CN109767819A

CN109767819A - Group technology, device, storage medium and the electronic equipment of case history

Info

Publication number: CN109767819A
Application number: CN201811512148.8A
Authority: CN
Inventors: 王阳; 赵立军; 张霞
Original assignee: Neusoft Corp
Current assignee: Neusoft Corp
Priority date: 2018-12-11
Filing date: 2018-12-11
Publication date: 2019-05-17
Anticipated expiration: 2038-12-11
Also published as: CN109767819B

Abstract

This disclosure relates to a kind of group technology of case history, device, storage medium and electronic equipment, this method comprises: extracting all target signatures wait be grouped in case history；According to the characteristic probability network pre-established and all target signatures, obtain the total energy value that DRG group corresponds to the case history to be grouped, this feature probability net is using DRG group and feature group as node, using the incidence relation between DRG group and feature group as side, with the network topology structure for the weight that the correlation probabilities value between DRG group and feature group is the side, which is the summation of multiple correlation probabilities values between multiple feature groups belonging to DRG group and all target signatures；Determine that the DRG group for having maximum total energy value is to be somebody's turn to do the corresponding target DRG group of case history to be grouped.Case history can be identified and is grouped by the network structure established according to the correlation of feature and DRG group, avoided the step of being manually grouped, improve the efficiency and accuracy rate of case history grouping.

Description

Group technology, device, storage medium and the electronic equipment of case history

Technical field

This disclosure relates to medical care evaluation field, and in particular, to a kind of group technology of case history, device, storage medium and Electronic equipment.

Background technique

DRG (Diagnosis Related Groups diagnoses associated packets) system is that the generally acknowledged comparison in the world today is first Into one of the means of payment, be a kind of patient classification's scheme, be used exclusively for the sorting code number mark of medical insurance advance payment system It is quasi-.The system is according to the age of patient, gender, length of stay, clinical diagnosis, illness, operation, disease severity, complication With complication and the factors such as lapse to medical care evaluation carried out to the case history of patient, and then case history is divided to 500-600 DRG Group, and by science measuring and calculating, give imprest money.That is, it is that Medical Insurance Organizations are paid with regard to disease that DRG system is practical Standard is reached an agreement with hospital, and when accepting the patient for participating in medical insurance for medical treatment, Medical Insurance Organizations are according to the pre- of the disease for hospital Standardized payment pays expense, a kind of Payment system undertaken beyond part by hospital to hospital.In recent years, with medical expense It rises steadily, the pressure of medical insurance is gradually increased, and DRG scheme is gradually introduced medicare system to medical resource by national governments It is managed and supervises.In the related technology, usually case history is divided according to the content (i.e. feature) in case history by manual type To different DRG groups, recycling one-class SVM, (one-class Support Vector Machine, single classification are supported Vector machine) or isolated forest abnormal case history is identified in DRG group.But one-class SVM is affected by super ginseng, Isolated forest can only handle a small amount of continuous variable.And both of which is only to find out the exception being grouped in case history Point can not provide more accurate grouping information.

Summary of the invention

To overcome the problems in correlation technique, purpose of this disclosure is to provide a kind of group technology of case history, device, Storage medium and electronic equipment.

To achieve the goals above, according to the first aspect of the embodiments of the present disclosure, a kind of group technology of case history, institute are provided The method of stating includes:

Extract all target signatures wait be grouped in case history；

According to the characteristic probability network and all target signatures pre-established, obtains the first DRG group and correspond to institute The total energy value of case history to be grouped is stated, the characteristic probability network is using DRG group and feature group as node, with DRG group and feature Incidence relation between group is side, the net established with the weight that the correlation probabilities value between DRG group and feature group is the side Network topological structure, the first DRG group are any DRG group in multiple DRG groups that the library DRG includes, and the total energy value is institute The summation of multiple correlation probabilities values between multiple feature groups belonging to the first DRG group and all target signatures is stated, often A feature group includes to meet multiple features of same grouping condition；

Determine that the DRG group for having the maximum total energy value is the corresponding target DRG group of the case history to be grouped.

