Method, device, storage medium and electronic equipment for data classification

Abstract

The present disclosure relates to a data classification method, device, storage medium and electronic equipment, which comprises the following steps: obtaining sample data corresponding to a target event under multiple data labels; obtaining a preset data screening model, and determining a preset key data label and a preset non-key data label from the data labels; determining key data corresponding to the key data label and non-key data corresponding to the non-key data label from the sample data; determining an associated data label related to the key data label from the non-key data label according to the key data and the non-key data through the preset data screening model; obtaining a preset data classification training model, training the preset data classification training model through the key data and the non-key data corresponding to the associated data label, and obtaining a target data classification model; obtaining target data to be classified in the target event, taking the target data as the input of the target data classification model, and obtaining a classification result.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the field of data classification, in particular, to a data classification method, device, storage medium and electronic equipment. BACKGROUND

[0002] At present, the judgment of the patient's physical condition in the medical field is mostly based on the analysis of various physical condition index data, such as gender, age, hemoglobin, trace elements and other index data. In general, the evaluation of the target event related to the patient (for example, the evaluation event of the timing of the dialysis of the patient with kidney disease after completing this dialysis) is mostly based on the clinical observational study of the patient's physical condition index data and the experience judgment of the medical staff. However, due to the different physical conditions of different patients, the corresponding index data is also more. Simply evaluating the target event through clinical observation and experience judgment not only has low efficiency, but also needs to improve the accuracy. SUMMARY

[0003] The purpose of the present disclosure is to provide a data classification method, device, storage medium and electronic equipment.

[0004] In a first aspect, a data classification method is provided, the method comprising: obtaining sample data corresponding to a target event under a plurality of data labels; obtaining a preset data screening model, determining a pre-set key data label and a non-key data label from the data labels, and determining key data corresponding to the key data label and non-key data corresponding to the non-key data label from the sample data; determining an associated data label related to the key data label from the non-key data label according to the key data and the non-key data through the preset data screening model; obtaining a preset data classification training model, training the preset data classification training model through the key data and the non-key data corresponding to the associated data label to obtain a target data classification model; obtaining target data to be classified in the target event, taking the target data as the input of the target data classification model, and obtaining a classification result.

[0005] Optionally, the determining, according to the key data and the non-key data, the associated data label related to the key data label from the non-key data label through the preset data screening model comprises: performing permutation and combination processing on the non-key data label to obtain a plurality of non-key data label sets, wherein different non-key data label sets correspond to different permutation and combination results; combining the key data label with each non-key data label set to obtain a plurality of data label set groups; taking, as inputs of a preset data screening model, the key data corresponding to the key data label and the non-key data corresponding to the non-key data label in each data label set group in sequence to obtain output results; determining a classification accuracy corresponding to each data label set group according to the output results; determining, from all the data label set groups, a target label set group with the highest classification accuracy, and determining the non-key data label in the target label set group as the associated data label.

[0006] Optionally, the determining, according to the output results, the classification accuracy corresponding to each data label set group comprises: obtaining an expected output result corresponding to each data label set group; and obtaining the classification accuracy according to the output result and the expected output result.

[0007] Optionally, the target data classification model comprises:

[0008]

[0009] wherein x1, x2,...x n represents an input variable of the target data classification model, x i represents the i-th data, Y represents an output variable of the target classification model, b i , c i , p i , a0 represents a parameter of the target data classification model, and i includes any one of 1 to n.

[0010] In a second aspect, a device for data classification is provided, which comprises: an acquisition module configured to acquire sample data corresponding to a target event under a plurality of data labels; a determination module configured to acquire a preset data screening model, determine a preset key data label and a preset non-key data label from the data labels, and determine key data corresponding to the key data label and non-key data corresponding to the non-key data label from the sample data; a data screening module configured to determine, according to the key data and the non-key data, an associated data label related to the key data label from the non-key data label through the preset data screening model; a model training module configured to acquire a preset data classification training model, train the preset data classification training model through the key data and the non-key data corresponding to the associated data label, and obtain a target data classification model; and a classification module configured to acquire target data to be classified in the target event, take the target data as an input of the target data classification model, and obtain a classification result.

[0011] Optionally, the data screening module comprises: a first determination submodule configured to perform permutation and combination processing on the non-key data labels to obtain a plurality of non-key data label sets, wherein different non-key data label sets correspond to different permutation and combination results; a second determination submodule configured to combine the key data label with each non-key data label set to obtain a plurality of data label set groups; a third determination submodule configured to sequentially take the key data corresponding to the key data label and the non-key data corresponding to the non-key data label in each data label set group as an input of a preset data screening model to obtain an output result; a fourth determination submodule configured to determine a classification accuracy corresponding to each data label set group according to the output result; and a fifth determination submodule configured to determine, from all the data label set groups, a target label set group with the highest classification accuracy, and determine the non-key data label in the target label set group as the associated data label.

[0012] Optionally, the fourth determination submodule is configured to acquire an expected output result corresponding to each data label set group, and determine the classification accuracy according to the output result and the expected output result.

[0013] Optionally, the target data classification model comprises:

[0014]

[0015] wherein x1, x2,...x n represents an input variable of the target data classification model, xi represents the i-th data, Y represents an output variable of the target classification model, b i , c i , p ia0 represents a parameter of the target data classification model, and i includes any one from 1 to n.

