A data mining system, method and device

Abstract

The present invention relates to a data mining method, system and related devices. The data mining method at least includes acquiring historical medical data to train a memory network model to achieve the establishment and update of a department classification model. The establishment and update of the department classification model at least includes the following steps: inputting the historical medical condition description into the memory network model for training to obtain the probability that the historical medical condition description corresponds to each department; evaluating the training effect of the memory network model, and determining whether to stop training based on the evaluation result; if the evaluation result is invalid training, updating the parameters of the memory network model for the next training; if the evaluation result is to stop training, generating the department classification model from the current memory network model.

Description

Technical Field

[0001] The present invention relates to the technical field of data mining, and in particular to a data mining system, method and device. Background Art

[0002] In current medical services, assigning patients to the correct department is time-consuming and laborious, and requires professional medical staff as guides to assign patients to the corresponding department based on the description of each patient's condition. However, most patients cannot accurately describe their condition and can only simply describe the onset of the disease. It is often the case that a patient visits multiple departments before determining the correct department to visit, thus delaying treatment. In severe cases, the wrong treatment direction may even endanger life.

[0003] In summary, the present invention provides a data mining system, method and device to solve the problems existing in the prior art.

[0004] In addition, on the one hand, there are differences in understanding among those skilled in the art; on the other hand, the applicant studied a large number of documents and patents when making the present invention, but due to space limitations, not all details and contents are listed in detail. However, this does not mean that the present invention does not have the characteristics of these prior arts. On the contrary, the present invention already has all the characteristics of the prior art, and the applicant reserves the right to add relevant prior art to the background technology. Summary of the invention

[0005] In view of the deficiencies of the prior art, the present invention provides a data mining method. The data mining method at least includes acquiring historical medical data to train a memory network model to achieve the establishment and update of a department classification model. The establishment and update of a department classification model includes at least the following steps: inputting the historical medical condition description into the memory network model for training to obtain the probability that the historical medical condition description corresponds to each department; evaluating the training effect of the memory network model, and determining whether to stop training based on the evaluation result; if the evaluation result is invalid training, determining the result is retraining, and updating the parameters of the memory network model for the next training; if the evaluation result is good training, determining the result is stopping training, and generating the department classification model from the current memory network model.

[0006] According to a preferred embodiment, the data mining method acquires and summarizes historical medical data. The historical medical data includes historical condition descriptions and corresponding historical departments. The data mining method trains a memory network model based on the historical medical data to obtain the department classification model.

[0007] According to a preferred embodiment, the data mining method guides the patient to click on a preset entry to obtain a description of the patient's condition that conforms to the expression of medical terms.

[0008] According to a preferred embodiment, the data mining method inputs the patient's condition description that conforms to the expression of medical terms into the department classification model to obtain the corresponding department, so as to automatically triage the patient's medical treatment.

[0009] According to a preferred embodiment, the department classification model is trained on historical medical data so that historical disease descriptions and corresponding historical departments establish a disease description word library. The disease description word library can provide vocabulary data support for the data mining method to guide patients, so that the preset terms selected by patients to express their medical purposes are consistent with medical terms.

[0010] According to a preferred embodiment, the data mining method collects the patient's non-medical terminology description of the condition, and establishes a mapping relationship between the non-medical terminology description and the medical terminology description of the same condition. In the case of guiding the patient to click on the preset terms to describe the patient's condition, the preset terms provided to the patient by the data mining method for the same condition include non-medical terminology descriptions and / or medical terminology descriptions. In the case of the patient clicking on the non-medical terminology description and / or the medical terminology description, the data mining method transmits the corresponding medical terminology description to the department classification model to determine the patient's treatment department. Preferably, in the process of guiding the patient to describe his own condition, the non-medical terminology terms can be easy for the patient to understand. Preferably, in the case of the patient clicking on the non-medical terminology description and / or the medical terminology description, the data mining method only transmits the corresponding medical terminology description to the department classification model, which can effectively reduce the amount of processed data compared to the method of transmitting both the non-medical terminology description and the medical terminology description to the department classification model for processing, thereby improving the processing efficiency of the department classification model, and the use of medical terminology description can increase the accuracy of the classification results of the department classification model.

