Historical encounter data flexible screening method and system

Abstract

The application relates to the technical field of data analysis, and particularly discloses a historical diagnosis data flexible screening method and system, which comprises the following steps: for any patient, querying the identity of the patient, reading a diagnosis array of the patient according to the identity, and sorting the diagnosis array based on a time identifier to obtain a diagnosis array sequence; determining the local importance of each data according to the diagnosis array sequence; clustering all the diagnosis arrays to determine the mean array of each diagnosis array, and determining the global importance of each diagnosis array according to the mean array; and determining the data deletion sequence according to the global importance of the diagnosis array and the local importance of each data. The application determines the importance of data according to the predictability and the required degree of reservation, determines the data deletion sequence according to the ascending order of the importance, and makes the unimportant data be deleted first, so that the method is more gentle and more consistent with the actual situation compared with the traditional periodic deletion method.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data analysis, and in particular to a historical inquiry data flexible screening method and system. BACKGROUND

[0002] With the development of society and the progress of science and technology, the data of many existing hospital areas are converted into digital data. These data can be stored in a database for a very long time, and the cost of the storage is very low, and there is enough storage capacity to store these data. However, the cost of the storage still exists, including the cost of itself, the cost of occupied space and the cost of maintenance. If the data is to be stored all the time, the cost will be higher and higher. Some data in the inquiry data may not be so important, and it can be deleted regularly. The existing deletion method is a hard deletion process according to time, and does not consider the importance of the data too much. Therefore, how to provide a more appropriate screening and deletion scheme based on the data itself is a technical problem to be solved by the technical scheme of the present application. SUMMARY

[0003] The present application aims to provide a historical inquiry data flexible screening method and system to solve the problems raised in the background art.

[0004] To achieve the above-mentioned purpose, the present application provides the following technical scheme:

[0005] A historical inquiry data flexible screening method, the method comprising:

[0006] Normalizing all inquiry data based on a preset medical record template to obtain data types and quantized values, and constructing an inquiry array; wherein the inquiry array and the inquiry data have the same identity and time identifiers;

[0007] For any patient, query its identity, read the inquiry array of the patient according to the identity, and sort based on the time identifier to obtain an inquiry array sequence;

[0008] Determine the local importance of each data according to the inquiry array sequence;

[0009] Cluster all inquiry arrays to determine the mean array of each class of inquiry array, and determine the global importance of each inquiry array according to the mean array;

[0010] Determine the data deletion order according to the global importance of the inquiry array and the local importance of each data, and execute the deletion process regularly and quantitatively based on the data deletion order.

[0011] As a further scheme of the present application: the step of normalizing all the consultation data based on the preset medical record template to obtain data types and quantized values thereof, and constructing a consultation array includes:

[0012] Reading type labels in the preset medical record template in sequence;

[0013] Locating data in the consultation data based on the type labels, and performing non-dimensional processing on the located data to obtain quantized values;

[0014] According to the order of the type labels, the quantized values are counted to obtain a consultation array; each element in the consultation array corresponds to a type label;

[0015] Querying an identity identifier and a time identifier in the consultation data, and copying and inserting the consultation data;

[0016] When the type labels in the medical record template contain priorities, the reading order of the type labels is adjusted based on the priorities; the positioning results corresponding to each type label are counted in real time, the mean value of the positioning results is calculated, and when a certain positioning process fails, the mean value is used as the positioning result, and a prompt label is inserted synchronously.

[0017] As a further scheme of the present application: the step of querying the identity identifier of any patient, reading the consultation array of the patient according to the identity identifier, and sorting the consultation array based on the time identifier to obtain a consultation array sequence includes:

[0018] Querying the identity identifier of any patient;

[0019] Matching the consultation array of the patient in the consultation array library according to the identity identifier;

[0020] Sorting the consultation array of the patient according to the time identifier to obtain a consultation array sequence.

