CN114255830A

CN114255830A - Method for classifying incremental medical data based on rule classifier and related equipment

Info

Publication number: CN114255830A
Application number: CN202111576268.6A
Authority: CN
Inventors: 孙建彬; 王小燕; 赵青松; 赵蕊蕊; 游雅倩; 张涛; 杨克巍; 姜江; 葛冰峰
Original assignee: National University of Defense Technology
Current assignee: National University of Defense Technology
Priority date: 2021-12-21
Filing date: 2021-12-21
Publication date: 2022-03-29

Abstract

The application provides a method for classifying incremental medical data based on a rule classifier and related equipment; the method comprises the following steps: presetting a plurality of categories for a medical data set, and determining the attribute and the category of each sample in the medical data set; determining a classification order when classification is performed for each category; in each classification, calculating the attribute reliability of each attribute, and obtaining reliability sequencing; selecting a preset number of attributes as an initial feature subset according to reliability sequencing, constructing a rule classifier, and classifying samples by using the rule classifier to obtain classification accuracy; adjusting attributes in the initial feature subsets according to the reliability sorting and the classification accuracy to obtain target feature subsets; classifying all categories according to the classification sequence to obtain a target feature subset and a rule classifier of each classification operation; the entire target feature subset and the rule classifier are integrated for classifying new samples in the medical data set.

Description

Method for classifying incremental medical data based on rule classifier and related equipment

Technical Field

The embodiment of the application relates to the technical field of data processing, in particular to a method for classifying incremental medical data based on a rule classifier and related equipment.

Background

In the existing medical data classification, for example, a PSG (polysomnography) evaluation method mostly uses a large number of different monitoring signals as a plurality of characteristics or attributes of data, and evaluates a sleep state based on a numerical value generated by each characteristic, but in practical application, data acquisition often has a precedence order, and conditions for acquiring some attributes are harsh or expensive, for example, the PSG evaluation method, a large number of monitoring signals come from monitoring different parts of a body, and different special instruments are used, which often means that a large amount of cost is required for a patient, and a monitoring process of a large number of signal data needs to be performed in a special place, which brings about very high monitoring cost and data analysis cost.

In some cases, for example, in classification of medical data for PSG evaluation, there are cases where rough classification can be performed using a part of attributes, and whether or not a subsequent attribute needs to be acquired is determined from the result, which greatly saves time and resources, and thus it is necessary to perform classification using existing limited data.

Based on this, there is a need for a scheme that enables efficient classification of medical data with limited data as well as incremental data.

Disclosure of Invention

In view of the above, an object of the present application is to provide a method and related apparatus for classifying incremental medical data based on a rule classifier.

In view of the above, the present application provides a method for classifying incremental medical data based on a rule classifier, applied to a database storing medical data sets, comprising:

presetting a plurality of classes for a medical data set, and determining a plurality of attributes of each sample in the medical data set and a class to which each sample is initially attached;

determining a classification order in which classification is performed for each of the classes among all the classes;

in each classification for a single category, for each attribute, calculating the reliability of the attribute according to the number of samples in each category when the attribute is used independently for classification, and obtaining reliability ranking according to all the reliability of the attribute;

selecting a preset number of attributes from all the attributes according to the reliability sequence to serve as an initial feature subset, constructing the rule classifier by using the initial feature subset, and classifying all the samples by using the rule classifier to obtain the classification accuracy rate of the initial feature subset;

adjusting the attributes in the initial feature subsets according to the reliability ranking and the classification accuracy to obtain target feature subsets;

classifying all the categories according to the classification sequence to obtain the target feature subset and the rule classifier of each classification operation; integrating all of the target feature subsets and the rule classifier for classifying new samples in the medical data set.

Further, in each classification for a single said class, for each said attribute, calculating an attribute reliability for the attribute according to the number of said samples in each said class when the attribute is used individually for classification, resulting in a reliability ranking according to all said attribute reliabilities, comprising:

in each classification for a single said category, performing a binary operation with respect to that category, and not the category;

for each of the attributes, when all the samples are classified by the attribute alone, determining the number of decisions of the samples belonging to the category by the attribute in each of the categories;

determining the attribute reliability of the attribute according to the determined number of the attribute;

and arranging all the attributes according to the attribute reliability of each attribute from top to bottom.

Further, constructing the rule classifier using the initial feature subset includes:

for each attribute of the initial feature subset, matching a plurality of reference values for the attribute, giving corresponding reference evidence for each reference value, and giving corresponding reference evidence weight for each reference evidence;

for each attribute of the initial feature subset, according to the attribute and the reference value corresponding to the attribute, constructing similarity distribution between the attribute and two categories in the two-category operation, and constructing a reference evidence matrix of the attribute according to the similarity distribution;

for each attribute, weighting the reference evidence of the attribute according to the similarity distribution on the attribute and the reference evidence matrix to obtain the attribute evidence of the attribute, weighting the reference evidence weight of the reference evidence of the attribute to obtain the attribute evidence weight of the attribute, and fusing all the attribute evidences by using the reliability and the attribute evidence weight to obtain the rule classifier.