Optionally, each of described library DRG DRG group correspond to it is multiple be grouped case history, it is each described to be grouped disease It include multiple features in going through, the multiple feature corresponds to multiple feature classes, in all targets extracted wait be grouped in case history Before feature, the method also includes:

It is directed to fisrt feature class, corresponding multiple be grouped in case history of the 2nd DRG group is obtained and belongs to the fisrt feature All sample characteristics of class, the 2nd DRG group are any DRG group in the library DRG, and the fisrt feature class is described more Any feature class in a feature class；

According to the corresponding grouping condition of the fisrt feature class, all sample characteristics are divided into multiple feature groups, So that the quantity of sample characteristics is in default distribution in the multiple feature group；

According to the quantity and the default distribution of sample characteristics in each feature group in the multiple feature group Corresponding probability density function obtains related between the 2nd DRG group and each feature group in the multiple feature group Property probability value；

Getting the correlation probabilities between the multiple DRG group and the corresponding all feature groups of the multiple DRG group After value, using the multiple DRG group and the corresponding all feature groups of the multiple DRG group as node, with the multiple DRG group and Incidence relation between the corresponding all feature groups of the multiple DRG group is side, with the multiple DRG group and the multiple DRG Correlation probabilities value between the corresponding all feature groups of group is the weight on the side, establishes the characteristic probability network.

Optionally, the characteristic probability network and all target signatures that the basis pre-establishes obtain the first DRG Group corresponds to the total energy value of the case history to be grouped, comprising:

Determine that there are all fisrt feature groups of incidence relation with the first DRG group according to the characteristic probability network；

The determining multiple second feature groups to match with all target signatures in all fisrt feature groups；

Multiple phases between the first DRG group and the multiple second feature group are determined according to the characteristic probability network The summation of closing property probability value corresponds to the total energy value of the case history to be grouped as the first DRG group.

Optionally, the feature Value Types of the feature are discrete type or continuous type, and each feature class is corresponding all Feature has same feature Value Types, when the feature Value Types of the corresponding all sample characteristics of the fisrt feature class are discrete type When, it is described according to the corresponding grouping condition of the fisrt feature class, all sample characteristics are divided into multiple feature groups, with The quantity for making sample characteristics in the multiple feature group is in default distribution, comprising:

The sample characteristics for having same characteristic features value in all sample characteristics are divided into a feature group, to obtain State multiple feature groups；

The multiple feature group is arranged, so that the quantity of sample characteristics is in described default in the multiple feature group Distribution；

It is the multiple characteristic component with number according to the sequence of arrangement.

Optionally, described when the feature Value Types of the corresponding all sample characteristics of the fisrt feature class are continuous type According to the corresponding grouping condition of the fisrt feature class, all sample characteristics are divided into multiple feature groups, so that described The quantity of sample characteristics is in default distribution in multiple feature groups, comprising:

It obtains and has the sample characteristics of maximum eigenvalue in all sample characteristics and have the sample of minimal eigenvalue Feature；

Multiple value intervals are separated between the maximum eigenvalue and the minimal eigenvalue etc.；

The sample characteristics that the same value interval is in all sample characteristics are divided to the same feature group, To obtain the multiple feature group；

It is that the multiple characteristic component is matched according to the size of the endpoint value of the corresponding value interval of the multiple feature group Number.

Optionally, the default distribution is normal distribution, each feature according in the multiple feature group The quantity of target signature and the corresponding probability density function of the default distribution in group, obtain the 2nd DRG group and Correlation probabilities value between the corresponding each feature group of the 2nd DRG group, comprising:

By input variable of the number as normpdf of each feature group, to obtain described the The correlation probabilities value between each feature group in two DRG groups and the multiple feature group；Wherein, the normal distribution probability Density function includes:

Wherein, x is input variable, and σ is the standard deviation of the quantity of target signature in the multiple feature group, and μ is described more The average value of the quantity of sample characteristics in a feature group.