[0016] In a third aspect, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program. The computer program is executed by a processor to implement the steps of the method in the first aspect.

[0017] In a fourth aspect, an electronic device is provided, and the electronic device includes a memory storing a computer program and a processor configured to execute the computer program stored in the memory to implement the steps of the method in the first aspect.

[0018] According to the above technical solution, the sample data corresponding to the target event under multiple data labels is obtained, the preset data screening model is obtained, the pre-set key data label and non-key data label are determined from the data labels, and the key data corresponding to the key data label and the non-key data corresponding to the non-key data label are determined from the sample data. According to the key data and the non-key data, the preset data screening model is used to determine the associated data label related to the key data label from the non-key data label. The preset data classification training model is obtained, the preset data classification training model is trained by the key data and the non-key data corresponding to the associated data label, and the target data classification model is obtained. The target data to be classified in the target event is obtained, the target data is taken as the input of the target data classification model, and the classification result is obtained. In this way, the target event can be evaluated by the target data classification model according to the target data, the data analysis efficiency is improved, and the accuracy of the evaluation of the target event is also improved.

[0019] Other features and advantages of the present disclosure will be described in detail in the following detailed description section. BRIEF DESCRIPTION OF DRAWINGS

[0020] The accompanying drawings are included to provide a further understanding of the present disclosure and constitute a part of the specification, and are used together with the following detailed description to explain the present disclosure, but do not constitute a limitation of the present disclosure. In the drawings:

[0021] Figure 1 is a flowchart of a first data classification method according to an exemplary embodiment;

[0022] Figure 2 is a flowchart of a second data classification method according to an exemplary embodiment;

[0023] Figure 3 is a block diagram of a first data classification device according to an exemplary embodiment;

[0024] Figure 4is a block diagram of an apparatus for a second data classification according to an exemplary embodiment;

[0025] Figure 5 is a block diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0026] The specific embodiments of the present disclosure will be described below in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely intended to illustrate and explain the present disclosure, and are not intended to limit the present disclosure.

[0027] The present disclosure provides a data classification method, apparatus, storage medium and electronic device. First, sample data corresponding to a target event under a plurality of data labels and a preset data screening model are obtained. Then, the sample data corresponding to the target event under the plurality of data labels is screened by the preset data screening model, and a target data label affecting the classification result of a target data classification model is determined. In this way, the accuracy of the classification result of the model is improved. The preset data classification training model is trained according to the target sample data corresponding to the target data label, and a target data classification model is obtained. Thus, the target data in the target event to be classified can be classified by the target data classification model, so that the target event is evaluated according to the target data. The data analysis efficiency is improved, and the accuracy of the evaluation of the target event is also improved. Further, technical support can be provided for evaluating the target event related to a patient, and the morbidity and mortality of the patient can be reduced.

[0028] The specific embodiments of the present disclosure will be described below in detail with reference to the accompanying drawings.

[0029] Figure 1 is a flowchart of a data classification method according to an exemplary embodiment, as shown in Figure 1 The method comprises the following steps:

[0030] S101, obtaining sample data corresponding to a target event under a plurality of data labels.

[0031] The target event can include an evaluation event of a patient with kidney disease on the timing of the patient's dialysis after completing the dialysis, an evaluation event of a patient with tumor on the timing of the patient's chemotherapy after completing the chemotherapy, etc. The sample data can include physical condition index data. The data label can be used to represent the attribute of each sample data. For example, the data label can include gender, age at dialysis, first dialysis time, hemoglobin, albumin, creatinine, urea, potassium, phosphorus, whether heart failure, whether nausea and vomiting, whether edema, whether diabetic nephropathy, whether uremic encephalopathy, etc.

[0032] In a possible implementation, the sample data can be obtained from a related database. For example, when evaluating the hemodialysis timing of a uremia patient, 1000 pieces of physical condition index data of uremia patients in the past three years can be randomly selected from a national hemodialysis case information registration system in the Chinese Research Data Services Platform (CNRDS) as the sample data.

[0033] Generally, before the preset model (for example, a preset data screening model, a preset data classification training model, and the like, which will be mentioned later) is trained according to the sample data, the sample data needs to be preprocessed to exclude noise data, improve the efficiency of model training, and reduce the complexity of model calculation. For example, when the sample data includes data labels such as gender, dialysis age, first dialysis time, hemoglobin, albumin, creatinine, urea, potassium, phosphorus, whether heart failure, whether nausea and vomiting, whether edema, whether diabetic nephropathy, and whether uremic encephalopathy, the sample data includes two types of data. One type is numerical data, including dialysis age, first dialysis time, hemoglobin, albumin, creatinine, urea, potassium, and phosphorus. The other type is Boolean data, including gender, whether heart failure, whether nausea and vomiting, whether edema, whether diabetic nephropathy, and whether uremic encephalopathy. At this time, the numerical data in the sample data can be normalized (for example, the numerical data can be taken as natural logarithm with base e), and the Boolean data can be assigned a value of 0 or 1, so as to complete the preprocessing operation of the sample data.

[0034] In S102, a preset data screening model is obtained, a preset key data label and a non-key data label are determined from the data label, and key data corresponding to the key data label and non-key data corresponding to the non-key data label are determined from the sample data.

[0035] The preset data screening model can include a BP neural network.