[0021] As a further scheme of the present application: the step of determining the local importance of each data according to the consultation array sequence includes:

[0022] Randomly selecting a preset proportion of the consultation arrays in the consultation array sequence as a feature set, and simultaneously selecting all the remaining consultation arrays as a test set;

[0023] Cyclically executing a preset number of times to obtain a preset number of feature sets and test sets;

[0024] For each pair of feature set and test set, a regression model is trained based on the feature set, and the average error of each data is determined based on the test set;

[0025] For each data, the average error in each training process is calculated, and the mean value of the average errors is calculated.

[0026] The average error-based mean determines the local importance of each data of the patient.

[0027] As a further aspect of the application: the step of clustering all the inquiry arrays, determining the mean array of each inquiry array, and determining the global importance of each inquiry array based on the mean array comprises:

[0028] For all inquiry arrays within a preset time range, the inquiry arrays are pairwise compared to calculate the similarity;

[0029] Based on the similarity, each inquiry array is clustered, and after clustering, the mean array of each inquiry array is determined simultaneously;

[0030] The mean array is input into a preset evaluation model to determine the evaluation value of the mean array;

[0031] The evaluation value is adjusted according to the number of arrays of each inquiry array to determine the global importance.

[0032] As a further aspect of the application: the step of determining the data deletion order based on the global importance of the inquiry array and the local importance of each data, and performing the deletion process based on the data deletion order comprises:

[0033] For each data of each patient, the local importance of the data is queried;

[0034] The global importance of the inquiry array in which the data is located is queried;

[0035] The local importance and the global importance are summed to obtain a comprehensive importance;

[0036] The ascending order of the comprehensive importance is used as the data deletion order, and the deletion process is performed based on the data deletion order.

[0037] The technical scheme of the application further provides a flexible screening system for historical inquiry data, the system comprising:

[0038] An inquiry array construction module is configured to normalize all inquiry data based on a preset medical record template to obtain data types and quantized values, and construct inquiry arrays; wherein the inquiry arrays and the inquiry data have the same identity and time identifiers;

[0039] An inquiry array sorting module is configured to query the identity of any patient, read the inquiry array of the patient based on the identity, and sort the inquiry array based on the time identifier to obtain an inquiry array sequence;

[0040] A local feature extraction module is configured to determine the local importance of each data based on the inquiry array sequence;

[0041] A global feature extraction module is configured to cluster all the diagnosis arrays, determine a mean array of each class of diagnosis arrays, and determine a global importance degree of each diagnosis array according to the mean array;

[0042] A data deletion module is configured to determine a data deletion sequence according to the global importance degree of the diagnosis array and the local importance degree of each data, and perform a deletion process in a time and quantity manner based on the data deletion sequence.

[0043] As a further scheme of the present application, the diagnosis array construction module comprises:

[0044] A type label reading unit is configured to read type labels in a preset medical record template in sequence;

[0045] A data preprocessing unit is configured to locate data in the diagnosis data based on the type labels, and perform dimensionless processing on the located data to obtain a quantitative value;

[0046] A quantitative value statistical unit is configured to statistically obtain the quantitative value according to the sequence of the type labels to obtain a diagnosis array; each element in the diagnosis array corresponds to a type label;

[0047] An identification insertion unit is configured to query an identity identifier and a time identifier in the diagnosis data, and copy and insert the diagnosis data;

[0048] When the type labels in the medical record template contain a priority, the reading sequence of the type labels is adjusted based on the priority; the positioning result corresponding to each type label is statistically obtained in real time, the mean value of the positioning result is calculated, the mean value is used as the positioning result when a certain positioning process fails, and a prompt label is synchronously inserted.

[0049] As a further scheme of the present application, the diagnosis array sorting module comprises:

[0050] An identity identifier query unit is configured to query an identity identifier of any patient;

[0051] An array matching unit is configured to match the diagnosis array of the patient in the diagnosis array library according to the identity identifier;

[0052] A sorting execution unit is configured to sort the diagnosis array of the patient according to the time identifier to obtain a diagnosis array sequence.