[0011] According to a preferred embodiment, the data mining method divides the patient's medical purpose into a first-class purpose that requires the participation of professional medical personnel and a second-class purpose that does not require the participation of professional medical personnel, according to whether the realization of the patient's medical purpose requires the participation of professional medical personnel. In the case where the patient's medical purpose is a first-class purpose, the data mining method obtains the patient's condition description by guiding the patient to click on the preset entry, and inputs the patient's condition description into the department classification model to obtain the corresponding department.

[0012] The present invention also provides a data mining device. The data mining device at least includes: a training module, an evaluation module, an adjustment module, a generation module, a storage module and a transmission module. The training module is used to input the historical medical condition description into the memory network model for training to obtain the probability that the historical medical condition description corresponds to each department. The evaluation module is used to evaluate the training effect of the training module and determine whether to stop training based on the evaluation result. The adjustment module is used to update the parameters of the memory network model for the next training when the evaluation module determines that the training is to be retrained. The generation module is used to generate a department classification model from the current memory network model when the evaluation module determines that the training is to be stopped. The storage module is used to store historical medical data. The transmission module sends the patient's condition description to the department classification model and returns the classification result of the department classification model.

[0013] The present invention also provides a data mining system. The system at least includes a user terminal, a medical terminal and a server. The user terminal can obtain the patient's condition description and upload it to the server. The medical terminal can upload historical medical data to the server. Preferably, the historical medical data includes historical condition descriptions and corresponding historical departments. The server establishes a department classification model based on the historical medical data, and inputs the condition description into the department classification model to determine the corresponding department. The server at least includes: a training module, an evaluation module, an adjustment module, a generation module, a storage module and a transmission module. The training module is used to input the historical condition description into the memory network model to obtain the probability that the historical condition description corresponds to each department. The evaluation module is used to evaluate the training effect of the training module and determine whether to stop training according to the evaluation result. The adjustment module is used to update the parameters of the memory network model when the evaluation result of the evaluation module is invalid training, so as to carry out the next training. The generation module is used to generate the department classification model from the current memory network model when the evaluation module determines that the training is stopped. The storage module is used to store the historical medical data uploaded by the medical terminal. The transmission module sends the patient's condition description to the department classification model, and returns the classification result of the department classification model to the user end.

[0014] Preferably, the department classification model is established in the following manner, that is, the training module retrieves the historical medical data stored in the storage module, inputs the historical condition description in the historical medical data into the memory network model for training, and obtains the probability that the historical condition description corresponds to each department. Preferably, the evaluation module evaluates the training results of the training module based on the historical departments corresponding to the historical condition descriptions. Preferably, the evaluation module determines whether to stop training according to the evaluation results. Preferably, the evaluation results of the evaluation module include good training, effective training and invalid training, and the corresponding determination results are stop training, continue training and retraining. In the case where the evaluation module determines that the training is stopped, the generation module generates the department classification model from the current memory network model. In the case where the evaluation module determines that the training is retraining, the adjustment module updates the parameters of the memory network model for the next training, and repeats the above steps until the evaluation module determines that the training is stopped, and the generation module generates the department classification model from the determined memory network model.

[0015] Preferably, the evaluation module arranges each department in the training result in descending order of probability values. Preferably, when the department with the highest probability in the training result is the same as the historical department corresponding to the historical condition description, the evaluation result of the evaluation module is passed, and the judgment result is to stop training. Preferably, when the probability of the historical department corresponding to the historical condition description in the training result is not the highest, but is in the top three of the probability values ​​of each department in descending order, and the sum of the top three probabilities is not less than half of the total probability, and the ratio of the probability value of the latter item to the previous item in descending order of probability values ​​is not less than three quarters, the evaluation result of the evaluation module is retraining, and the judgment result is to continue training. Preferably, when the training result does not meet the above two conditions, the evaluation result of the evaluation module is invalid training, and the judgment result is retraining. Preferably, when the evaluation result of the evaluation module is retraining and the judgment result is to continue training, the training module continues to train without changing the parameters of the memory network model until the evaluation module determines that the training is stopped, and the generation module generates the department classification model based on the determined memory network model.