[0036] In actual application scenarios, some sample data corresponding to data labels are mandatory sample data. For example, when the target event is the evaluation of the hemodialysis timing of a uremic patient, the sample data corresponding to the data labels of gender, age at dialysis, first dialysis time, hemoglobin, albumin, creatinine, and urea are mandatory sample data. In this case, the key data labels are gender, age at dialysis, first dialysis time, hemoglobin, albumin, creatinine, and urea, and the non-key data labels are potassium, phosphorus, heart failure, nausea and vomiting, edema, diabetic nephropathy, and uremic encephalopathy. This is only an example, and the present disclosure is not limited in this regard. The key data labels and the non-key data labels can be specifically set according to different application scenarios, and the present disclosure is not limited in this regard.

[0037] In S103, the associated data labels related to the key data labels are determined from the non-key data labels according to the key data and the non-key data by using the preset data screening model.

[0038] In this step, the non-key data labels can be arranged and combined to obtain a plurality of non-key data label sets, wherein different non-key data label sets correspond to different arrangement and combination results. The key data labels are combined with each non-key data label set to obtain a plurality of data label set groups. The key data corresponding to the key data labels and the non-key data corresponding to the non-key data labels in each data label set group are sequentially input into the preset data screening model as input, and output results are obtained. The classification accuracy corresponding to each data label set group is determined according to the output results. The target label set group with the highest classification accuracy is determined from all the data label set groups, and the non-key data labels in the target label set group are determined as the associated data labels. In addition, when the classification accuracy corresponding to each data label set group is determined according to the output results, the expected output results corresponding to each data label set group can be obtained, and then the classification accuracy is obtained according to the output results and the expected output results.

[0039] The output result can include the survival time of the patient (for example, in months), for example, the output result can be "the survival time of the patient is 30 months", in a possible implementation, the survival time of the patient can also be classified, and the output result can include the category of the survival time of the patient, for example, the survival time of the patient can be classified into two categories, and 36 months is used as a critical value, when the survival time of the patient is less than 36 months, the patient is determined to be in a survival poor group, when the survival time of the patient is greater than or equal to 36 months, the patient is determined to be in a survival good group, of course, the survival time can also be classified into more than two categories, and the critical value can be set according to the actual application scenario, and the disclosure does not limit the critical value; the expected output result can include the actual survival time of each patient or the category of the actual survival time, since the sample data records the actual survival time corresponding to the first dialysis time of each patient, the actual survival time of each patient or the category of the actual survival time is known, that is, the expected output result corresponding to each patient under the data label set group is known.

[0040] In addition, since the data label in each data label set group can correspond to the sample data of multiple patients, the output result can include the output result corresponding to each patient in the multiple patients corresponding to each data label set group, that is, when the sample data corresponding to the data label in each data label set group (that is, the key data label and the non-key data label in each data label set group) is input into the preset data screening model as the input of the preset data screening model to obtain the output result, the sample data of the multiple patients corresponding to the data label set group can be sequentially input into the preset data screening model, so that the output result corresponding to each patient can be obtained, and correspondingly, the expected output result corresponding to each data label set group can include the expected output result corresponding to each patient in the multiple patients under the data label set group.

[0041] When the classification accuracy corresponding to each data label set group is determined according to the output result, the expected output result corresponding to each patient under each data label set group can be obtained, and then the classification accuracy can be obtained according to the output result and the expected output result corresponding to each patient, specifically, the output result of each patient can be compared with the expected output result of the patient, and when it is determined that the output result is the same as the corresponding expected output result according to the comparison result, it is determined that the output result is correct, so that the ratio of the number of patients whose output result is correct to the total number of patients corresponding to the data label set group can be calculated to obtain the classification accuracy corresponding to the data label set group.

[0042] S104, acquire a preset data classification training model, train the preset data classification training model by the key data and the non-key data corresponding to the associated data label, and obtain a target data classification model.

[0043] The preset data classification training model can include a fuzzy T-S neural network, and the target data classification model can include:

[0044]

[0045] wherein x1, x2,...x n represents an input variable of the target data classification model (for example, the input variable can be body condition index data), x i represents the i-th data in the input variable, Y represents an output variable of the target classification model (for example, the output variable can be survival time), b i (i=1, 2,..., n), c i (i=1, 2,..., n), p i (i=1, 2,..., n), a0 represents a parameter of the target data classification model.

[0046] In a possible implementation, the key data and the associated data can be taken together as inputs of the preset data classification training model by a particle swarm algorithm, the preset data classification training model is trained, model parameters of the classification model are determined, for example, the model parameters can include parameters (i.e., b i , c i ) of membership functions in the antecedent network in the fuzzy T-S neural network and coefficients (i.e., p i ) of linear structures in the consequent network in the fuzzy T-S neural network in the target data classification model.

[0047] S105, acquire target data to be classified in the target event, take the target data as an input of the target data classification model, and obtain a classification result.