[0053] As a further scheme of the present application, the local feature extraction module comprises:

[0054] A feature set construction unit is configured to randomly select diagnosis arrays of a preset proportion in the diagnosis array sequence as a feature set, and simultaneously select all the remaining diagnosis arrays as a test set;

[0055] a loop execution unit, configured to execute a preset number of times to obtain a preset number of times of feature sets and test sets;

[0056] a training unit, configured to train a regression model based on the feature set and determine the average error of each data based on the test set for each pair of feature set and test set;

[0057] an error analysis unit, configured to calculate the average error of each data in each training process and calculate the mean value of the average error;

[0058] an error application unit, configured to determine the local importance of each data of the patient based on the mean value of the average error.

[0059] Compared with the prior art, the beneficial effects of the present application are that the present application analyzes the inquiry data to determine the predictability of each data, identifies the inquiry data itself to determine the degree of retention demand, determines the importance of the data according to the predictability and the degree of retention demand, and determines the data deletion order according to the ascending order of importance, so that the unimportant data is deleted first, which is more gentle and more consistent with the actual situation compared with the traditional periodic deletion method. BRIEF DESCRIPTION OF DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application.

[0061] Figure 1 Flowchart of the historical inquiry data flexible screening method.

[0062] Figure 2 Block diagram of the composition structure of the historical inquiry data flexible screening system. DETAILED DESCRIPTION

[0063] In order to make the technical problems to be solved by the present application, the technical solutions and the beneficial effects more clear, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0064] Figure 1 Flowchart of the historical inquiry data flexible screening method, in the embodiments of the present application, a historical inquiry data flexible screening method, the method comprises:

[0065] Step S100: normalizing all inquiry data based on a preset medical record template to obtain data types and quantized values, and constructing an inquiry array; wherein the inquiry array and the inquiry data have the same identity and time identifier;

[0066] Medical consultation data is the interaction information between patients and medical staff in the hospital. In the same hospital, the medical record templates are generally the same, but as time goes by, the medical record templates will also be updated. At this time, the medical consultation data will be divided into "new and old". The old data is the data counted using the previous template, and the new data is the data counted using the latest template. When the templates are different, the order and format of these data may be different. Therefore, before processing the data, all medical consultation data need to be formatted uniformly. The format unification process is to convert all medical consultation data based on a unified medical record template to obtain unified data. The unified medical record template generally uses the latest medical record template.

[0067] After normalizing all the consultation data, the data conversion process is introduced to quantize each data in the template. The main purpose is to convert different character strings into quantized values. Each data type and its quantized value are counted to obtain an array, called the consultation array. Each element in the consultation array corresponds to a data type. In addition, the consultation array and the consultation data have the same identity and time identifier, which is used to indicate which patient's consultation data the consultation array is and at what time.

[0068] Step S200: query the identity of any patient, read the patient's medical inquiry array according to the identity, and sort it based on the time identifier to obtain a medical inquiry array sequence;

[0069] For any patient, query their identity. The simplest identity is their ID number. All medical question arrays are stored in the database. According to the patient's identity, the patient's medical question array can be read and sorted. The sorting order is chronological. Since the medical question array contains a time identifier, the sorting process is not complicated. After the sorting is completed, the sorted medical question array of each patient is obtained, which is called the medical question array sequence.

[0070] Step S300: determining the local importance of each data according to the query array sequence;

[0071] By analyzing the series of medical questionnaires for each patient, the importance of each data can be evaluated. The analysis process is to compare each data with its own corresponding data at different times. The importance obtained is called local importance. Each data of each patient has a local importance.

[0072] Step S400: clustering all the inquiry arrays, determining the mean array of each type of inquiry array, and determining the global importance of each inquiry array based on the mean array;

[0073] Clustering all the diagnosis arrays in the database can obtain multiple diagnosis arrays, for each diagnosis array, calculating the mean array of all diagnosis arrays of the diagnosis array, analyzing the mean array to determine the importance of the diagnosis array, at this time, the importance is the importance of a diagnosis array, called global importance.