Further, the adjusting the attribute in the initial feature subset by the reliability ranking and the classification accuracy to obtain a target feature subset includes:

selecting one attribute with the highest reliability ranking from the attributes which are not added into the initial feature subset, and adding the selected attribute into the initial feature subset;

classifying all of the samples using all of the attributes in the initial feature subset based on the constructed rule classifier;

calculating the classification accuracy for the initial feature subset based on classification results;

adjusting the attributes in the initial feature subset according to a numerical change of the classification accuracy;

and taking the adjusted initial feature subset as the target feature subset.

Further, adjusting the attributes in the initial feature subset according to a numerical change of the classification accuracy includes:

in response to determining that the classification accuracy is increased, retaining the attributes that were added last, and continuing to select, among the attributes that are not added to the initial feature subset, the one of the attributes with the top reliability ranking to add to the initial feature subset;

in response to determining that the classification accuracy is not increased, removing the attributes that were added last and leaving the attributes contained in the initial feature subset unchanged.

Further, selecting one of the attributes with the top reliability ranking to be added to the initial feature subset comprises:

setting a reliability threshold for all the attributes;

selecting one attribute with the top reliability ranking, and judging whether the reliability of the attribute is greater than the reliability threshold value;

in response to determining that the attribute reliability of the attribute is greater than or equal to a reliability threshold, adding the attribute to the initial feature subset;

in response to determining that the attribute reliability of the attribute is less than a reliability threshold, not adding the attribute to the initial feature subset and leaving the attribute contained in the initial feature subset unchanged.

Further, according to the attribute and the reference value corresponding to the attribute, a similarity distribution between the attribute and two of the categories in the two-category operation is formed, and a reference evidence matrix of the attribute is constructed according to the similarity distribution, including:

calculating, for each of said reference values in each of said attributes, a first difference between that reference value and another of said reference values that is adjacent thereto, over each of said samples; and calculating a second difference between the attribute and another adjacent reference value; determining the matching degree of the attribute and the reference value according to the proportion of the first difference value and the second difference value;

based on the attribute, constructing the similarity distribution using the matching degrees of all the reference values and two of the classes in the two-class operation;

for each of the classes, calculating a sum of the matching degrees belonging to the same attribute and the same reference value among all the samples as a first parameter of the similarity distribution;

for each of the classes, calculating a sum of the first parameters of all the attributes in all the samples as a second parameter of the similarity distribution;

constructing the similarity distribution by using the first parameter, the second parameter and the matching degree;

under the condition that the category is determined, for each reference value matched with each attribute, constructing the likelihood of taking the attribute as the reference value according to the proportion of the first parameter and the second parameter in the similarity distribution;

obtaining reference evidence credibility of each reference evidence when the sample is judged by each reference evidence through the likelihood of respectively normalizing each category;

and under the condition that the attribute is determined, constructing the reference evidence matrix according to the reference evidence credibility and the category of each reference evidence of the attribute.

Further, weighting the reference evidence of the attribute to obtain an attribute evidence of the attribute, and weighting the reference evidence weight of the reference evidence of the attribute to obtain an attribute evidence weight of the attribute, including:

responding to the fact that the attribute value is determined to be between two adjacent reference values based on the distribution of the reference evidence matrix, and obtaining the evidence credibility of the attribute according to the matching degree of the two reference values and the reference evidence credibility; obtaining the attribute evidence weight of the attribute according to the matching degree of the two reference values and the reference evidence weight;

and combining the attribute evidence weight, the evidence credibility and the reliability to obtain the attribute evidence of the attribute.

Further, fusing all the attribute evidences by using the reliability and the attribute evidence weight to obtain the rule classifier, including:

for each attribute evidence, fusing all attribute evidences of the sample by using the reliability factor and the attribute evidence weight according to a fusion rule in an evidence reasoning rule to obtain the discrimination credibility of the attribute evidence for judging the sample to be each category;

and constructing the rule classifier by using each category and the discrimination reliability of the category.

Based on the same inventive concept, the application also provides a device for classifying incremental medical data based on a rule classifier, which is connected with a database storing medical data sets, and comprises: the system comprises a preprocessing module, a single classification module and a global classification module;

the preprocessing module is configured to preset a plurality of classes for a medical data set, and determine a plurality of attributes possessed by each sample in the medical data set and a class to which each sample is initially affiliated.