[0048] In the case that the target event is an evaluation event of a dialysis time for a patient with kidney disease after the patient completes the dialysis, the input of the target data classification model can include data corresponding to key labels such as gender, age at dialysis, first dialysis time, hemoglobin, albumin, creatinine, urea, and data corresponding to associated data labels such as potassium, phosphorus, edema, diabetic nephropathy, uremic encephalopathy, and the like of the patient with kidney disease determined by a preset data screening model, and the output of the target data classification model can include the survival time of the patient with kidney disease (for example, in months), and further, the survival time of the patient with kidney disease can be classified, and the classification result is the category of the survival time of the patient with kidney disease, for example, the survival time can be classified into two categories, and 36 months is used as a critical value, when the survival time of the patient with kidney disease is less than 36 months, the patient with kidney disease is determined to be in a poor survival group, and when the survival time of the patient with kidney disease is greater than or equal to 36 months, the patient with kidney disease is determined to be in a good survival group, of course, the survival time can be classified into more than two categories, and the critical value can be set according to the actual application scenario, which is not limited in the present disclosure.

[0049] By using the above method, the target data to be classified in the target event can be classified by the target data classification model, so as to evaluate the target event according to the target data, thereby improving the data analysis efficiency and the accuracy of the evaluation of the target event.

[0050] Figure 2 is a flowchart of a data classification method according to an example embodiment, in this example, taking the target event as an evaluation event of a dialysis time for a patient with uremia after the patient completes the dialysis, as shown in Figure 2 The method comprises the following steps:

[0051] S201, obtaining sample data corresponding to an evaluation event of a dialysis time for a patient with uremia under a plurality of data labels.

[0052] The sample data can include physical condition index data, and the data label can be used to represent the attribute of each sample data, for example, the data label can include gender, age at dialysis, first dialysis time, hemoglobin, albumin, creatinine, urea, potassium, phosphorus, whether heart failure, whether nausea and vomiting, whether edema, whether diabetic nephropathy, whether uremic encephalopathy, and the like.

[0053] In a possible implementation, the sample data can be obtained from a related database. For example, when evaluating the hemodialysis timing of a uremia patient, 1000 pieces of physical condition index data of uremia patients in the past three years can be randomly selected from a national hemodialysis case information registration system in the Chinese Research Data Services Platform (CNRDS) as the sample data.

[0054] Generally, before the preset model (for example, a preset data screening model, a preset data classification training model, and the like, which will be mentioned later) is trained according to the sample data, the sample data needs to be preprocessed to exclude noise data, improve the efficiency of model training, and reduce the complexity of model calculation. For example, when the sample data includes data labels such as gender, dialysis age, first dialysis time, hemoglobin, albumin, creatinine, urea, potassium, phosphorus, whether heart failure, whether nausea and vomiting, whether edema, whether diabetic nephropathy, and whether uremic encephalopathy, the sample data includes two types of data. One type of data is numerical data, including dialysis age, first dialysis time, hemoglobin, albumin, creatinine, urea, potassium, and phosphorus. The other type of data is Boolean data, including gender, whether heart failure, whether nausea and vomiting, whether edema, whether diabetic nephropathy, and whether uremic encephalopathy. At this time, the numerical data in the sample data can be normalized (for example, the numerical data can be taken as natural logarithm with base e), and the Boolean data can be assigned a value of 0 or 1, so as to complete the preprocessing operation of the sample data.

[0055] S202, obtaining a preset data screening model, determining a preset key data label and a non-key data label from the data label, and determining key data corresponding to the key data label and non-key data corresponding to the non-key data label from the sample data.

[0056] The preset data screening model can include a BP neural network.

[0057] In actual application scenarios, some data labels correspond to mandatory sample data, for example, in the evaluation of the hemodialysis timing of a uremic patient, the sample data corresponding to the data labels of gender, age at dialysis, first dialysis time, hemoglobin, albumin, creatinine, and urea are mandatory sample data. At this time, the key data labels are gender, age at dialysis, first dialysis time, hemoglobin, albumin, creatinine, and urea, and the non-key data labels are potassium, phosphorus, heart failure, nausea and vomiting, edema, diabetic nephropathy, and uremic encephalopathy. This is only an example, and the present disclosure is not limited thereto. The key data labels and the non-key data labels can be specifically set according to different application scenarios, and the present disclosure is not limited thereto.

[0058] S203, performing permutation and combination processing on the non-key data labels to obtain a plurality of non-key data label sets.

[0059] Different non-key data label sets correspond to different permutation and combination results.

[0060] For example, taking the non-key data labels including potassium, phosphorus, and whether heart failure as an example, after permutation and combination processing on the above three labels, seven non-key data label sets are obtained, which are (potassium), (phosphorus), (whether heart failure), (potassium, phosphorus), (potassium, whether heart failure), (phosphorus, whether heart failure), and (potassium, phosphorus, whether heart failure). The above example is only an example, and the present disclosure is not limited thereto.

[0061] S204, combining the key data labels with each non-key data label set to obtain a plurality of data label set groups.

[0062] Different data label set groups include different non-key data label sets.

[0063] For example, taking the key data tags including gender, age at dialysis, time of first dialysis, hemoglobin, albumin, creatinine, and urea as an example, it is assumed that seven non-key data tag sets are obtained after S203, i.e., (potassium), (phosphorus), (whether heart failure), (potassium, phosphorus), (potassium, whether heart failure), (phosphorus, whether heart failure), and (potassium, phosphorus, whether heart failure). At this time, the key data tags are combined with each of the non-key data tag sets, and seven data tag set groups are obtained, i.e., (gender, age at dialysis, time of first dialysis, hemoglobin, albumin, creatinine, urea, potassium), (gender, age at dialysis, time of first dialysis, hemoglobin, albumin, creatinine, urea, phosphorus), (gender, age at dialysis, time of first dialysis, hemoglobin, albumin, creatinine, urea, whether heart failure), (gender, age at dialysis, time of first dialysis, hemoglobin, albumin, creatinine, urea, potassium, phosphorus), (gender, age at dialysis, time of first dialysis, hemoglobin, albumin, creatinine, urea, potassium, whether heart failure), (gender, age at dialysis, time of first dialysis, hemoglobin, albumin, creatinine, urea, phosphorus, whether heart failure), and (gender, age at dialysis, time of first dialysis, hemoglobin, albumin, creatinine, urea, potassium, phosphorus, whether heart failure). The above example is only illustrative, and the present disclosure is not limited in this regard.