[0074] Step S500: determining the data deletion order according to the global importance of the diagnosis array and the local importance of each data, and executing the deletion process in a time and quantity based manner based on the data deletion order;

[0075] From the above, it can be known that each diagnosis array corresponds to a global importance, and each data in the diagnosis array corresponds to a local importance, for each data, the global importance of the diagnosis array in which the data is located is taken as the global importance of the data, and the global importance and the local importance can obtain a comprehensive importance, the data deletion order is determined by the comprehensive importance, the lower the comprehensive importance, the earlier the data deletion order, and the meaning of executing the deletion process in a time and quantity based manner based on the data deletion order is that every interval of a preset time length deletes a preset data amount of data.

[0076] It should be noted that in the actual application of the technical scheme of the present application, a time span is determined in advance, such as a quarter, at this time, the present application is applied to the data within a quarter, and the obtained diagnosis data is the diagnosis data within the quarter.

[0077] Regarding step S100, the step of constructing the diagnosis array based on the normalization processing of all diagnosis data based on the preset medical record template to obtain the data type and the quantized value includes:

[0078] Reading the type label in the preset medical record template in sequence;

[0079] Based on the type label, positioning the data in the diagnosis data, and performing non-dimensional processing on the positioned data to obtain the quantized value;

[0080] According to the order of the type label, the quantized value is counted to obtain the diagnosis array, each element in the diagnosis array corresponds to a type label;

[0081] Querying the identity and time identifier in the diagnosis data, and copying and inserting the diagnosis data.

[0082] reading the latest medical record template, reading type tags in the latest medical record template in turn, the type tags being the names of each data type, locating data in the inquiry data within a preset time range (within a quarter) based on the type tags, and performing dimensionless processing on the located data to obtain a quantitative value; the dimensionless processing on the located data is actually two steps, the first step is to convert the located data into a numerical format, for a string, the staff generally needs to pre-encode all the values of the string of this type to convert it into a numerical format; after the first step is completed, the second step is to convert the numerical value into unitless data, which is implemented in a target manner including obtaining the maximum value and the minimum value of the numerical value, calculating the difference between any numerical value and the minimum value as the actual difference value, then calculating the difference between the maximum value and the minimum value as the span, calculating the ratio of the actual difference value and the span to obtain the dimensionless processed numerical value; finally, the quantitative values are counted according to the order of the type tags to obtain an inquiry array, after obtaining the inquiry array, the identity and time identifiers in the inquiry data are queried and copied and inserted into the inquiry data.

[0083] It is worth mentioning that when the type tags in the medical record template contain priorities, the reading order of the type tags is adjusted based on the priorities, which affects the type tags corresponding to each element in the inquiry array; in addition, the positioning results corresponding to each type tag are counted in real time, and the mean value of the positioning results is calculated, when a certain positioning process fails, it indicates that the data of the type tag is not found in the inquiry data of a certain patient, at this time, the mean value is used as the positioning result, and a prompt tag is inserted synchronously, which is used to inform the artificial end that the artificial end will check subsequently, which will not be described here.

[0084] Regarding step S200, the step of querying the identity of any patient, reading the inquiry array of the patient according to the identity, and sorting based on the time identifier to obtain an inquiry array sequence includes:

[0085] Querying the identity of any patient;

[0086] Matching the inquiry array of the patient in the inquiry array library according to the identity;

[0087] Sorting the inquiry array of the patient according to the time identifier to obtain an inquiry array sequence.

[0088] For any patient, querying his identity, matching the inquiry array of the patient in the inquiry array library according to the identity, the inquiry array containing a time identifier, determining the time order of the inquiry array according to the time identifier, and counting the inquiry array based on the time order to obtain an inquiry array sequence.