Further, a classification order in performing classification on each of the classes is determined among all the classes.

The single classification module is configured to, in each classification for a single class, calculate, for each attribute, attribute reliability for the attribute according to the number of samples in each class when the attribute is used individually for classification, and obtain reliability ranking according to all the attribute reliabilities.

Further, a predetermined number of attributes are selected from all the attributes according to the reliability ranking as an initial feature subset, the initial feature subset is used for constructing the rule classifier, and all the samples are classified by the rule classifier, so that the classification accuracy of the initial feature subset is obtained.

Further, the attributes in the initial feature subset are adjusted according to the reliability ranking and the classification accuracy rate to obtain a target feature subset.

The global classification module is configured to classify all the categories according to the classification sequence to obtain the target feature subset and the rule classifier of each classification operation; integrating all of the target feature subsets and the rule classifier for classifying new samples in the medical data set.

Based on the same inventive concept, the present application further provides an electronic device, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the method for classifying incremental medical data based on a rule classifier as described in any one of the above items when executing the program.

Based on the same inventive concept, the present application further provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions for causing the computer to perform the method for classifying incremental medical data based on a rule classifier as described above.

From the above, the method for classifying incremental medical data based on the rule classifier and the related device provided by the application classify each category according to the classification order of actual requirements, classify only one category in each classification, sort all attributes based on the attribute reliability in each classification, establish an initial feature subset only by using the attributes capable of improving the classification accuracy, adjust the attributes in the initial feature subset by comprehensively considering the attribute reliability sorting and the classification accuracy, realize the rough classification by using the first-come data attribute training rule classifier, and then continuously train other classifiers according to the subsequently obtained attributes, and gradually perform more detailed classification.

Drawings

In order to more clearly illustrate the technical solutions in the present application or the related art, the drawings needed to be used in the description of the embodiments or the related art will be briefly introduced below, and it is obvious that the drawings in the following description are only embodiments of the present application, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without creative efforts.

Fig. 1 is a flowchart of a method for classifying incremental medical data based on a rule classifier according to an embodiment of the present application;

FIG. 2 is a block diagram of a device for classifying incremental medical data based on a rule classifier according to an embodiment of the present application;

FIG. 3a is a schematic diagram a illustrating a classification process of incremental medical data based on a rule classifier according to an embodiment of the present application;

FIG. 3b is a schematic diagram b illustrating a classification process of incremental medical data based on a rule classifier according to an embodiment of the present application;

fig. 4 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is further described in detail below with reference to the accompanying drawings in combination with specific embodiments.

It should be noted that technical terms or scientific terms used in the embodiments of the present application should have a general meaning as understood by those having ordinary skill in the art to which the present application belongs, unless otherwise defined. The use of "first," "second," and similar terms in the embodiments of the present application do not denote any order, quantity, or importance, but rather the terms are used to distinguish one element from another. The word "comprising" or "comprises", and the like, means that the element or item listed before the word covers the element or item listed after the word and its equivalents, but does not exclude other elements or items.

As described in the background section, the related method for classifying incremental medical data based on a rule classifier is also difficult to meet the requirements of practical applications.

In the process of implementing the present application, the applicant finds that, in the existing medical data classification, for example, a PSG (polysomnography) evaluation method, a large number of different monitoring signals are mostly used as multiple features or attributes of data, and sleep state is evaluated based on a value generated by each feature, but in practical applications, data acquisition often occurs in a sequential order, and the acquisition conditions of some attributes are harsh or expensive.

For example, in the PSG evaluation method related to medical data classification, a large number of monitoring signals are from different parts of the body, and different special instruments are used, which often means that a large amount of cost is required for patients, and the monitoring process of a large number of signal data needs to be performed at different time intervals, for example, when a patient is going to sleep, data attributes related to the sleep state in the rapid eye movement period are useless, so that the sleep state of the patient can be roughly classified by using partial attributes, and whether subsequent attributes need to be acquired is judged again through the classification result, which greatly saves time and resources.

Taking the polysomnography as an example, in order to save the monitoring cost and the data processing cost, the medical data samples can be roughly classified, only whether the patient is in the sleep state or not is distinguished, when the patient is in the sleep state, the deep sleep or the shallow sleep of the patient is classified according to other data attributes acquired subsequently, and then the data attributes related to the eye movement sleep state and the slow wave sleep state are acquired, so that the medical data are classified into more detailed classes.

It is to be appreciated that the method can be performed by any apparatus, device, platform, cluster of devices having computing and processing capabilities.

Hereinafter, the technical method of the present application will be described in detail by specific examples.

