 72, the next queue time point is 8:20. Therefore, the number of queues at 8:20 is found through one prediction. The current time point plus the time point of the first time length is taken as the first time point. If the next prediction number is 42, the user's detection clinic location, the first time point and the corresponding next prediction number are used to construct the next queue detection vector. The next queue detection vector is [02, 0, 0, 01, 24, 02, 46]. The next queue detection vector is input into the trained first prediction network to obtain the next two prediction numbers. The next two prediction numbers represent the number of queues at 8:40.

[0069] If the first time period is 5 minutes, then the number of people in the queue at 8:00 AM, which is close to the time of arrival, is used as the first predicted queue size. If the first time period is 10 minutes, then the number of people in the queue at 8:00 AM, which is close to the time of arrival, is used as the first predicted queue size. If the first time period is 15 minutes, then the number of people in the queue at 8:20 AM, which is close to the time of arrival, is used as the first predicted queue size. If the first time period is 35 minutes, then the number of people in the queue at 8:40 AM, which is close to the time of arrival, is used as the first predicted queue size. Find the number of people in the queue at a closer time point.

[0070] The first prediction network is a deep learning network (DNN) comprising 5 neuron layers.

[0071] m user clinic locations correspond to m first predicted arrangement quantities.

[0072] Optionally, extracting features based on the first connectivity matrix, the second connectivity vector, and the third connectivity vector to obtain a first fused feature vector includes: The first connectivity matrix is ​​input into the first convolutional neural network to extract features and obtain a first feature map.

[0073] In this embodiment, the first convolutional neural network is a convolutional neural network (CNN) composed of 6 layers of 2*2 convolution kernels.

[0074] The first feature map is input into a first deep neural network to extract features to obtain a first feature vector.

[0075] In this embodiment, the first deep neural network is a deep neural network (DNN) including three neuron layers.

[0076] The first connectivity matrix is ​​input into the second deep neural network to extract features and obtain a second eigenvector.

[0077] In this embodiment, the second deep neural network is a deep neural network (DNN) including 5 neuron layers.

[0078] The first connectivity matrix is ​​input into a third deep neural network to extract features and obtain a third eigenvector.

[0079] In this embodiment, the third deep neural network is a deep neural network (DNN) including 6 neuron layers.

[0080] The first eigenvector, the second eigenvector, and the third eigenvector have the same number of elements.

[0081] The number of neurons in the output layers of the first deep neural network, the second deep neural network, and the third deep neural network is the same.

[0082] The first eigenvector, the second eigenvector, and the third eigenvector are averaged by their corresponding subscripts to obtain a first fused eigenvector.

[0083] Example 2

[0084] Based on the above-mentioned outpatient scheduling method based on deep learning, an embodiment of the present invention further provides an outpatient scheduling system based on deep learning, the system comprising: The acquisition module is used to obtain the user's clinic information; the user's clinic information includes the user's current location and m user detection clinic locations; the user detection clinic location represents the location of the clinic where the user needs to undergo examination.

[0085] The feature module is used to obtain a first fused feature vector based on the user's clinic information; the first fused feature vector represents the characteristics of the location and time of the user's arrival at the m user detection clinic locations.

[0086] The detection module is used to detect the time change pattern based on the user clinic information and the first fusion feature vector through the first prediction network, and obtain the recommended user clinic vector and the planned time length; the recommended user clinic vector represents the user clinic location recommended to the user for sequential detection; the planned time length represents the shortest waiting time after arriving at the user clinic location in the recommended user clinic vector.

[0087] The algorithm and display provided herein are not inherently related to any particular computer, virtual system or other device. Various general-purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious that the structure required for constructing this type of system. In addition, the present invention is not directed to any specific programming language. It should be understood that various programming languages ​​can be utilized to realize the content of the present invention described herein, and the above description of specific languages ​​is for the purpose of disclosing the best mode of the present invention.

[0088] In the description provided herein, numerous specific details are described. However, it is understood that embodiments of the present invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques are not shown in detail so as not to obscure the understanding of this description.