Optionally, it is directed to the target signature that feature Value Types in all target signatures are discrete type, it is described in institute State the determining multiple second feature groups to match with all target signatures in all fisrt feature groups, comprising:

Determine second feature class belonging to the target signature；

The multiple third feature groups for corresponding to the second feature class are determined in all fisrt feature groups；

It is determining in the multiple third feature group to have spy locating for the feature of same characteristic features value with the target signature Sign group, as the second feature group to match with the fisrt feature；

The determining second feature group to match with each target signature, as the multiple second feature group.

Optionally, it is directed to the target signature that feature Value Types in all target signatures are continuous type, it is described in institute State the determining multiple second feature groups to match with all target signatures in all fisrt feature groups, comprising:

Determine third feature class belonging to the target signature；

The multiple fourth feature groups for corresponding to the third feature class are determined in all fisrt feature groups, it is described more A fourth feature group is to be separated according to the maximum eigenvalue of the corresponding all features of the third feature class and minimal eigenvalue etc. Multiple value intervals；

The first value interval locating for the target signature is determined in the multiple value interval；

Corresponding with first value interval feature group is determined in the multiple fourth feature group, as with the mesh The second feature group that mark feature matches；

According to the second aspect of an embodiment of the present disclosure, a kind of apparatus for grouping of case history is provided, described device includes:

Characteristic extracting module, for extracting all target signatures wait be grouped in case history；

Energy value obtains module, for obtaining according to the characteristic probability network and all target signatures pre-established The first DRG group is taken to correspond to the total energy value of the case history to be grouped, the characteristic probability network is to be with DRG group and feature group Node, using the incidence relation between DRG group and feature group as side, using the correlation probabilities value between DRG group and feature group as institute The network topology structure that the weight on side is established is stated, the first DRG group is any DRG in multiple DRG groups that the library DRG includes Group, multiple phases of the total energy value between the first DRG group and multiple feature groups belonging to all target signatures The summation of closing property probability value, each feature group include to meet multiple features of same grouping condition；

Case history grouping module, for determining that the DRG group for having the maximum total energy value is the case history pair to be grouped The target DRG group answered.

Optionally, each of described library DRG DRG group correspond to it is multiple be grouped case history, it is each described to be grouped disease It include multiple features in going through, the multiple feature corresponds to multiple feature classes, described device further include:

Sample acquisition module obtains that the 2nd DRG group is corresponding multiple to be grouped in case history for being directed to fisrt feature class Belonging to all sample characteristics of the fisrt feature class, the 2nd DRG group is any DRG group in the library DRG, described the One feature class is any feature class in the multiple feature class；

Feature grouping module is used for according to the corresponding grouping condition of the fisrt feature class, by all sample characteristics Multiple feature groups are divided into, so that the quantity of sample characteristics is in default distribution in the multiple feature group；

Correlation determining module, for the quantity according to sample characteristics in each feature group in the multiple feature group, And the corresponding probability density function of the default distribution, obtain the 2nd DRG group in the multiple feature group Correlation probabilities value between each feature group；

Network establishes module, for getting the multiple DRG group and the corresponding all feature groups of the multiple DRG group Between correlation probabilities value after, using the multiple DRG group and the corresponding all feature groups of the multiple DRG group as node, Using the incidence relation between the multiple DRG group and the corresponding all feature groups of the multiple DRG group as side, with the multiple Correlation probabilities value between DRG group and the corresponding all feature groups of the multiple DRG group is the weight on the side, described in foundation Characteristic probability network.

Optionally, the energy value obtains module, comprising:

Fisrt feature group determines submodule, for existing according to the characteristic probability network is determining with the first DRG group All fisrt feature groups of incidence relation；

Second feature group determines submodule, for determining and all target signatures in all fisrt feature groups The multiple second feature groups to match；

Energy value computational submodule, for according to the characteristic probability network determine the first DRG group with it is the multiple The summation of multiple correlation probabilities values between second feature group corresponds to the case history to be grouped as the first DRG group Total energy value.