Referring to fig. 1, a method for classifying incremental medical data based on a rule classifier according to an embodiment of the present application includes the following steps:

step S101, presetting a plurality of classes for a medical data set, and determining a plurality of attributes of each sample in the medical data set and a class to which each sample is initially attached; and further determining a classification order when classification is performed on each of the classes among all the classes.

In the embodiment of the present application, taking human sleep health monitoring as a specific example, in the monitoring technology of human sleep health, PSG is mostly adopted as an evaluation means, where PSG needs to monitor multiple physical signs of a patient through multiple sensors to obtain multiple signals related to the sleep state of the monitored patient, for example, in one data sample obtained by each monitoring, dozens of physical signs such as the respiratory rhythm, heart rate, electroencephalogram, electrocardiogram, electromyogram, SPO2 (pulse oxygen saturation) and HR (heart rate) of the patient may be included, and in data obtained by each measurement, the multiple signals are taken as multiple attributes of the data sample obtained by the measurement, and the patient is inferred to be in the sleep state such as deep sleep, shallow sleep or awake according to the dozens of attributes.

In this embodiment, the acquired patient data may be used as a medical data set, the patient data acquired each time may be used as one data sample in the medical data set, which is also referred to as a sample in the present application for short, and different sleep states may be used as a plurality of categories for classifying the sample.

In the present embodiment, as shown in fig. 3a and 3b, the medical data samples are classified a plurality of times, each classification is divided into only one of the above-mentioned classes, that is, in each classification operation, two classification operations are performed.

In one mode, as shown in fig. 3a, the classification result is two categories, namely the specific category and the non-specific category, and according to the attribute obtained by the continuous increment, the sample in the non-specific category is subjected to the binary classification operation of another specific category again until all detailed categories are distinguished at one time.

In another mode, as shown in fig. 3b, two classification operations of rough classification may be performed on the medical data set according to the existing data attributes, and then two classes obtained by the rough classification are continuously refined according to the continuously obtained data attributes, that is, the two classes are gradually classified into more detailed classes.

In this embodiment, the first manner shown in fig. 3a is taken as an example for explanation, and it should be additionally explained that the two manners are only slightly different in classification order, and can be selected according to the situation of a specific scene, and there is no difference in the construction of the rule classifier and the method operation.

Further, there are 6 categories labeled by human sleep health monitoring data, one awake state w, and five sleep states: S1-S5. The sleep state is divided into rapid eye movement period sleep and non-rapid eye movement period sleep, the non-rapid eye movement period is divided into 1 period, 2 periods, 3 periods and 4 periods, wherein the 1 period is a sleep latency period, the 2 period is a light sleep period, and the 3 periods and 4 periods are deep sleep periods. Sample classes S1 through S4 correspond to stages 1, 2, 3, and 4 of the non-rapid eye movement period, and sample class S5 corresponds to the rapid eye movement period. Applicants' studies have found that the sample size for sleep states S1 and S5 is smaller because the non-rapid eye movement periods 1 and 4 are closer to periods 2 and 3, respectively. In order to approximate the sample size of each category and to facilitate the distinction between categories, the non-rapid eye movement periods 1 and 2 are collectively referred to as shallow sleep states, and the non-rapid eye movement periods 3 and 4 are collectively referred to as deep sleep states. After redefining the sleep class, the sleep samples are classified into 4 classes, which are respectively the waking state y₁Shallow sleep state y₂Deep sleep state y₃Fast eye movement sleep state y₄。

In this embodiment, PSG evaluation may involve 12 monitoring signals, the acquisition of which requires multiple sensors, which is considered expensive and inconvenient for most users requiring daily sleep monitoring. The attributes SPO2 and HR which are usually used for distinguishing sleep from wakefulness are relatively easy to acquire, and when the sleep states are classified, if a person is judged to be in the wakefulness state, the sleep state does not need to be determined, and the acquisition and classification processes of the attributes related to other sleep states can be simplified; and further acquiring related attributes after judging that the patient enters the sleep state to determine the sleep state is a light sleep state or other sleep states, and finally determining the sleep state is a deep sleep state or a rapid eye movement period sleep state.

In this embodiment, a total of 3 classifications are determined, with the awake state y first being distinguished₁And non-awake states, i.e. all other sleep states: y is₂、y₃And y₄(ii) a And then distinguish the light sleep state y₂And non-shallow sleep states, i.e., two other sleep states: y is₃And y₄(ii) a Finally distinguishing shallow sleep states y₃And rapid eye movement sleep state y₄And 1 regular classifier needs to be constructed in each of the three classifications.

It should be noted that, this embodiment only uses a limited number of categories as an example, in practical operation, the number of categories may be any number, and in the case that the number of samples is sufficient, the classification of the categories may be performed in an infinitely fine manner.