[0064] It should be noted that the data tags in each of the data tag set groups can correspond to sample data of multiple patients. For example, when the sample data includes physical condition index data of patients A, B, and C, the data tags in each of the data tag set groups can correspond to the physical condition index data of patients A, B, and C.

[0065] S205, sequentially taking the key data corresponding to the key data tags and the non-key data corresponding to the non-key data tags in each of the data tag set groups as inputs of a preset data screening model to obtain an output result.

[0066] The output result can include the survival time of the patient (for example, in months). For example, the output result can be "the survival time of the patient is 30 months". In one possible implementation, the survival time of the patient can be classified. In this case, the output result can include the category of the survival time of the patient. For example, the survival time can be classified into two categories, and 36 months can be used as a critical value. When the survival time of the patient is less than 36 months, the patient can be determined to be in a survival inferior group. When the survival time of the patient is greater than or equal to 36 months, the patient can be determined to be in a survival superior group. Of course, the survival time can be classified into more than two categories, and the critical value can be set according to an actual application scenario, and the present disclosure is not limited in this regard.

[0067] It should be noted that, since the data labels in each data label set group can correspond to sample data of multiple patients, the output result can include the output result corresponding to each patient of the multiple patients corresponding to each data label set group, that is, when the sample data corresponding to the data labels (i.e., the key data labels and the non-key data labels in each data label set group) in each data label set group are taken as the input of the preset data screening model to obtain the output result, the sample data of the multiple patients corresponding to the data label set group can be sequentially input into the preset data screening model, so that the output result corresponding to each patient can be obtained. For example, when the data labels in each data label set group correspond to the physical condition index data of patients A, B and C, in the execution of this step, the physical condition index data of patient A corresponding to the data labels in the data label set group can be input into the preset data screening model first to obtain the output result corresponding to patient A; then the physical condition index data of patient B corresponding to the data labels in the data label set group can be input into the preset data screening model to obtain the output result corresponding to patient B; finally, the physical condition index data of patient C corresponding to the data labels in the data label set group can be input into the preset data screening model to obtain the output result corresponding to patient C. The above is only an example and the present disclosure does not limit this.

[0068] S206, determining a classification accuracy corresponding to each data label set group according to the output result.

[0069] In this step, the classification accuracy can be obtained by obtaining the expected output result corresponding to each patient under each data label set group, and then obtaining the classification accuracy according to the output result and the expected output result corresponding to each patient. For example, the expected output result can include the actual survival time of each patient or the category in which the actual survival time is located. Since the sample data records the actual survival time corresponding to the first dialysis time of each patient, the actual survival time of each patient or the category in which the actual survival time is located is known, that is, the expected output result corresponding to each patient under the data label set group is known.

[0070] Since the data labels in each data label set group can correspond to sample data of multiple patients, and the output result corresponding to each patient can be obtained after S205 is executed, when calculating the classification accuracy, the output result of each patient can be compared with the expected output result of the patient, and when it is determined according to the comparison result that the output result is the same as the corresponding expected output result, it can be determined that the output result is correct. In this way, the classification accuracy corresponding to the data label set group can be obtained by calculating the ratio of the number of patients whose output result is correct to the total number of patients corresponding to the data label set group.

[0071] For example, the output result of the patient survival time is classified as a category, in this example, the category of the patient survival time includes a survival quality group (survival time greater than or equal to 36 months) and a survival poor quality group (survival time less than 36 months), the data label set group shown in Table 1 corresponds to the physical condition index data of patients A, B and C, when calculating the classification accuracy of the data label set group shown in Table 1, first, the physical condition index data corresponding to patient A is input into the preset data screening model to obtain the output result corresponding to patient A, that is, the survival time of patient A belongs to the survival quality group, and because the category of the actual survival time of patient A is known, for example, when the actual survival time of patient A is 40 months, it can be determined that the expected output result corresponding to patient A is the survival quality group, at this time, it can be determined that the output result of the model corresponding to patient A is the same as the expected output result, that is, the classification result of the preset data screening model for the survival time of patient A is correct, and the classification result of the preset data screening model for the survival time of patients B and C can be determined in the same way, assuming that the classification result for the survival time of patient B is incorrect (for example, the actual survival time of patient B belongs to the survival quality group, and after inputting the physical condition index data of patient B into the preset data screening model, the output result obtained is that the survival time of patient B belongs to the survival poor quality group, it is determined that the classification result of the preset data screening model for the survival time of patient B is incorrect), the classification result for the survival time of patient C is correct, that is, when the physical condition index data of patients A, B and C corresponding to the data label set group shown in Table 1 is input into the preset data screening model, according to the output result of the model, it can be determined that the classification result of the survival time of two patients is correct, and the classification result of the survival time of one patient is incorrect, therefore, the classification accuracy of the data label set group shown in Table 1 is 2 / 3, about 66.7%, the above example is only for illustration, and the present disclosure is not limited thereto.