[0089] Regarding step S300, the step of determining the local importance of each data according to the inquiry array sequence includes:

[0090] Randomly select a preset proportion of the inquiry array sequence as a feature set, and meanwhile, select the remaining other inquiry arrays as a test set;

[0091] Cyclically execute a preset number of times to obtain a preset number of pairs of feature sets and test sets;

[0092] For each pair of feature set and test set, train a regression model based on the feature set, and determine the average error of each data based on the test set;

[0093] For each data, calculate the average error in each training process, and calculate the mean value of the average error;

[0094] Determine the local importance of each data of the patient based on the mean value of the average error.

[0095] The above content provides a specific calculation process of local importance, which uses the idea of random forest, randomly selects a preset proportion of inquiry arrays in the inquiry array sequence, for example, assuming that the inquiry array sequence contains 20 inquiry arrays, and the selected proportion is 80%, at this time, 16 inquiry arrays can be selected, and these inquiry arrays are used as a feature set, and meanwhile, the remaining other inquiry arrays are used as a test set (4 inquiry arrays). The selection process is random, and the preset number of times is cyclically executed, so that different splitting methods can be obtained, that is, if the result of one splitting is used as a pair of feature set and test set, then multiple execution processes can obtain multiple pairs of feature set and test set. For each pair of feature set and test set, a regression model is trained based on the feature set, and the test set can be used to test the regression model to obtain an error rate. It should be noted that the regression model is used to process each data in the inquiry array. If the above example is used as a reference, the test set contains 4 inquiry arrays, and each data can be tested 4 times, and the mean value is calculated as the average error of the data. At this time, each data has an average error in each training process, and the average error obtained by cyclically executing the training process for a preset number of times has a preset number of times, and the mean value of the average error in each training process is calculated, and the local importance of each data of the patient is determined based on the mean value of the average error.

[0096] It should be noted that for the random selection process, a condition can also be set, that is, each inquiry array needs to be at least an element in the test set to prevent some data from being used only as a training benchmark for the regression model.

[0097] The mean of the average error represents the predictability of the data (the smaller the mean of the average error, the higher the predictability), and the higher the predictability, the lower the importance according to the present application, because even if it is deleted, the accuracy of prediction according to the remaining data is higher; therefore, the local importance is proportional to the mean of the average error; further, in the sequence of the patient's inquiry array, each type label corresponds to a group of data, and the above example is used as a reference. The second element of the 20 inquiry arrays can extract 20 data, and the local importance of any data in the 20 data is proportional to the mean of the average error according to the above content; on this basis, a condition is introduced, that is, the data is inversely proportional to the total number of remaining data of the 20 data, indicating that the more the same type of data, the more the original data that can be used to predict the data, at this time, the data is less important. In simple terms, for any data, the local importance is inversely proportional to the remaining number of data of the same type label

[0098] Regarding step S400, the step of clustering all inquiry arrays, determining the mean array of each class of inquiry arrays, and determining the global importance of each inquiry array according to the mean array includes:

[0099] For all inquiry arrays within a preset time range, compare the inquiry arrays pairwise to calculate the similarity;

[0100] Cluster each class of inquiry arrays based on the similarity, and simultaneously determine the mean array of each class of inquiry arrays after clustering is completed;

[0101] Input the mean array into a preset evaluation model to determine the evaluation value of the mean array;

[0102] Adjust the evaluation value according to the number of arrays of each class of inquiry arrays to determine the global importance.

[0103] Assuming that the time range is a quarter (it can also be longer, such as a year), for all inquiry arrays within a preset time range, compare the inquiry arrays pairwise to calculate the similarity, cluster each class of inquiry arrays based on the similarity, and simultaneously determine the mean array of each class of inquiry arrays after clustering is completed. These processes can be applied to the simplest array operation; input the mean array into a preset evaluation model to determine the evaluation value of the mean array. The evaluation model is an evaluation model for each index, and existing data importance evaluation standards can be used. The output of the evaluation model is the evaluation value of the mean array, which is used to represent the importance of a class of inquiry arrays, called global importance.