In this embodiment, the medical data set may be set to have K samples, and for convenience of description, the above 12 attributes provided for each sample are set to be M attributes: x ═ x₁,x₂,...,x_MThe 4 categories of the medical data set are represented as N categories: Θ ═ y₁,…y_n,…,y_NWhere N is 4, y_nIs the nth category, then a sample vector can be represented as: x { (x)₁,x₂,...,x_M),y_nAnd can be decomposed into M sample pairs { (x)_i,y_n),i＝1,2,...,M}。

Further, a plurality of reference values need to be set for each attribute, and in this embodiment, the number of reference values of each attribute is set to 6, and is expressed as:

where i represents the ith attribute, J_i＝6,i＝1,2,...,M。

Further, a corresponding reference evidence is set for each reference value, the reliability of the attribute calculated as follows is equivalent to a reliability factor of the reference evidence about the attribute, and a reference evidence weight is set for each reference evidence.

The value of the reference evidence weight may be determined subjectively after considering a plurality of factors, so as to reflect the relative importance of the reference evidence corresponding to the reference evidence weight compared with other reference evidences.

Step S102, in each classification aiming at a single category, for each attribute, calculating the reliability of the attribute according to the number of samples in each category when the attribute is used independently for classification, and obtaining reliability sequencing according to all the reliability of the attribute.

In the embodiment of the present application, since the operations in each classification are the same operation, only the first classification is described as an example in the present embodiment.

As described above, each classification is performed for a single specific class, but two classification operations are performed, and thus two classes are necessarily obtained.

In each classification, the attribute reliability of all held attributes is calculated firstly, and the attribute reliability is equivalent to the reliability factor of each reference evidence of the attribute.

Specifically, for each attribute, it is necessary to determine the number of samples that can be individually classified by the attribute according to two classes classified this time, and determine the attribute reliability r of the attribute i according to the definition formula of the attribute reliability as shown below_i：

Wherein Q is_iIs possible through attribute x_iDirectly determining the number of samples belonging to each class if

I.e. the attribute reliability r of the attribute i_iIf 1, then attribute x is considered_iFor the most reliable attribute, pass attribute x_iThe most samples that can be directly classified.

After the attribute reliability of all the attributes is calculated, all the attributes are sorted according to the sequence of the attribute reliability from large to small.

Step S103, selecting a preset number of attributes from all the attributes according to the reliability sequence as an initial feature subset, constructing the rule classifier by using the initial feature subset, and classifying all the samples by using the rule classifier to obtain the classification accuracy rate of the initial feature subset.

In an embodiment of the present application, in each classification, a predetermined number of attributes are selected from all the attributes based on the above-determined attribute reliability ranking, and the selected attributes are taken as an initial feature subset.

In this embodiment, each time classification is performed, 2 attributes are first selected from all the attributes as an initial feature subset, and a reliability threshold is set for the initial feature subset.

Further, a rule classifier is constructed based on the attributes in the initial feature subset, the reference values of the attributes, the reference evidences, the reference evidence weights and the reliability factors, and the classification accuracy of the initial feature subset based on the current attributes is calculated.

Specifically, for each attribute in the initial feature subset in the classification, the similarity between the attribute and each reference value thereof can be obtained by using a matching degree calculation formula shown as follows:

α_i,j'＝0,j'＝1,...J_i,j'≠j,j+1

wherein alpha is_i,jAn ith attribute of the M attributes representing the kth sample

And the jth reference value of the attribute

The degree of matching describing the attribute

And a reference value

The degree of similarity between them.

Specifically, as shown in the above formula, on each sample, for each reference value in each attribute, a first difference value of the reference value and another reference value adjacent thereto is calculated; and calculating a second difference between the attribute and another adjacent reference value; and determining the matching degree of the attribute and the reference value according to the proportion of the first difference value and the second difference value.

Due to the intervention of the reference value, the attribute x can be modified_iThe relationship with the category y is approximately converted into the attribute x_iSpecific reference value of

And the relation between the category y, in combination with the matching degree described above, can further combine the sample pair (x) in the medical data set_iY) is transformed and uniquely represented as a similarity distribution for category y as shown in table 1 below:

TABLE 1 sample pairs (x)_iY) in attribute x_iDistribution of similarity of

Wherein, a_n,jIs an attribute value

And a reference value

Can calculate the matching degree and belongs to y_nThe sum of the matching degrees of all sample pairs of the class is taken as a first parameter.

Is in the category y_nThe sum of the matching degrees of all the reference values of the sample pairs of (1) is taken as a second parameter.

Is at a reference value

In the case of determination, for attribute values

With reference values in all classes of sample pairs

Is taken as a third parameter, and has

The first parameter, the second parameter, and the third parameter together form the similarity distribution as shown above.