[0089] The various component embodiments of the present invention may be implemented in hardware, as software modules running on one or more processors, or as a combination thereof. Those skilled in the art will appreciate that, in practice, a microprocessor or digital signal processor (DSP) may be used to implement some or all of the functionality of some or all of the components of the apparatus according to the embodiments of the present invention. The present invention may also be implemented as an apparatus or device program (e.g., a computer program or computer program product) for performing part or all of the methods described herein. Such a program implementing the present invention may be stored on a computer-readable medium or in the form of one or more signals. Such signals may be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

Claims

1. The outpatient scheduling method based on deep learning is characterized by: include: Get user outpatient information; The user clinic information includes the user's current location and the locations of m user testing clinics; The user detection clinic location indicates the location of the clinic where the user needs to undergo examination; Based on the user clinic information, a first fused feature vector is obtained; the first fused feature vector represents the characteristics of the location and time of the user's arrival at the m user detection clinic locations; Through the first prediction network, based on the user clinic information and the first fusion feature vector, the time change pattern is detected to obtain a recommended user clinic vector and a planned time length; the recommended user clinic vector represents the user clinic location recommended to the user for sequential detection; the planned time length represents the shortest waiting time after arriving at the user clinic location in the recommended user clinic vector.

2. The outpatient scheduling method based on deep learning according to claim 1, characterized in that: The obtaining of a first fusion feature vector based on the user outpatient information includes: Based on the current location of the user and the locations of the detection clinics of multiple users, a first connectivity matrix, a second connectivity vector and a third connectivity vector are obtained; Based on the first connectivity matrix, the second connectivity vector and the third connectivity vector, features are extracted to obtain a first fused feature vector; the first fused feature vector represents features that fuse geographic location information, processing status and queuing status.

3. The outpatient scheduling method based on deep learning according to claim 1, characterized in that: The obtaining of a first connectivity matrix, a second connectivity vector, and a third connectivity vector based on the current location of the user and the locations of the detection clinics of multiple users includes: Obtain the number of pending transactions, processing speed, queue number, and queue speed; the number of pending transactions represents the number of pending transactions in a user detection clinic location; the processing speed represents the speed of processing transactions in a user detection clinic location; the queue number represents the number of people queuing in a user detection clinic location; and the queue speed represents the speed of queuing in a user detection clinic location; Filling the distance between the user's current location and the locations of the m user detection clinics into a first connectivity matrix; the rows of the first connectivity matrix include the user's current location and the locations of the m user detection clinics; the columns of the first connectivity matrix include the user's current location and the locations of the m user detection clinics; The number of pending transactions at the m user detection clinic locations is divided by the corresponding processing speed to obtain a second connectivity vector; The number of queues at the m user detection clinic locations is divided by the corresponding queue speed to obtain a third connectivity vector.

4. The outpatient scheduling method based on deep learning according to claim 1, characterized in that: The training method of the first prediction network includes: Acquire a training queue data set for multiple days; the training queue data set includes the number of training queues for multiple testing clinic locations at n queuing time points; the training queue number represents the number of people queuing at one testing clinic location at one time point; n represents 24 hours divided into n equal parts; the testing clinic locations represent the locations of all clinics in the hospital; Based on the detection clinic location and the number of training queues, a training fusion feature vector is obtained; Construct a queue detection vector based on 1 testing clinic location, n / 12 queue time points, and the corresponding number of training queues; the number of elements in the queue detection vector is 1+n / 12+n / 12; Inputting the queue detection vector and the training fusion feature vector into a first prediction network to obtain a predicted queue number; the predicted queue number represents the number of people queuing at the next queue time point at the same testing clinic location; One testing clinic location corresponds to one first prediction network; The training queue data at the next queue time point is compared with the predicted queue quantity to obtain the loss, and back propagation is performed to train the first prediction network.

5. The outpatient scheduling method based on deep learning according to claim 4 is characterized in that: The training queue data at the next queue time point is compared with the predicted queue quantity to calculate the loss, and the first convolutional neural network, the first deep neural network, the second deep neural network and the third deep neural network are trained at the same time.