[0104] On this basis, if the number of arrays in a type of inquiry array is more, if this situation is considered to be more valuable for analysis, a regulation coefficient greater than 1 can be provided to multiply the global importance, that is, the global importance can be enlarged; on the contrary, if it is considered that the data sample is sufficient in this case, its importance is not high, a regulation coefficient less than 1 can be provided to multiply the global importance, that is, the global importance can be reduced.

[0105] It is worth mentioning that the generation process of the inquiry array adopts the same medical record template, and the dimensions of the inquiry arrays are the same, which enables the above array calculation process to proceed smoothly.

[0106] Regarding step S500, the step of determining the data deletion order according to the global importance of the inquiry array and the local importance of each data, and performing the deletion process in a time-quantity-based manner based on the data deletion order includes:

[0107] For each data of each patient, the local importance of the data is queried;

[0108] The global importance of the inquiry array in which the data is located is queried;

[0109] The local importance and the global importance are added to obtain a comprehensive importance;

[0110] The ascending order of the comprehensive importance is taken as the data deletion order, and the deletion process is performed in a time-quantity-based manner based on the data deletion order.

[0111] In an example of the technical scheme of the present application, for each data of each patient, the local importance of the data is queried, the global importance of the inquiry array in which the data is located is queried, the local importance and the global importance are added to obtain a comprehensive importance, the ascending order of the comprehensive importance is taken as the data deletion order, and the deletion process is performed in a time-quantity-based manner based on the data deletion order, so that the more unimportant data is deleted first, wherein the specific parameters of the time-quantity-based are determined by the staff according to the specific situation.

[0112] It is worth mentioning that since the local importance of any data is inversely proportional to the remaining number of data of the same type label, the local importance of the remaining data will change after some data are deleted, and accordingly, the comprehensive importance will also change, thereby building a recursive adjustment process.

[0113] Figure 2 The composition structure block diagram of the historical inquiry data flexible screening system, in the embodiment of the present application, a historical inquiry data flexible screening system, the system 10 includes:

[0114] The inquiry array construction module 11 is configured to normalize all inquiry data based on a preset medical record template to obtain data types and quantized values, and construct an inquiry array; the inquiry array and the inquiry data have the same identity and time identifiers;

[0115] The inquiry array sorting module 12 is configured to query the identity of any patient, read the inquiry array of the patient according to the identity, and sort the inquiry array based on the time identifier to obtain an inquiry array sequence.

[0116] The local feature extraction module 13 is configured to determine the local importance of each data according to the inquiry array sequence.

[0117] The global feature extraction module 14 is configured to cluster all inquiry arrays, determine the mean array of each type of inquiry array, and determine the global importance of each inquiry array according to the mean array.

[0118] The data deletion module 15 is configured to determine the data deletion order according to the global importance of the inquiry array and the local importance of each data, and execute the deletion process in a timely and quantitative manner based on the data deletion order.

[0119] Further, the inquiry array construction module 11 comprises:

[0120] The type label reading unit is configured to read type labels in the preset medical record template in sequence.

[0121] The data preprocessing unit is configured to locate data in the inquiry data based on the type labels, and perform dimensionless processing on the located data to obtain quantized values.

[0122] The quantized value statistical unit is configured to count the quantized values according to the order of the type labels to obtain an inquiry array; each element in the inquiry array corresponds to a type label.

[0123] The identifier insertion unit is configured to query the identity and time identifier in the inquiry data, and copy and insert the inquiry data.

[0124] When the type labels in the medical record template contain priorities, the reading order of the type labels is adjusted based on the priorities; the positioning results corresponding to each type label are counted in real time, the mean value of the positioning results is calculated, the mean value is used as the positioning result when a certain positioning process fails, and a prompt label is inserted synchronously.