[0016] According to a preferred embodiment, the user terminal may be a self-service machine set up in a hospital. Preferably, the user terminal may include an interaction module, a processing module, a database module and a communication module. Preferably, the interaction module can guide the patient to log in and / or register to the user terminal to obtain patient data including the patient's medical purpose, and send the patient data to the server. Preferably, the interaction module can describe the patient's condition through the preset terms selected by the patient. Preferably, medical staff, scientific research institutions, etc. collect non-medical terminology descriptions of the patient's condition, and establish a mapping relationship between the non-medical terminology description of the same condition and the medical terminology description and store them in the database module. Preferably, the department classification model is trained on historical medical data so that the historical condition description and the corresponding historical department establish a condition description vocabulary. The condition description vocabulary can provide vocabulary data support for the data mining method to guide the patient, so that the preset terms selected by the patient to express his or her own medical purpose conform to the medical terminology.

[0017] Preferably, when guiding the patient to click on a preset entry to describe the patient's condition, the preset entry provided to the patient by the interactive module for the same condition includes a non-medical terminology description and / or a medical terminology description. When the patient clicks on a non-medical terminology description and / or a medical terminology description, the processing module transmits the corresponding medical terminology description to the server through the communication module to determine the department the patient is visiting. Preferably, the transmission module of the server transmits the corresponding medical terminology description to the department classification model to obtain a classification result, and returns the classification result of the department classification model to the user end.

[0018] Preferably, in the process of guiding patients to describe their own conditions, non-medical terminology entries can be easier for patients to understand. Preferably, when a patient clicks on a non-medical terminology description and / or a medical terminology description, the user terminal only transmits the corresponding medical terminology description to the department classification model, which can effectively reduce the amount of processed data compared to the method of transmitting both the non-medical terminology description and the medical terminology description to the department classification model for processing, thereby improving the processing efficiency of the department classification model, and the use of medical terminology description can increase the accuracy of the classification results of the department classification model.

[0019] Preferably, in response to receipt of the classification result, the user terminal combines the classification result and the structural data of the pre-stored hospital to generate a medical navigation video. Preferably, the communication module and the transmission module establish a data transmission channel to transmit data between the user terminal and the server. Preferably, the classification result of the department classification model includes at least one destination department. In response to receipt of the classification result, the processing module combines the structural data of the pre-stored hospital in the database module to generate a patient's route and model a part of the model corresponding to the hospital for the patient's route. Preferably, the local model modeled by the processing module can be sent to the patient's personal smart terminal in the form of a video stream. The smart terminal can receive the video stream by pairing with the user terminal.

[0020] Preferably, the data mining system provided by the present invention can also update the department classification model. The updating of the department classification model is carried out in the following manner, that is, medical personnel, scientific research institutions, etc. upload medical information as new historical medical data to the server through the medical terminal, thereby updating the historical medical data in the storage module. The training module retrieves the new historical medical data stored in the storage module, and inputs the historical disease description in the new historical medical data into the memory network model for training, and obtains the probability that the historical disease description corresponds to each department. The evaluation module evaluates the training results of the training module based on the historical departments corresponding to the historical disease descriptions. When the evaluation module determines that the training is to stop, the generation module generates the department classification model from the current memory network model, thereby completing the update of the department classification model. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a simplified schematic diagram of a data mining system according to a preferred embodiment of the present invention;

[0022] Figure 2 is a simplified schematic diagram of a server according to a preferred embodiment of the present invention;

[0023] Figure 3 is a simplified schematic diagram of a user terminal of a preferred embodiment provided by the present invention;

[0024] Figure 4 It is a conceptual diagram of a user terminal (self-service machine) of a preferred implementation mode provided by the present invention.