6. The outpatient scheduling method based on deep learning according to claim 3, characterized in that: The first prediction network detects the time variation pattern based on the user clinic information and the first fusion feature vector to obtain the recommended user clinic vector and the planned time length, including: Obtain user speed, current time point, and multiple current queue numbers; the user speed represents the user's walking speed; the current time point represents the time point corresponding to the user's current location; the current queue number represents the number of people queuing at a user testing clinic location at the current time point; Obtaining, through the first prediction network, m first waiting time lengths, m first detection time points, and m first predicted queue sizes based on the user speed, the current time point, the m current queue sizes, and the m queue speeds; Obtaining, through the first prediction network, m-1 second waiting time lengths, m-1 second detection time points, and m-1 second predicted queue sizes based on the user speed, the first detection time point, m-1 first predicted queue sizes, and m-1 queue speeds; Traverse all user detection clinic locations to obtain 1*2*3…(m-1)*m user routes and corresponding processing time lengths; the user route represents a route passing through all user detection clinic locations; the processing time length represents the sum of the first waiting time length, the second waiting time length, ..., the mth waiting time length on a user route; The processing time length that is shorter than other processing time lengths is used as the planning time length, and the user detection clinic positions corresponding to the user routes corresponding to the planning time lengths are sequentially used as elements in the recommended user clinic vector.

7. The outpatient scheduling method based on deep learning according to claim 6, characterized in that: The obtaining, through the first prediction network, m first waiting time lengths, m first detection time points, and m first predicted queue numbers based on the user speed, the current time point, m current queue numbers, and m queue speeds, includes: Divide the distance from the user's current position in the first connectivity matrix to the locations of the m user's detection clinics by the user's speed to obtain m first time lengths; the first time length represents the time length for the user's current position to reach the location of one user's clinic; Based on the user's detection clinic location, the current time point, and the corresponding current queue number, m first predicted permutation numbers are obtained; Divide the first predicted queue number by the queue speed to obtain the first queue time length; Add the current time point to the first time length and the first arrangement time length to obtain the first detection time point; m user clinic locations correspond to m first detection time points; The number of pending transactions at the first detection point is obtained and divided by the corresponding processing speed to obtain the first waiting time length; m first detection time points correspond to m first waiting time lengths.

8. The outpatient scheduling method based on deep learning according to claim 7, characterized in that: The method of obtaining m first predicted arrangement numbers based on the user's detected clinic location, the current time point, and the corresponding current queue number includes: The current queue detection vector is constructed by combining the user's detection clinic location, current time point and the corresponding current queue number; According to the current queue detection vector, a first predicted queue number is obtained through a trained first prediction network; the first predicted queue number represents the number of queues at a time point after a first time length from the current time point; m user clinic locations correspond to m first predicted arrangement quantities.

9. The outpatient scheduling method based on deep learning according to claim 2, characterized in that: The extracting features based on the first connectivity matrix, the second connectivity vector, and the third connectivity vector to obtain a first fused feature vector includes: Inputting the first connectivity matrix into the first convolutional neural network, extracting features, and obtaining a first feature map; Inputting the first feature map into a first deep neural network to extract features and obtain a first feature vector; Inputting the first connectivity matrix into a second deep neural network, extracting features, and obtaining a second eigenvector; Inputting the first connectivity matrix into a third deep neural network to extract features and obtain a third eigenvector; The first eigenvector, the second eigenvector and the third eigenvector have the same number of elements; The first eigenvector, the second eigenvector, and the third eigenvector are averaged by their corresponding subscripts to obtain a first fused eigenvector.

10. The outpatient scheduling system based on deep learning is characterized by: include: Acquisition module, used to obtain user outpatient information; The user clinic information includes the user's current location and the locations of m user testing clinics; The user detection clinic location indicates the location of the clinic where the user needs to undergo examination; A feature module is configured to obtain a first fused feature vector based on the user's clinic information; the first fused feature vector represents the characteristics of the location and time of the user's arrival at the locations of the m user detection clinics; The detection module is used to detect the time change pattern based on the user clinic information and the first fusion feature vector through the first prediction network, and obtain the recommended user clinic vector and the planned time length; the recommended user clinic vector represents the user clinic location recommended to the user for sequential detection; the planned time length represents the shortest waiting time after arriving at the user clinic location in the recommended user clinic vector.

Images