[0125] Specifically, the inquiry array sorting module 12 comprises:

[0126] The identity query unit is configured to query the identity of any patient.

[0127] An array matching unit is configured to match the diagnosis array of the patient from the diagnosis array library according to the identity;

[0128] An order execution unit is configured to sort the diagnosis array of the patient according to the time identifier to obtain a diagnosis array sequence.

[0129] Further, the local feature extraction module 13 comprises:

[0130] A feature set construction unit is configured to randomly select a preset proportion of diagnosis arrays in the diagnosis array sequence as a feature set, and select the remaining diagnosis arrays as a test set;

[0131] A loop execution unit is configured to execute a preset number of times to obtain the preset number of times of the feature set and the test set;

[0132] A training unit is configured to train a regression model based on the feature set and determine the average error of each data based on the test set for each pair of feature set and test set;

[0133] An error analysis unit is configured to calculate the average error of each data in each training process and calculate the mean value of the average error.

[0134] An error application unit is configured to determine the local importance of each data of the patient based on the mean value of the average error.

[0135] The above merely describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, and improvement within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A flexible screening method for historical medical consultation data, characterized in that: The method comprises: Based on the preset medical record template, all the medical consultation data are normalized to obtain the data type and its quantitative value, and a medical consultation array is constructed; wherein the medical consultation array and the medical consultation data have the same identity and time stamp; For any patient, query their identity, read the patient's medical inquiry array according to the identity and sort it based on the time stamp to obtain the medical inquiry array sequence; Determine the local importance of each data according to the query array sequence; Clustering all the inquiry arrays, determining the mean array of each type of inquiry array, and determining the global importance of each inquiry array based on the mean array; Determine the data deletion order based on the global importance of the query array and the local importance of each data, and execute the deletion process based on the data deletion order in a timely and quantitative manner; The step of determining the local importance of each data according to the query array sequence includes: Randomly select a preset proportion of question arrays from the question array sequence as the feature set, and use all other question arrays as the test set; The loop is executed for a preset number of times to obtain the preset number of pairs of feature sets and test sets; For each pair of feature set and test set, train the regression model based on the feature set and calculate the average error of each data based on the test set; For each data, calculate its average error in each training process and calculate the mean of the average errors; Determine the local importance of each data of the patient based on the mean of the average error; The steps of clustering all the inquiry arrays, determining the mean array of each type of inquiry array, and determining the global importance of each inquiry array according to the mean array include: For all the medical consultation arrays within the preset time range, perform pairwise comparisons on the medical consultation arrays and calculate the similarity; Cluster each type of medical inquiry array based on similarity. After clustering is completed, determine the mean array of each type of medical inquiry array at the same time; Inputting the mean array into a preset evaluation model to determine an evaluation value of the mean array; The evaluation value is adjusted according to the number of arrays in each type of inquiry array to determine the global importance.

2. The flexible screening method for historical medical inquiry data according to claim 1, characterized in that: The steps of normalizing all medical consultation data based on a preset medical record template to obtain data types and quantitative values ​​and constructing a medical consultation array include: Read the type labels in the preset medical record template in sequence; Locating data in the medical inquiry data based on the type label, performing dimensionless processing on the located data to obtain a quantized value; According to the sequential statistical quantization values ​​of the type labels, a medical inquiry array is obtained; each element in the medical inquiry array corresponds to a type label; Query the identity and time stamp in the medical consultation data, copy and insert the medical consultation data; Among them, when the type tag in the medical record template contains a priority, the reading order of the type tag is adjusted based on the priority; the positioning results corresponding to each type tag are counted in real time, and the mean of the positioning results is calculated. When a positioning process fails, the mean is used as the positioning result, and a prompt tag is inserted synchronously.

3. The flexible screening method for historical medical inquiry data according to claim 1, characterized in that: The steps of querying the identity of any patient, reading the patient's medical inquiry array according to the identity, and sorting the array based on the time identifier to obtain the medical inquiry array sequence include: For any patient, query their identity; Matching the patient's medical inquiry array in the medical inquiry array library according to the identity identifier; The patient's medical consultation array is sorted according to the time stamp to obtain a medical consultation array sequence.