[0072] S207, from all the data label set groups, determine the target label set group with the highest classification accuracy, and determine the non-key data label in the target label set group as the associated data label.

[0073] The association data label can include data labels related to the key data label. For example, it is assumed that after S204 is executed, three data label set groups are obtained, which are data label set group 1: (gender, age at dialysis, first dialysis time, hemoglobin, albumin, creatinine, urea, potassium) (as shown in Table 1), data label set group 2: (gender, age at dialysis, first dialysis time, hemoglobin, albumin, creatinine, urea, phosphorus) (as shown in Table 2), and data label set group 3: (gender, age at dialysis, first dialysis time, hemoglobin, albumin, creatinine, urea, potassium, phosphorus) (as shown in Table 3). After the body condition index data of patients A, B and C corresponding to the data labels in the data label set group 1 shown in Table 1 is input into the preset data screening model in sequence, the classification accuracy is 66.67%; after the body condition index data of patients A, B and C corresponding to the data labels in the data label set group 2 shown in Table 2 is input into the preset data screening model in sequence, the classification accuracy is 33.33%; after the body condition index data of patients A, B and C corresponding to the data labels in the data label set group 3 shown in Table 3 is input into the preset data screening model in sequence, the classification accuracy is 100%, at this time, it can be determined that the target label set group is the data label set group 3 (gender, age at dialysis, first dialysis time, hemoglobin, albumin, creatinine, urea, potassium, phosphorus) shown in Table 3, further, it can be determined that the association data label is the non-key data label potassium and phosphorus in the target label set group. The above example is only for illustration, and the present disclosure is not limited thereto.

[0074]

[0075] Table 1

[0076]

[0077] Table 2

[0078]

[0079] Table 3 S208, obtaining a preset data classification training model, training the preset data classification training model through the key data and the non-key data corresponding to the association data label, and obtaining a target data classification model.

[0080] The target data classification model can include:

[0081]

[0082] wherein (x1, x2,...xn represents an input variable of the target data classification model (for example, the input variable can be the physical condition index data), x i may represent the i (i = 1, 2,... n) data in the input variable (x1, x2,... x n represents an output variable of the target classification model (for example, the output variable can be the survival time), b i (i = 1, 2,... n), c i (i = 1, 2,... n), p i (i = 1, 2,... n), a0 represents the parameters of the target data classification model.

[0083] In a possible implementation, the key data and the associated data can be taken as inputs of the preset data classification training model together by a particle swarm algorithm, the preset data classification training model is trained, and model parameters of the classification model are determined, for example, the model parameters can include parameters (that is, b i , c i ) of the membership function in the antecedent network in the fuzzy T-S neural network and coefficients (that is, p i ) of the linear structure in the consequent network in the fuzzy T-S neural network.

[0084] S209, obtaining target data to be classified in the target event, taking the target data as an input of the target data classification model, and obtaining a classification result.

[0085] When the target event is an evaluation event of the hemodialysis timing of a uremia patient, the input of the target data classification model can include data corresponding to key labels such as gender, dialysis age, first dialysis time, hemoglobin, albumin, creatinine, urea, and the like in the physical condition index data of a uremia patient and data corresponding to associated data labels such as potassium, phosphorus, whether edema, whether diabetic nephropathy, whether uremic encephalopathy, and the like in the physical condition index data of the uremia patient determined by a preset data screening model, and the output of the target data classification model can include the survival time (for example, can be in months) of the uremia patient. Further, the survival time of the uremia patient can be classified, at this time, the classification result is the category in which the survival time of the uremia patient is located, for example, the survival time can be classified into two categories, and 36 months is taken as a critical value, when the survival time of the uremia patient is less than 36 months, the uremia patient can be determined as a survival poor group, and when the survival time of the uremia patient is greater than or equal to 36 months, the uremia patient can be determined as a survival good group. Of course, the survival time can be classified into more than two categories, and the present disclosure does not limit this.

[0086] In addition, the critical value can also be determined according to the actual application scenario, for example, the critical value can be set to 12 months or 24 months, or other critical values, and the present disclosure does not limit this.

[0087] It should be noted that in the present example, a mathematical model for evaluating dialysis timing is established based on a machine learning algorithm (such as a fuzzy T-S neural network), thereby providing technical support for establishing a dialysis timing standard for uremic patients, combining data mining methods with medical research, and establishing a systematic blood dialysis evaluation system. Further, the mortality and morbidity of dialysis patients can be reduced, and the medical burden of blood dialysis can be reduced.

[0088] By using the above method, the target data in the target event to be classified can be classified by the target data classification model, so that the target event is evaluated according to the target data, thereby improving the data analysis efficiency and the accuracy of the evaluation of the target event.

[0089] Figure 3 is a block diagram of a data classification device according to an example embodiment, as shown in Figure 3 The device includes:

[0090] The acquisition module 301 is configured to acquire sample data corresponding to a target event under a plurality of data labels.

[0091] The determination module 302 is configured to acquire a preset data screening model, determine a pre-set key data label and a non-key data label from the data label, and determine key data corresponding to the key data label and non-key data corresponding to the non-key data label from the sample data.