According to Table 1Knowing that in case of class determination we can construct attribute x_iIs a reference value

Likelihood of (1), denoted as c_n,j：

Further, reference evidence for each reference value may be determined

Is defined as: when attribute x_iIs taken as a reference value

Then, the sample is determined as type y_nHas a reference evidence confidence of

And, the reliability of the above attributes is equated as a reliability factor of the reference evidence

Further, reference evidence

Reference evidence confidence of

The likelihood of normalization can be found using the following equation:

further, according to the above calculation, for each attribute x_iA reference evidence matrix for this property can be obtained as shown in table 2:

TABLE 2 Attribute x_iReference evidence matrix of

Further, according to the distribution of the reference evidence matrix, for the kth sample with M attributes in this embodiment:

if its ith attribute

Value in the interval

In that case, the activation and reference values are compared

And

corresponding two adjacent reference evidences in the reference evidence matrix

And

and weighted to compute the attribute evidence for that attribute.

Specifically, as shown in the following formula, the reference evidence is completed by using the matching degree of the attribute and the two reference values and the reference evidence credibility of the two reference evidences

And

weighting of (2):

wherein the result p is obtained_n,iRepresents: in that

Is in the interval

Internal, i.e. activation

And

in the case of (2), the class of the sample k is regarded as y_nEvidence confidence of, i.e. about, attributes

Attribute evidence e of (1)_iThe evidence confidence of (1).

Further, for reference evidence

And

reference evidence weight of

And

weighting is done in the same way, taking the formula shown below:

wherein the result w is obtained_iRepresents: about attributes

Attribute evidence e of (1)_iThe attribute evidence weight of (1).

The attribute reliability of the attribute obtained by the calculation is equivalent to the reliability factor of the attribute evidence, and the attribute evidence weight, the reliability factor and the evidence credibility of the attribute evidence are combined to obtain the complete attribute evidence e_i。

Further, according to the above manner, all M pieces of attribute evidence of M attributes can be obtained: e.g. of the type₁,e₂,...,e_M。

Further, based on a fusion rule in the evidence reasoning rule, combining the attribute evidence weight and the reliability factor to fuse the M pieces of attribute evidence.

Wherein, two attribute pieces of evidence e independent of each other₁And e₂Can be fused by an evidence reasoning fusion rule to obtain e₁And e₂Belief function p in combination with jointly supporting propositions theta_θ,e(2)The calculation method is shown as the following formula:

when M evidences are fused, jointly supporting the reliability function M of proposition theta_θ,e(i)Can also be expressed as:

after the M pieces of evidence are fused by combining the attribute evidence weight and the reliability factor, the following fusion functions about the sample category are obtained and are used as a rule classifier to classify the medical data set:

P(x^k)＝{(y_n,p_n,e(M)),n＝1,...,N}

wherein, y_nDenotes the nth class, p_n,e(M)Indicating that the sample belongs to class y_nConfidence of (2) is determined from the fused result P (x)^k) Judging the sample x^kClass is maximum confidence p_n,e(m)Corresponding category y_n。

Based on the obtained classification result, an accuracy calculation formula for describing the classification performance of the current rule classifier is constructed and expressed as follows:

wherein totalNum represents the total number of samples participating in the classification operation, rightNum represents the number of samples correctly classified, and the accuracy describes the classification capability of the samples based on the existing attributes of the initial feature subset, namely the classification performance of the rule classifier constructed based on the existing attributes of the initial feature subset.

And S104, adjusting the attributes in the initial feature subset according to the reliability sequence and the classification accuracy rate to obtain a target feature subset.

In the embodiment of the present application, the top 1 attribute with reliability ranking is selected from the other remaining attributes that are not added to the initial feature subset, and is compared with the reliability threshold set above.

Further, when the attribute is selected to be less than the reliability threshold, it is not added to the initial feature subset and the existing attribute of the initial feature subset is maintained.

And when the attribute is selected to be more than or equal to the reliability threshold, adding the attribute into the value initial feature subset, constructing a rule classifier by using the current initial feature subset according to the same mode, and calculating the classification accuracy of the current initial feature subset.

Based on the obtained classification accuracy, if the classification accuracy is not increased, removing the attribute added into the initial feature subset at last, keeping the attribute in the initial feature subset, and finishing the classification.

If the classification accuracy is increased, the attribute added to the initial feature subset last is retained, and the top 1 attribute with the highest reliability rank is selected again from the other remaining attributes not added to the initial feature subset, and the above comparison of the reliability thresholds and the comparison of the classification accuracy in step S104 are repeated again to determine whether the attribute added again is retained in the initial feature subset.