4. The flexible screening method for historical medical inquiry data according to claim 1, characterized in that: The steps of determining the data deletion order according to the global importance of the query array and the local importance of each data, and executing the deletion process in a timely and quantitative manner based on the data deletion order include: For each data of each patient, query the local importance of the data; Query the global importance of the consultation array where the data is located; Add the local importance and the global importance to get the comprehensive importance; The ascending order of comprehensive importance is used as the data deletion order, and the deletion process is executed regularly and quantitatively based on the data deletion order.

5. A flexible screening system for historical medical consultation data, characterized in that: The system comprises: A medical inquiry array construction module is used to normalize all medical inquiry data based on a preset medical record template, obtain the data type and its quantized value, and construct a medical inquiry array; wherein the medical inquiry array and the medical inquiry data have the same identity and time stamp; The medical inquiry array sorting module is used to query the identity of any patient, read the patient's medical inquiry array according to the identity, and sort it based on the time stamp to obtain the medical inquiry array sequence; The local feature extraction module is used to determine the local importance of each data according to the query array sequence; A global feature extraction module is used to cluster all the question arrays, determine the mean array of each type of question array, and determine the global importance of each question array based on the mean array; A data deletion module is used to determine the data deletion order according to the global importance of the query array and the local importance of each data, and to perform the deletion process in a timely and quantitative manner based on the data deletion order; The local feature extraction module includes: A feature set construction unit is used to randomly select a preset proportion of question arrays from the question array sequence as a feature set, and use all other question arrays as a test set; A loop execution unit, configured to loop and execute a preset number of times to obtain a preset number of pairs of feature sets and test sets; a training unit, configured to, for each pair of a feature set and a test set, train a regression model based on the feature set, and calculate an average error of each data based on the test set; The error analysis unit is used to calculate the average error of each data in each training process and the mean of the average errors; an error application unit, configured to determine the local importance of each data of the patient based on a mean value of the average error; The steps of clustering all the medical question arrays, determining the mean array of each type of medical question array, and determining the global importance of each medical question array based on the mean array include: For all the medical consultation arrays within the preset time range, perform pairwise comparisons on the medical consultation arrays and calculate the similarity; Cluster each type of medical inquiry array based on similarity. After clustering is completed, determine the mean array of each type of medical inquiry array at the same time; Inputting the mean array into a preset evaluation model to determine an evaluation value of the mean array; The evaluation value is adjusted according to the number of arrays in each type of inquiry array to determine the global importance.

6. The flexible screening system for historical medical inquiry data according to claim 5, characterized in that: The inquiry array building module includes: A type label reading unit, used to read type labels in a preset medical record template in sequence; a data preprocessing unit, configured to locate data in the medical inquiry data based on the type label, and perform dimensionless processing on the located data to obtain a quantized value; A quantization value statistics unit is used to count the quantization values ​​according to the order of the type labels to obtain a question array; each element in the question array corresponds to a type label; An identification insertion unit is used to query the identity identification and time identification in the medical inquiry data, and copy and insert the medical inquiry data; Among them, when the type tag in the medical record template contains a priority, the reading order of the type tag is adjusted based on the priority; the positioning results corresponding to each type tag are counted in real time, and the mean of the positioning results is calculated. When a positioning process fails, the mean is used as the positioning result, and a prompt tag is inserted synchronously.

7. The flexible screening system for historical medical inquiry data according to claim 5, characterized in that: The inquiry array sorting module includes: An identity query unit, used to query the identity of any patient; An array matching unit, configured to match the patient's medical inquiry array in the medical inquiry array library according to the identity identifier; The sorting execution unit is used to sort the patient's medical inquiry array according to the time mark to obtain a medical inquiry array sequence.

Images