[0092] The data screening module 303 is configured to determine an associated data label related to the key data label from the non-key data label according to the key data and the non-key data by using the preset data screening model.

[0093] The model training module 304 is configured to acquire a preset data classification training model, train the preset data classification training model by using the key data and the non-key data corresponding to the associated data label, and obtain a target data classification model.

[0094] The classification module 305 is configured to acquire target data to be classified in the target event, take the target data as an input of the target data classification model, and obtain a classification result.

[0095] Optionally, Figure 4 is a block diagram of a data classification device according to an example embodiment, as shown in Figure 3 The device includes: Figure 4As shown, the data screening module 303 includes: a first determining sub-module 3031, configured to perform permutation and combination processing on the non-key data labels to obtain a plurality of non-key data label sets, wherein different non-key data label sets correspond to different permutation and combination results; a second determining sub-module 3032, configured to combine the key data label with each non-key data label set to obtain a plurality of data label set groups; a third determining sub-module 3033, configured to sequentially take the key data corresponding to the key data label and the non-key data corresponding to the non-key data label in each data label set group as input of a preset data screening model to obtain an output result; a fourth determining sub-module 3034, configured to determine a classification accuracy corresponding to each data label set group according to the output result; and a fifth determining sub-module 3035, configured to determine, from all the data label set groups, a target label set group with the highest classification accuracy, and determine the non-key data labels in the target label set group as the associated data labels.

[0096] Optionally, the fourth determining sub-module 3034 is configured to obtain an expected output result corresponding to each data label set group, and obtain the classification accuracy according to the output result and the expected output result.

[0097] Optionally, the target data classification model includes:

[0098]

[0099] wherein (x1, x2,...x n ) represents an input variable of the target data classification model, x i represents the i-th (i=1, 2,...n) data in (x1, x2,...x n ), Y represents an output variable of the target classification model, b i (i=1, 2,..., n), c i (i=1, 2,..., n), p i (i=1, 2,..., n), and a0 represents a parameter of the target data classification model.

[0100] As to the apparatus in the above-described embodiments, the specific manners in which various modules perform operations have been described in details in the embodiments of the method, and thus will not be described in details here.

[0101] By using the apparatus, the target data to be classified in a target event can be classified by the target data classification model, so that the target event is evaluated according to the target data, and the data analysis efficiency is improved, and the accuracy of the evaluation of the target event is also improved.

[0102] Figure 5is a block diagram of an electronic device 500 according to an exemplary embodiment. As shown, the electronic device 500 can include a processor 501, a memory 502. The electronic device 500 can also include one or more of a multimedia component 503, an input / output (I / O) interface 504, and a communication component 505. Figure 5

[0103] The processor 501 is configured to control overall operations of the electronic device 500 to complete all or part of the steps of the data classification method described above. The memory 502 is configured to store various types of data to support operations of the electronic device 500, which can include, for example, instructions of any application or method operating on the electronic device 500, and application-related data such as contact data, transmitted and received messages, pictures, audio, video, and the like. The memory 502 can be implemented by any type of volatile or non-volatile memory device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk, or an optical disk. The multimedia component 503 can include a screen and an audio component. The screen can be, for example, a touch screen, and the audio component is configured to output and / or input audio signals. For example, the audio component can include a microphone configured to receive external audio signals. The received audio signals can be further stored in the memory 502 or transmitted through the communication component 505. The audio component also includes at least one speaker configured to output audio signals. The I / O interface 504 provides an interface between the processor 501 and other interface modules, which can be a keyboard, a mouse, a button, and the like. The buttons can be virtual buttons or physical buttons. The communication component 505 is configured to perform wired or wireless communication between the electronic device 500 and other devices. The wireless communication, such as Wi-Fi, Bluetooth, near field communication (NFC), 2G, 3G, or 4G, or a combination of one or more of them, so the corresponding communication component 505 can include a Wi-Fi module, a Bluetooth module, an NFC module. ​

[0104] In an exemplary embodiment, the electronic device 500 can be implemented by one or more Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), controller, microcontroller, microprocessor or other electronic elements for performing the above-described data classification method.

[0105] In another exemplary embodiment, a computer readable storage medium including program instructions is also provided, which, when executed by a processor, implements the steps of the above-described data classification method. For example, the computer readable storage medium can be the above-described memory 502 including program instructions, which can be executed by the processor 501 of the electronic device 500 to complete the above-described data classification method.

[0106] The preferred embodiments of the present disclosure are described in detail above with reference to the accompanying drawings, but the present disclosure is not limited to the specific details of the above-described embodiments. Within the technical concept of the present disclosure, various simple modifications can be made to the technical solutions of the present disclosure, and these simple modifications all belong to the protection scope of the present disclosure.

[0107] In addition, it should be noted that each specific technical feature described in the above specific embodiments can be combined in any appropriate manner without contradiction, and in order to avoid unnecessary repetition, the present disclosure will not further describe various possible combinations.

[0108] Furthermore, any combination of the various different embodiments of the present disclosure can also be made, as long as it does not deviate from the idea of the present disclosure, and it should also be considered as disclosed by the present disclosure.