Further, if all attributes are added to the initial feature subset, the classification accuracy still remains to be increased, but since no redundant attributes can be added continuously, the current classification operation can also be ended.

Based on the classification operations described above, the samples in the medical data set may be classified into two categories: awake state y₁And includes a light sleep state y₂Deep sleep state y₃Rapid eye movement sleep state y₄The other category.

Step S105, classifying all the categories according to the classification sequence to obtain the target feature subset and the rule classifier of each classification operation; integrating all of the target feature subsets and the rule classifier for classifying new samples in the medical data set.

In the embodiment of the present application, as shown in fig. 3a, the remaining three categories of all 4 categories are classified in the same operation as the above-described step S103 and step S104.

Further, in the second classification and the third classification, since all attributes need to be sorted for attribute reliability when the initial feature subset addition attribute is established in each classification, in addition to classifying the sample based on the existing attribute, in some cases, if a newly added attribute is provided, the newly added attribute can also be sorted for attribute reliability, thereby realizing the screening of the incremental attribute.

After all three classification operations are completed, 3 feature subsets, as shown in fig. 3a, and three regular classifiers can be obtained.

Further, all feature subsets and all rule classifiers may be integrated for classifying new samples in the medical data set.

Therefore, the incremental medical data classification method based on the rule classifier and the related equipment provided by the application classify each category according to the classification order of actual requirements, classify only one category in each classification, sort all attributes based on the attribute reliability in each classification, establish an initial feature subset only by using the attributes capable of improving the classification accuracy, adjust the attributes in the initial feature subset by comprehensively considering the attribute reliability sorting and the classification accuracy, realize the rough classification by using the first-come data attribute training rule classifier, then continuously train other classifiers according to the subsequently acquired attributes, and gradually carry out more detailed classification.

It should be noted that the method of the embodiments of the present application may be executed by a single device, such as a computer or a server. The method of the embodiment can also be applied to a distributed scene and completed by the mutual cooperation of a plurality of devices. In such a distributed scenario, one of the devices may only perform one or more steps of the method of the embodiments of the present application, and the devices may interact with each other to complete the method.

It should be noted that the above describes some embodiments of the present application. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments described above and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.

Based on the same inventive concept, corresponding to any embodiment method, the embodiment of the application also provides a device for classifying the incremental medical data based on the rule classifier.

Referring to fig. 2, the apparatus for classifying incremental medical data based on a rule classifier, connected to a database storing medical data sets, includes: a preprocessing module 201, a single classification module 202 and a global classification module 203;

the preprocessing module 201 is configured to preset a plurality of classes for a medical data set, and determine a plurality of attributes possessed by each sample in the medical data set and a class to which each sample is initially affiliated.

The single-classification module 202 is configured to, in each classification for a single class, calculate, for each attribute, the reliability of the attribute according to the number of samples in each class when the attribute is used individually for classification, and obtain a reliability ranking according to all the reliability of the attribute.

The global classification module 203 is configured to classify all the categories according to the classification order, so as to obtain the target feature subset and the rule classifier of each classification operation; integrating all of the target feature subsets and the rule classifier for classifying new samples in the medical data set.

For convenience of description, the above devices are described as being divided into various modules by functions, and are described separately. Of course, the functions of the modules may be implemented in the same or multiple software and/or hardware when implementing the embodiments of the present application.

The apparatus of the foregoing embodiment is used to implement the corresponding method for classifying incremental medical data based on a rule classifier in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiment, which are not described herein again.

Based on the same inventive concept, corresponding to any of the above embodiments, the embodiments of the present application further provide an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the program, the processor implements the method for classifying incremental medical data based on a rule classifier as described in any of the above embodiments.

Fig. 4 is a schematic diagram illustrating a more specific hardware structure of an electronic device according to this embodiment, where the electronic device may include: a processor 1010, a memory 1020, an input/output interface 1030, a communication interface 1040, and a bus 1050. Wherein the processor 1010, memory 1020, input/output interface 1030, and communication interface 1040 are communicatively coupled to each other within the device via bus 1050.

The processor 1010 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an Application Specific Integrated Circuit (ASIC), or one or more Integrated circuits, and is configured to execute related programs to implement the technical solutions provided in the embodiments of the present Application.

The Memory 1020 may be implemented in the form of a ROM (Read Only Memory), a RAM (Random Access Memory), a static storage device, a dynamic storage device, or the like. The memory 1020 may store an operating system and other application programs, and when the technical solution provided by the embodiment of the present application is implemented by software or firmware, the relevant program codes are stored in the memory 1020 and called to be executed by the processor 1010.