[0025] Reference numerals list

[0026] 100: data mining system; 110: user end; 111: interaction module; 112: processing module; 113: database module; 114: communication module; 120: medical end; 130: server; 131: training module; 132: evaluation module; 133: adjustment module; 134: generation module; 135: storage module; 136: transmission module. DETAILED DESCRIPTION

[0027] The following is combined with Figures 1 to 3 The invention provides a data mining method, system and related devices. The invention generates a department classification model based on a large amount of historical medical data. When a patient visits a doctor, the invention converts the patient's non-medical terminology description of the condition into a description of the condition that conforms to medical terminology, and processes the patient's condition description through the department classification model to determine the department for consultation, thereby realizing automatic triage.

[0028] The present invention mines historical medical data to convert the patient's non-medical terminology description of the condition into a medical terminology description of the condition, thereby avoiding the problem of wrong consultation department and treatment direction due to incorrect condition description.

[0029] Example 1

[0030] The present invention also provides a data mining system 100. Figure 1 Preferably, the data mining system 100 includes at least a user terminal 110, a medical terminal 120 and a server 130. The user terminal 110 can obtain the patient's condition description and upload it to the server 130. The medical terminal 120 can upload historical medical data to the server 130. Preferably, the historical medical data includes historical condition descriptions and corresponding historical departments. The server 130 establishes a department classification model based on the historical medical data, and inputs the condition description into the department classification model to determine the corresponding department.

[0031] See also Figure 2The server 130 at least includes: a training module 131, an evaluation module 132, an adjustment module 133, a generation module 134, a storage module 135 and a transmission module 136. The training module 131 is used to input the historical condition description into the memory network model to obtain the probability of the historical condition description corresponding to each department. The evaluation module 132 is used to evaluate the training effect of the training module 131 and determine whether to stop training according to the evaluation result. The adjustment module 133 is used to update the parameters of the memory network model for the next training when the evaluation result of the evaluation module 132 is invalid training. The generation module 134 is used to generate a department classification model from the current memory network model when the evaluation module 132 determines that the training is stopped. The storage module 135 is used to store the historical medical data uploaded by the medical terminal 120. The transmission module 136 sends the patient's condition description to the department classification model and returns the classification result of the department classification model to the user terminal 110.

[0032] Preferably, the establishment of the department classification model is achieved in the following manner, that is, the training module 131 retrieves the historical medical data stored in the storage module 135, inputs the historical condition description in the historical medical data into the memory network model for training, and obtains the probability of the historical condition description corresponding to each department. Preferably, the evaluation module 132 evaluates the training results of the training module 131 based on the historical departments corresponding to the historical condition description. Preferably, the evaluation module 132 determines whether to stop training according to the evaluation results. Preferably, the evaluation results of the evaluation module 132 include good training, effective training and invalid training, and the corresponding determination results are stop training, continue training and retraining. In the case where the evaluation module 132 determines that the training is stopped, the generation module 134 generates a department classification model from the current memory network model. In the case where the evaluation module 132 determines that the training is retraining, the adjustment module 133 updates the parameters of the memory network model to perform the next training, and the above steps are repeated until the evaluation module 132 determines that the training is stopped, and the generation module 134 generates a department classification model from the determined memory network model.

[0033] Preferably, the evaluation module 132 arranges each department in the training result in descending order of probability values. Preferably, when the department with the highest probability in the training result is the same as the historical department corresponding to the historical medical description, the evaluation result of the evaluation module 132 is passed, and the judgment result is to stop the training. Preferably, when the probability of the historical department corresponding to the historical medical description is not the highest in the training result, but is in the top three in the descending order of the probability values ​​of each department, and the sum of the top three probabilities is not less than half of the total probability, and the ratio of the probability value of the latter item to the previous item in the descending order of the probability value is not less than three quarters, the evaluation result of the evaluation module 132 is retraining, and the judgment result is to continue training. Preferably, when the training result does not meet the above two conditions, the evaluation result of the evaluation module 132 is invalid training, and the judgment result is retraining. Preferably, when the evaluation result of the evaluation module 132 is retraining and the judgment result is to continue training, the training module 131 continues training without changing the parameters of the memory network model until the evaluation module 132 judges to stop training, and the generation module 134 generates a department classification model based on the judged memory network model.