Claims

1. A method for data classification, characterized in that, The method includes: Obtain sample data corresponding to the target event under multiple data labels; the target event includes evaluation events related to the patient's body, and the sample data includes physical condition index data; the data labels are used to represent the attributes of the sample data, and the data labels include gender, age at dialysis, time of first dialysis, hemoglobin, albumin, creatinine, urea, potassium, phosphorus, whether heart failure occurs, whether nausea and vomiting occur, whether edema occurs, whether diabetic nephropathy occurs, and whether uremic encephalopathy occurs; A preset data filtering model is obtained, and pre-set key data labels and non-key data labels are determined from the data labels. Key data corresponding to the key data labels and non-key data corresponding to the non-key data labels are determined from the sample data. The key data are the sample data that must be selected when evaluating the target event, and the non-key data are the sample data that can be selected when evaluating the target event. Based on the key data and the non-key data, the preset data filtering model is used to determine the associated data tags related to the key data tags from the non-key data tags; A preset data classification training model is obtained, and the preset data classification training model is trained using the key data and the non-key data corresponding to the associated data labels to obtain a target data classification model; the target data classification model is used to evaluate the target events related to the patient's body; The target data to be classified in the target event is obtained, and the target data is used as input to the target data classification model to obtain the classification result. The classification result is used to characterize the category of the survival time of the patient corresponding to the target data. Different categories correspond to different ranges of survival time. The target data includes the data corresponding to the key data label and the data corresponding to the associated data label.

2. The method according to claim 1, characterized in that, The step of determining the associated data tags related to the key data tags from the non-key data tags using the preset data filtering model based on the key data and the non-key data includes: The non-critical data tags are permuted and combined to obtain multiple sets of non-critical data tags, wherein different sets of non-critical data tags correspond to different permutation and combination results; The key data tags are combined with each of the non-key data tag sets to obtain multiple data tag set groups; In each data tag set group, the key data corresponding to the key data tag and the non-key data corresponding to the non-key data tag are used as inputs to the preset data filtering model to obtain the output results. The classification accuracy corresponding to each data label set group is determined based on the output results; From all the data label sets, determine the target label set with the highest classification accuracy, and The non-critical data tags in the target tag set group are identified as the associated data tags.

3. The method according to claim 2, characterized in that, Determining the classification accuracy for each data label set group based on the output results includes: Obtain the expected output result for each of the data label sets; The classification accuracy is obtained based on the output result and the expected output result.

4. The method according to any one of claims 1 to 3, characterized in that, The target data classification model includes: in, This represents the input variable of the target data classification model. Indicates the first Here, Y represents the output variable of the target classification model. , , , The parameter represents the target data classification model, where i represents any value from 1 to n.

5. A data classification apparatus, characterized in that, The device includes: The acquisition module is used to acquire sample data corresponding to a target event under multiple data labels; the target event includes evaluation events related to the patient's body, and the sample data includes physical condition index data; the data labels are used to represent the attributes of the sample data, and the data labels include gender, age at dialysis, time of first dialysis, hemoglobin, albumin, creatinine, urea, potassium, phosphorus, whether heart failure occurs, whether nausea and vomiting occur, whether edema occurs, whether diabetic nephropathy occurs, and whether uremic encephalopathy occurs; The determination module is used to acquire a preset data filtering model, determine pre-set key data labels and non-key data labels from the data labels, and determine the key data corresponding to the key data labels and the non-key data corresponding to the non-key data labels from the sample data; the key data are the sample data that must be selected when evaluating the target event, and the non-key data are the sample data that can be selected when evaluating the target event; A data filtering module is used to determine, based on the key data and the non-key data, related data tags associated with the key data tags from the non-key data tags using the preset data filtering model; The model training module is used to acquire a preset data classification training model, and to train the preset data classification training model using the key data and the non-key data corresponding to the associated data labels to obtain a target data classification model; the target data classification model is used to evaluate the target events related to the patient's body; The classification module is used to acquire target data to be classified in the target event, use the target data as input to the target data classification model, and obtain classification results. The classification results are used to characterize the category of the survival time of the patient corresponding to the target data. Different categories correspond to different ranges of survival time. The target data includes data corresponding to the key data tags and data corresponding to the associated data tags.

6. The apparatus according to claim 5, characterized in that, The data filtering module includes: The first determining submodule is used to perform permutation and combination processing on the non-critical data tags to obtain multiple non-critical data tag sets, wherein different non-critical data tag sets correspond to different permutation and combination results; The second determining submodule is used to combine the key data tags with each of the non-key data tag sets to obtain multiple data tag set groups; The third determining submodule is used to sequentially take the key data corresponding to the key data tags and the non-key data corresponding to the non-key data tags in each data tag set as input to the preset data filtering model and obtain the output results. The fourth determination submodule is used to determine the classification accuracy corresponding to each of the data label sets based on the output results; The fifth determination submodule is used to determine the target label set with the highest classification accuracy from all the data label sets, and to determine the non-key data labels in the target label set as the associated data labels.

7. The apparatus according to claim 6, characterized in that, The fourth determining submodule is used to obtain the expected output result corresponding to each of the data label set groups; and to obtain the classification accuracy based on the output result and the expected output result.

8. The apparatus according to any one of claims 5 to 7, characterized in that, The target data classification model includes: in, This represents the input variable of the target data classification model. Indicates the first Here, Y represents the output variable of the target classification model. , , , The parameter represents the target data classification model, where i represents any value from 1 to n.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method described in any one of claims 1-4.

10. An electronic device, characterized in that, include: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method according to any one of claims 1-4.

Images