Optionally, the feature Value Types of the feature are discrete type or continuous type, and each feature class is corresponding all Feature has same feature Value Types, the feature grouping module, comprising:

Fisrt feature group acquisition submodule, the sample characteristics for will have same characteristic features value in all sample characteristics It is divided into a feature group, to obtain the multiple feature group；

Feature group arranges submodule, for arranging the multiple feature group, so that sample in the multiple feature group The quantity of eigen is in the default distribution；

Fisrt feature group # submodule, for being the multiple characteristic component with number according to the sequence of arrangement.

Optionally, the feature grouping module, comprising:

Characteristic value acquisition submodule, for obtain have in all sample characteristics maximum eigenvalue sample characteristics and The sample characteristics for having minimal eigenvalue；

Interval division submodule, for separating multiple values between the maximum eigenvalue and the minimal eigenvalue etc. Section；

Second feature group acquisition submodule, for the sample of the same value interval will to be in all sample characteristics Eigen is divided to the same feature group, to obtain the multiple feature group；

Second feature group # submodule, for according to the endpoint value of the corresponding value interval of the multiple feature group Size is the multiple characteristic component with number.

Optionally, the default distribution is normal distribution, and the correlation determining module is used for:

Optionally, it is directed to the target signature that feature Value Types in all target signatures are discrete type, described second Feature group determines submodule, is used for:

Determine second feature class belonging to the target signature；

Optionally, it is directed to the target signature that feature Value Types in all target signatures are continuous type, described second Feature group determines submodule, is used for:

Determine third feature class belonging to the target signature；

According to the third aspect of an embodiment of the present disclosure, a kind of computer readable storage medium is provided, calculating is stored thereon with Machine program realizes the group technology for the case history that embodiment of the present disclosure first aspect provides when the computer program is executed by processor The step of.

According to a fourth aspect of embodiments of the present disclosure, a kind of electronic equipment is provided, comprising:

Memory is stored thereon with computer program；

Processor, for executing the computer program in the memory, to realize embodiment of the present disclosure first party The step of group technology for the case history that face provides.

Through the above technical solutions, the disclosure can extract all target signatures wait be grouped in case history；According to building in advance Vertical characteristic probability network and above-mentioned all target signatures obtain the gross energy that the first DRG group corresponds to the case history to be grouped Value, this feature probability net be using DRG group and feature group as node, using the incidence relation between DRG group and feature group as side, with Correlation probabilities value between DRG group and feature group is the network topology structure that the weight on the side is established, and the first DRG group is Any DRG group in multiple DRG groups that the library DRG includes, the total energy value are the first DRG group and above-mentioned all target signature institutes The summation of the multiple correlation probabilities values between multiple feature groups belonged to, each this feature group include to meet same grouping condition Multiple features；Determine that the DRG group for having the maximum total energy value is to be somebody's turn to do the corresponding target DRG group of case history to be grouped.It can lead to It crosses and case history is identified and is grouped according to the network structure that the correlation of feature and DRG group is established, avoid the step being manually grouped Suddenly, the efficiency and accuracy rate of case history grouping are improved.

Other feature and advantage of the disclosure will the following detailed description will be given in the detailed implementation section.

Detailed description of the invention

Attached drawing is and to constitute part of specification for providing further understanding of the disclosure, with following tool Body embodiment is used to explain the disclosure together, but does not constitute the limitation to the disclosure.In the accompanying drawings:

Fig. 1 is a kind of flow chart of the group technology of case history shown according to an exemplary embodiment；

Fig. 2 is the flow chart for implementing the group technology of another case history exemplified according to Fig. 1；

Fig. 3 is the flow chart for implementing a kind of total energy value acquisition methods exemplified according to Fig.2,；

Fig. 4 is the flow chart for implementing a kind of feature partition method exemplified according to Fig.2,；

Fig. 5 is the flow chart for implementing another feature partition method exemplified according to Fig.2,；

Fig. 6 is a kind of block diagram of the apparatus for grouping of case history shown according to an exemplary embodiment；

Fig. 7 is the block diagram for implementing the apparatus for grouping of another case history exemplified according to Fig.6,；

Fig. 8 is that a kind of energy value for implementing to exemplify according to Fig.7, obtains the block diagram of module；

Fig. 9 is the block diagram for implementing a kind of feature grouping module exemplified according to Fig.7,；

Figure 10 is the block diagram for implementing another feature grouping module exemplified according to Fig.7,；

Figure 11 is the block diagram of a kind of electronic equipment shown according to an exemplary embodiment.