[0034] Preferably, the user terminal 110 may be a self-service machine installed in a hospital. Figure 3 Preferably, the user terminal 110 may include an interaction module 111, a processing module 112, a database module 113 and a communication module 114. Preferably, the interaction module 111 can guide the patient to log in and / or register to the user terminal 110 to obtain patient data including the patient's medical purpose, and send the patient data to the server 130. Preferably, the interaction module 111 can describe the patient's condition through the preset terms selected by the patient. Preferably, medical personnel, scientific research institutions, etc. collect non-medical terminology descriptions of the patient's condition, and establish a mapping relationship between the non-medical terminology description of the same condition and the medical terminology description and store them in the database module 113. Preferably, the department classification model is trained on historical medical data so that the historical condition description and the corresponding historical department establish a condition description vocabulary. The condition description vocabulary can provide vocabulary data support for the data mining method to guide the patient, so that the preset terms selected by the patient to express his or her own medical purpose are consistent with medical terms.

[0035] Preferably, when the patient is instructed to click on a preset entry to describe the patient's condition, the preset entry provided to the patient by the interactive module 111 for the same condition includes a non-medical terminology description and / or a medical terminology description. When the patient clicks on a non-medical terminology description and / or a medical terminology description, the processing module 112 transmits the corresponding medical terminology description to the server 130 through the communication module 114 to determine the department the patient is visiting. Preferably, the transmission module 136 of the server 130 transmits the corresponding medical terminology description to the department classification model to obtain a classification result, and returns the classification result of the department classification model to the user terminal 110.

[0036] Preferably, in the process of guiding patients to describe their own conditions, non-medical terminology entries can be easy for patients to understand. Preferably, when the patient clicks on the non-medical terminology description and / or the medical terminology description, the user terminal 110 only transmits the corresponding medical terminology description to the department classification model, which can effectively reduce the amount of processed data compared to the method of transmitting both the non-medical terminology description and the medical terminology description to the department classification model for processing, thereby improving the processing efficiency of the department classification model, and the use of medical terminology description can increase the accuracy of the classification results of the department classification model.

[0037] Preferably, in response to receiving the classification result, the user terminal 110 combines the classification result and the structural data of the pre-stored hospital to generate a medical navigation video. Preferably, the communication module 114 and the transmission module 136 establish a data transmission channel to transmit data between the user terminal 110 and the server 130. Preferably, the classification result of the department classification model includes at least one destination department. In response to receiving the classification result, the processing module 112 combines the structural data of the pre-stored hospital in the database module 113 to generate a patient's route and model a part of the model corresponding to the hospital for the patient's route. Preferably, the local model modeled by the processing module 112 can be sent to the patient's personal smart terminal in the form of a video stream. The smart terminal can receive the video stream by pairing with the user terminal 110.

[0038] Preferably, in response to receiving the destination department, the user terminal 110 can generate a treatment route for the corresponding patient based on the corresponding patient's target location (the address of the corresponding department), the starting location (the location of the user terminal 110) and the sight. The treatment route for the corresponding patient is formed in the following manner, that is, the processing module 112 generates an initial treatment route based on the target location, the starting location and the sight of the corresponding patient in combination with the hospital structure data stored in the database module 113. In response to the generation of the initial treatment route, the user terminal 110 connects to the hospital monitoring network to obtain the road conditions of each node on the initial treatment route, and corrects the errors on the initial treatment route to generate the final treatment route. In the case where the hospital temporarily closes a certain area on the initial treatment route, the user terminal 110 will re-plan the closed path to prioritize the patient to the channel temporarily opened by the hospital management department.

[0039] Preferably, the user terminal 110 can obtain its own position data through the configured positioning unit, and then obtain the starting position and viewing direction of the corresponding patient by mirroring the own position data. The starting position and viewing direction of the corresponding patient are determined in the following manner, that is, the user terminal 110 uses the positioning unit to perform position detection to determine its own current position and orientation, and after mirroring its own current position and orientation along the display side, the viewing direction and positioning of the current patient of the user terminal 110 are determined, and the starting position and viewing direction of the patient in the corresponding local model are determined based on the viewing direction and positioning of the current patient in combination with the hospital structure data.