The input/output interface 1030 is used for connecting an input/output module to input and output information. The i/o module may be configured as a component in a device (not shown) or may be external to the device to provide a corresponding function. The input devices may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output devices may include a display, a speaker, a vibrator, an indicator light, etc.

The communication interface 1040 is used for connecting a communication module (not shown in the drawings) to implement communication interaction between the present apparatus and other apparatuses. The communication module can realize communication in a wired mode (such as USB, network cable and the like) and also can realize communication in a wireless mode (such as mobile network, WIFI, Bluetooth and the like).

Bus 1050 includes a path that transfers information between various components of the device, such as processor 1010, memory 1020, input/output interface 1030, and communication interface 1040.

It should be noted that although the above-mentioned device only shows the processor 1010, the memory 1020, the input/output interface 1030, the communication interface 1040 and the bus 1050, in a specific implementation, the device may also include other components necessary for normal operation. Furthermore, it will be understood by those skilled in the art that the above-described apparatus may also include only those components necessary to implement the embodiments of the present application, and not necessarily all of the components shown in the figures.

Based on the same inventive concept, corresponding to any of the above-described embodiment methods, the present application further provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the method for classifying rule classifier based incremental medical data according to any of the above embodiments.

Computer-readable media of the present embodiments, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device.

The storage medium of the above embodiment stores computer instructions for causing the computer to execute the method for classifying incremental medical data based on a rule classifier according to any of the above embodiments, and has the beneficial effects of corresponding method embodiments, which are not described herein again.

Those of ordinary skill in the art will understand that: the discussion of any embodiment above is meant to be exemplary only, and is not intended to intimate that the scope of the disclosure, including the claims, is limited to these examples; within the context of the present application, features from the above embodiments or from different embodiments may also be combined, steps may be implemented in any order, and there are many other variations of the different aspects of the embodiments of the present application as described above, which are not provided in detail for the sake of brevity.

In addition, well-known power/ground connections to Integrated Circuit (IC) chips and other components may or may not be shown in the provided figures for simplicity of illustration and discussion, and so as not to obscure the embodiments of the application. Furthermore, devices may be shown in block diagram form in order to avoid obscuring embodiments of the application, and this also takes into account the fact that specifics with respect to implementation of such block diagram devices are highly dependent upon the platform within which the embodiments of the application are to be implemented (i.e., specifics should be well within purview of one skilled in the art). Where specific details (e.g., circuits) are set forth in order to describe example embodiments of the application, it should be apparent to one skilled in the art that embodiments of the application can be practiced without, or with variation of, these specific details. Accordingly, the description is to be regarded as illustrative instead of restrictive.

While the present application has been described in conjunction with specific embodiments thereof, many alternatives, modifications, and variations of these embodiments will be apparent to those of ordinary skill in the art in light of the foregoing description. For example, other memory architectures (e.g., dynamic ram (dram)) may use the discussed embodiments.

The embodiments of the present application are intended to embrace all such alternatives, modifications and variances that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, substitutions, improvements, and the like that may be made without departing from the spirit and principles of the embodiments of the present application are intended to be included within the scope of the present application.

Claims

1. A method for classifying incremental medical data based on a rule classifier is applied to a database storing medical data sets, and comprises the following steps:

2. The method of claim 1, wherein in each classification for a single said class, for each said attribute, calculating an attribute reliability for the attribute based on the number of said samples in each said class when classified using the attribute alone, resulting in a reliability ranking based on all said attribute reliabilities, comprises:

3. The method of claim 2, wherein said constructing the rule classifier using the initial subset of features comprises:

4. The method according to any one of claims 1 to 3, wherein adjusting the attributes in the initial feature subset according to the reliability ranking and the classification accuracy to obtain a target feature subset comprises:

and taking the adjusted initial feature subset as the target feature subset.

5. The method of claim 4, wherein said adjusting said attributes in said initial subset of features according to a numerical change in said classification accuracy comprises:

6. The method of claim 4, wherein said selecting said one of said attributes with said reliability ranking top to be added to said initial feature subset comprises:

setting a reliability threshold for all the attributes;

7. The method according to claim 3, wherein constructing a similarity distribution between the attribute and two of the categories in the two classification operations according to the attribute and the reference value corresponding to the attribute, and constructing a reference evidence matrix of the attribute according to the similarity distribution comprises:

8. The method according to claim 7, wherein the weighting the reference evidence of the attribute to obtain the attribute evidence of the attribute, and weighting the reference evidence weight of the reference evidence of the attribute to obtain the attribute evidence weight of the attribute comprises:

9. The method according to claim 8, wherein said fusing all of said attribute evidences using said reliability and said attribute evidence weights to obtain said rule classifier comprises:

10. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable by the processor, characterized in that the processor implements the method according to any of claims 1 to 9 when executing the computer program.