Specific embodiment

Example embodiments are described in detail here, and the example is illustrated in the accompanying drawings.Following description is related to When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment Described in embodiment do not represent all implementations consistent with this disclosure.On the contrary, they be only with it is such as appended The example of the consistent device and method of some aspects be described in detail in claims, the disclosure.

Before introducing the grouping of case history of disclosure offer, first to target involved in embodiment each in the disclosure Application scenarios are introduced, which includes that (Diagnosis Related Groups, diagnosis is related to divide a DRG Group) system, which stores multiple DRG groups, and each DRG group, which corresponds to, multiple has been grouped case history.

Fig. 1 is a kind of flow chart of the group technology of case history shown according to an exemplary embodiment, as shown in Figure 1, answering For above-mentioned DRG system, this method comprises:

Step 101, all target signatures wait be grouped in case history are extracted.

Illustratively, being somebody's turn to do case history to be grouped is the case history that hospital is committed to the DRG system, is needed through the DRG system to this Case history to be grouped is grouped, with the medical resource spent according to the determining case history to be grouped of the DRG group divided.Above-mentioned All target signatures are should be wait be grouped in case history documented content under each column, wherein each column is a feature class, the spy Levy class may include: " symptom ", " drug number ", " treatment means ", " whether performing the operation ", " hospitalization cost ", " Operation Fee With " and " length of stay " etc..In other words, which is that each wait be grouped, " symptom " above-mentioned in case history, " drug is compiled Number " and the columns such as " hospitalization cost " under the information inserted.

Step 102, according to the characteristic probability network and above-mentioned all target signatures pre-established, the first DRG group is obtained Total energy value corresponding to the case history to be grouped.

Wherein, this feature probability net is using DRG group and feature group as node, with being associated between DRG group and feature group Relationship is side, and with the network topology structure that the weight that the correlation probabilities value between DRG group and feature group is the side is established, this One DRG group is any DRG group in multiple DRG groups that the library DRG includes.Each DRG group in the library DRG corresponds to multiple be grouped Case history, each this have been grouped in case history comprising multiple features, and above-mentioned multiple features correspond to multiple feature classes.It is understood that Before above-mentioned steps 102, it is thus necessary to determine that in the DRG system existing multiple DRG groups and it is multiple be grouped it is all in case history Correlation degree between feature, then in a step 102 according to the existing spy in the target signature and DRG system wait be grouped in case history The corresponding relationship of sign determines the correlation degree (i.e. total energy value) of the target signature and each DRG group in the case history to be grouped, should Correlation degree is the foundation being grouped in the following steps 103 to the case history to be grouped.The total energy value is the first DRG group The summation of multiple correlation probabilities values between multiple feature groups belonging to above-mentioned all target signatures, each this feature group packet Containing the multiple features for meeting same grouping condition.It is multiple to be grouped in case history existing that the statement of " feature group " herein is practical The feature repeatedly occurred, to be the corresponding feature of this feature group multiple be grouped the quantity of feature existing in feature group The number occurred in case history.

Step 103, determine that the DRG group for having the maximum total energy value is to be somebody's turn to do the corresponding target DRG group of case history to be grouped.

It illustratively, can after the total energy value of each DRG group in the DRG system obtained in through the above steps 102 It is compared with the total energy value to all DRG groups, and it is maximum (i.e. most with the case history degree of correlation to be grouped to select total energy value High) target DRG group, and then the case history to be grouped is divided in target DRG group, complete the grouping of the case history to be grouped Journey.