[0040] Preferably, the user terminal 110 can generate a virtual character image in the generated virtual model. Preferably, the starting position and viewing direction of the virtual character image are determined in the following manner, that is, the user terminal 110 determines the viewing direction and positioning of the current patient of the user terminal 110 according to its own current position and orientation, and provides the viewing direction and positioning to the user terminal 110, which determines the starting position and viewing direction of the corresponding virtual character image, so that the starting position and viewing direction of the corresponding virtual character image are the same as the actual position and viewing direction of the current patient of the user terminal 110, thereby facilitating the patient to substitute the model into reality.

[0041] The user terminal 110 can provide voice prompts to the patient using the user terminal 110 through the interaction module 111 to prompt the corresponding patient to select his or her personalized virtual character image, and determine the patient's current position and orientation through the positioning function of the smart terminal to thereby determine the starting position and viewing direction of the virtual character image selected by the patient, thereby ensuring that the position and viewing direction of the virtual character image selected by the patient are consistent with the patient's actual position and viewing direction.

[0042] When the corresponding local model for determining the starting position and viewing direction of the patient is displayed to the patient in the form of a video, the patient can determine the characteristic reference objects on the navigation path in the navigation model by adjusting the viewing angle of the corresponding virtual character image based on the movement of the virtual character corresponding to the patient's actual position and viewing direction in the virtual model. When the medical treatment route generated by the patient by operating the user terminal 110 intersects with the medical treatment route of at least another patient, the virtual character image corresponding to the current patient will appear in the navigation video generated by the other patient by operating the user terminal 110 to determine the starting position and viewing direction of the patient and sent to the smart terminal, thereby reflecting the actual flow of people in the corresponding area in the local model.

[0043] Preferably, the data mining system 100 provided by the present invention can also update the department classification model. The updating of the department classification model is carried out in the following manner, that is, medical personnel, scientific research institutions, etc. upload medical information as new historical medical data to the server 130 through the medical terminal 120, thereby updating the historical medical data in the storage module 135. The training module 131 retrieves the new historical medical data stored in the storage module 135, and inputs the historical disease description in the new historical medical data into the memory network model for training, and obtains the probability that the historical disease description corresponds to each department. The evaluation module 132 evaluates the training results of the training module 131 based on the historical departments corresponding to the historical disease descriptions. When the evaluation module 132 determines that the training is to stop, the generation module 134 generates a department classification model from the current memory network model, thereby completing the update of the department classification model.

[0044] Example 2

[0045] This embodiment is a further improvement on Embodiment 1, and the repeated contents will not be repeated here.

[0046] The present embodiment provides a data mining method. The data mining method at least includes acquiring historical medical data to train the memory network model to realize the establishment and update of the department classification model. The establishment and update of the department classification model at least includes the following steps: inputting the historical medical condition description into the memory network model for training to obtain the probability that the historical medical condition description corresponds to each department; evaluating the training effect of the memory network model, and determining whether to stop training based on the evaluation result; if the evaluation result is invalid training, the determination result is retraining, and updating the parameters of the memory network model for the next training; if the evaluation result is good training, the determination result is to stop training, and generate a department classification model from the current memory network model.

[0047] Preferably, the data mining method acquires and summarizes historical medical data. The historical medical data includes historical condition descriptions and corresponding historical departments. The data mining method trains the memory network model based on the historical medical data to obtain a department classification model.

[0048] Preferably, the data mining method guides the patient to click on a preset entry to obtain a description of the patient's condition that conforms to the expression of medical terms.

[0049] Preferably, the data mining method inputs the patient's condition description expressed in medical terms into a department classification model to obtain the corresponding department, so as to automatically triage the patient's medical treatment.

[0050] Preferably, the department classification model is trained on historical medical data so that historical disease descriptions and corresponding historical departments establish a disease description vocabulary. The disease description vocabulary can provide vocabulary data support for the data mining method to guide patients, so that the preset terms selected by patients to express their medical purposes are consistent with medical terms.