In conclusion the disclosure can extract all target signatures wait be grouped in case history；According to the feature pre-established Probability net and above-mentioned all target signatures obtain the total energy value that the first DRG group corresponds to the case history to be grouped, this feature Probability net is using DRG group and feature group as node, using the incidence relation between DRG group and feature group as side, with DRG group and spy Correlation probabilities value between sign group is the network topology structure that the weight on the side is established, and the first DRG group is that the library DRG includes Any DRG group in multiple DRG groups, the total energy value are multiple spies belonging to the first DRG group and above-mentioned all target signatures The summation of multiple correlation probabilities values between sign group, each this feature group include to meet multiple features of same grouping condition； Determine that the DRG group for having the maximum total energy value is to be somebody's turn to do the corresponding target DRG group of case history to be grouped.It can be by according to feature The network structure established with the correlation of DRG group is identified and is grouped to case history, and the step of being manually grouped is avoided, and improves case history The efficiency and accuracy rate of grouping.

Fig. 2 is the flow chart for implementing the group technology of another case history exemplified according to Fig. 1, as shown in Fig. 2, Before step 101, this method can also include:

Step 104, be directed to fisrt feature class, obtain the 2nd DRG group it is corresponding it is multiple be grouped in case history belong to this All sample characteristics of one feature class.

Wherein, the 2nd DRG group is any DRG group in the library DRG, which is in above-mentioned multiple feature classes Any feature class.It should be noted that the step 104 is to the following steps 106 only under a feature class in a DRG group All sample characteristics number statistics and probability calculation process for be illustrated.It is needed in actual DRG system scenarios Number statistics and probability calculation are carried out to the big measure feature under multiple feature classes in multiple DRG groups, wherein in each DRG group Each feature class under feature number statistics and probability calculation process and the step 104 to the number described in step 106 It counts identical with probability calculation process.

Illustratively, above-mentioned all sample characteristics be it is above-mentioned it is multiple be grouped in case history documented content under each column, Wherein, the corresponding feature class in each column.Above-mentioned multiple case histories (including above-mentioned case history to be grouped) that have been grouped are set herein to adopt With identical tabular form, that is, it is identical to be each grouped the feature class for including in case history.For example, the fisrt feature class is (i.e. Some corresponding to any DRG group has been grouped a certain column of case history) it is " symptom ", then it has each been grouped in case history comprising feature Class " symptom ".Wherein, all sample characteristics for belonging to the fisrt feature class may include: " trachyphonia ", " hyperplasia ", " discomfort ", " pain Bitterly " and " redness " etc..It should be noted that feature involved in the embodiment of the present disclosure is divided into two spies of discrete type and continuous type Value indicative type, discrete type characteristic value is character text, for example, above-mentioned " trachyphonia ", " hyperplasia ", " discomfort ", " pain " and " red It is swollen " etc. be discrete type feature, and the characteristic value for including under feature class " hospitalization cost " is the numerical value of continuous type, a feature The feature for including under class is provided with same type of characteristic value.It should be noted that the area of the feature of the discrete type and continuous type Dividing is not only foundation by text or number of feature text, for example, the feature for including under feature class " drug number " " 20140910348940 " do not have continuity, therefore this feature " 20140910348940 " is discrete to have though for number The feature of type characteristic value.

Step 105, according to the corresponding grouping condition of the fisrt feature class, above-mentioned all sample characteristics are divided into multiple spies Sign group, so that the quantity of sample characteristics is in default distribution in above-mentioned multiple feature groups.

Illustratively, the corresponding grouping condition of the fisrt feature class by sample characteristics under the fisrt feature class feature Value Types It determines, when the feature Value Types of sample characteristics under the fisrt feature class are discrete type, which is by the fisrt feature Characteristic value (practical is character text) identical sample characteristics of sample characteristics are divided into a feature group in class, then unite The quantity for counting feature in feature group multiple has been grouped the number occurred in case history above-mentioned as the sample characteristics.For example, this One feature class is " symptom ", all sample characteristics under the fisrt feature class be divided into A (trachyphonia), B (hyperplasia), C (discomfort), This 5 feature groups of D (pain) and E (redness).Wherein, include 50 features " trachyphonia " in A feature group, also mean that, " sound Neigh " this feature in above-mentioned multiple be grouped in case history occurs 50 times.