[0051] Preferably, the data mining method collects the patient's non-medical terminology description of the condition, and establishes a mapping relationship between the non-medical terminology description and the medical terminology description of the same condition. In the case of guiding the patient to click on the preset terms to describe the patient's condition, the preset terms provided to the patient by the data mining method for the same condition include non-medical terminology descriptions and / or medical terminology descriptions. In the case of the patient clicking on the non-medical terminology description and / or the medical terminology description, the data mining method transmits the corresponding medical terminology description to the department classification model to determine the patient's treatment department. Preferably, in the process of guiding the patient to describe his own condition, the non-medical terminology terms can be easy for the patient to understand. Preferably, in the case of the patient clicking on the non-medical terminology description and / or the medical terminology description, the data mining method only transmits the corresponding medical terminology description to the department classification model, which can effectively reduce the amount of data to be processed compared to the method of transmitting both the non-medical terminology description and the medical terminology description to the department classification model for processing, thereby improving the processing efficiency of the department classification model, and the use of medical terminology description can increase the accuracy of the classification results of the department classification model.

[0052] Preferably, the data mining method divides the patient's medical purpose into a first-class purpose that requires the participation of professional medical personnel and a second-class purpose that does not require the participation of professional medical personnel, according to whether the realization of the patient's medical purpose requires the participation of professional medical personnel. In the case where the patient's medical purpose is a first-class purpose, the data mining method obtains the patient's condition description by guiding the patient to click on the preset entry, and inputs the patient's condition description into the department classification model to obtain the corresponding department.

[0053] Example 3

[0054] This embodiment is a further improvement of Embodiment 1 and Embodiment 3, and the repeated contents are not repeated here. This embodiment provides a data mining device. The data mining device at least includes: a training module 131, an evaluation module 132, an adjustment module 133, a generation module 134, a storage module 135 and a transmission module 136. The training module 131 is used to input the historical condition description into the memory network model for training, and obtain the probability of the historical condition description corresponding to each department. The evaluation module 132 is used to evaluate the training effect of the training module 131, and determine whether to stop training according to the evaluation result. The adjustment module 133 is used to update the parameters of the memory network model for the next training when the evaluation module 132 determines that the training is to be retrained. The generation module 134 is used to generate a department classification model from the current memory network model when the evaluation module 132 determines that the training is to be stopped. The storage module 135 is used to store historical medical data. The transmission module 136 sends the patient's condition description to the department classification model and returns the classification result of the department classification model.

[0055] It should be noted that the above-mentioned specific embodiments are exemplary, and those skilled in the art can come up with various solutions inspired by the disclosure of the present invention, and these solutions also belong to the disclosure scope of the present invention and fall within the protection scope of the present invention. Those skilled in the art should understand that the present invention specification and its drawings are illustrative and do not constitute a limitation on the claims. The scope of protection of the present invention is defined by the claims and their equivalents. Throughout the text, the features guided by "preferably" are only an optional method and should not be understood as being required. Therefore, the applicant reserves the right to abandon or delete the relevant preferred features at any time. The present invention specification contains multiple inventive concepts, such as "preferably", "according to a preferred embodiment" or "optionally", all of which indicate that the corresponding paragraph discloses an independent concept, and the applicant reserves the right to file a divisional application based on each inventive concept.

Claims

1. A data mining system, characterized in that: The system at least comprises a user terminal (110), a medical terminal (120) and a server (130); When a patient clicks on a non-medical terminology description and / or a medical terminology description, the user terminal (110) is able to obtain the patient's condition description and only upload the corresponding medical terminology description that has a mapping relationship with the non-medical terminology description of the same condition to the server (130); the medical terminal (120) is able to upload historical medical data to the server (130); the server (130) establishes a department classification model based on the historical medical data, and inputs the condition description into the department classification model to determine the corresponding department; wherein the historical medical data includes historical condition descriptions and corresponding historical departments.

2. The data mining system according to claim 1, characterized in that: The server (130) returns the classification result of the department classification model to the user terminal (110); In response to receiving the classification result, the user terminal (110) generates a medical consultation navigation video by combining the classification result with pre-stored hospital structure data.

3. The data mining system according to claim 1 or 2, characterized in that: The server (130) is configured with: a storage module (135) and a transmission module (136); The storage module (135) is used to store the historical medical data uploaded by the medical terminal (120); The transmission module (136) sends the patient's condition description to the department classification model, and returns the classification result of the department classification model to the user terminal (110).

4. The data mining system according to claim 3, characterized in that: The server (130) is further configured with: a training module (131), an evaluation module (132) and a generation module (134); The training module (131) is used to input the historical condition description into the memory network model to obtain the probability that the historical condition description corresponds to each department; The evaluation module (132) is used to evaluate the training effect of the training module (131) and determine whether to stop the training according to the evaluation result; The generation module (134) is used to generate the department classification model from the current memory network model when the evaluation module (132) determines that the training is to be stopped.

5. The data mining system according to claim 4, characterized in that: The server (130) is further configured with: an adjustment module (133); The adjustment module (133) is used to update the parameters of the memory network model for the next training when the evaluation module (132) determines that the result is retraining.

6. The data mining system according to claim 3, characterized in that: The user terminal (110) at least includes an interaction module (111); the interaction module (111) can guide the patient to log in and / or register to the user terminal (110) to obtain patient data including the patient's medical purpose, and send the patient data to the server (130); wherein the interaction module (111) can describe the patient's condition through a preset entry selected by the patient.

7. A data mining device, characterized in that: The data mining device establishes data connections with the user terminal (110) and the medical terminal (120) respectively, and is characterized in that: When the patient clicks on the non-medical terminology description and / or the medical terminology description, the user terminal (110) obtains the patient's condition description and uploads only the corresponding medical terminology description that has a mapping relationship with the non-medical terminology description of the same condition to the data mining device; The medical terminal (120) uploads the historical medical data to the data mining device; The data mining device establishes a department classification model based on the historical medical data, and inputs the condition description into the department classification model to determine the corresponding department; wherein the historical medical data includes historical condition descriptions and corresponding historical departments.

8. The data mining device according to claim 7, characterized in that: The data mining device at least comprises: a training module (131), an evaluation module (132), an adjustment module (133), a generation module (134), a storage module (135) and a transmission module (136); The training module (131) is used to input the historical condition description into the memory network model for training, and obtain the probability that the historical condition description corresponds to each department; The evaluation module (132) is used to evaluate the training effect of the training module (131) and determine whether to stop the training according to the evaluation result; The adjustment module (133) is used to update the parameters of the memory network model for the next training when the evaluation module (132) determines that the result is retraining; The generating module (134) is used to generate a department classification model from the current memory network model when the evaluation module (132) determines that the training is to be stopped; The storage module (135) is used to store historical medical data; The transmission module (136) sends the patient's condition description to the department classification model and returns the classification result of the department classification model.

9. A data mining method, characterized in that: The data mining method at least comprises: When the patient clicks on the non-medical terminology description and / or the medical terminology description, the user terminal (110) obtains the patient's condition description and uploads only the corresponding medical terminology description that has a mapping relationship with the non-medical terminology description of the same condition to the server (130); The medical terminal (120) uploads the historical medical data to the server (130); The server (130) establishes a department classification model based on the historical medical data, and inputs the condition description into the department classification model to determine the corresponding department; wherein the historical medical data includes historical condition descriptions and corresponding historical departments.

10. The data mining method according to claim 9, characterized in that: The establishment and update of the department classification model at least includes the following steps: Inputting the historical medical condition description into a memory network model for training, and obtaining the probability that the historical medical condition description corresponds to each department; Evaluate the training effect of the memory network model and decide whether to stop training based on the evaluation results; If the evaluation result is invalid training, the determination result is retraining, and the parameters of the memory network model are updated to perform the next training; When the evaluation result is good training, the judgment result is to stop training and generate the department classification model from the current memory network model.

Images