Based on this, it is to be understood that practical only to need when the feature Value Types of sample characteristics under feature class are discrete type The number of the same next feature of feature class (i.e. a character text) appearance is counted, uses " feature group ", " feature " herein The difference of " characteristic value " is stated to be continuous type with the feature Value Types for working as sample characteristics under feature class described below When the step of frequency of occurrence of feature is counted it is corresponding.Has discrete type feature for example, being directed under feature class " symptom " It is worth the feature of " trachyphonia ", it is all " trachyphonia " that corresponding content is identical with " characteristic value " for " feature group ", " feature " here, but Unlike, " feature group " lays particular emphasis on the quantity of performance characteristic " trachyphonia " appearance, and " characteristic value " lays particular emphasis on performance characteristic " trachyphonia " Feature Value Types (i.e. discrete type).

Illustratively, unlike the feature for having discrete type characteristic value, the feature for having continuous type characteristic value can not be used For the consistent mode of character text as grouping condition, therefore, it is necessary to will be provided with the feature of continuous type characteristic value according to locating for it Value interval is divided into multiple feature groups.When the feature Value Types of sample characteristics under feature class are continuous type, the grouping condition The maximum eigenvalue and minimal eigenvalue of the sample characteristics under the fisrt feature class are in for acquisition, then in maximum eigenvalue and most Small characteristic value etc. separates the value interval of preset quantity, and using each value interval as a feature group, then will take in this The multiple feature value divisions for being worth section are a feature group.For example, being directed to feature class " hospitalization cost ", it is in this feature class The maximum eigenvalue of sample characteristics under " hospitalization cost " is " 20000 " and minimal eigenvalue is " 5000 ", at this time can be in spy Separated between value indicative " 20000 " and feature " 5000 " etc. [5000,10000], [10001,15000] and [15001,20000] this 3 A value interval, and successively respectively correspond feature group X, feature group Y and feature group Z.Characteristic value is " 5040 " and " 6000 " as a result, Feature be divided to this feature group X, characteristic value is that the feature of " 12000 " and " 13987 " is divided to this feature group Y, feature Value is that the feature of " 18234 " and " 17540 " is divided to this feature group Z.It can be seen that being directed under feature class " hospitalization cost " The feature for having continuous type characteristic value " 6000 ", practical " feature group " here be characterized value area locating for " 6000 " Between, " characteristic value " is applied not only to the feature Value Types (i.e. continuous type) of performance characteristic " 6000 ", is also used as judging characteristic " 6000 " Locating " feature group " foundation.

Step 106, according to the quantity of sample characteristics in each feature group in above-mentioned multiple feature groups and this default point The corresponding probability density function of cloth state obtains between each feature group in the 2nd DRG group and above-mentioned multiple feature groups Correlation probabilities value.

Illustratively, which is normal distribution, which includes: to make the number of each feature group For the input variable of normpdf, to obtain each spy in the 2nd DRG group and above-mentioned multiple feature groups Correlation probabilities value between sign group；Wherein, which includes:

Wherein, x is input variable, and σ is the standard deviation of the quantity of target signature in above-mentioned multiple feature groups, and μ is above-mentioned more The average value of the quantity of sample characteristics in a feature group.

Illustratively, after by above-mentioned steps 104 to 106, available to one DRG group and the DRG group are corresponding Belong to the correlation probabilities value between multiple feature groups of the same feature class.Wherein, a feature class is a computing unit. By taking the corresponding all feature groups of a DRG group are n feature group as an example, this n feature group is divided into 3 feature classes, feature class A Corresponding a feature group, feature class B correspond to b feature group, and feature class C corresponds to c feature group, wherein n=a+b+c.It can manage Solution is needed when calculating the correlation probabilities value between DRG group and feature group with the corresponding a feature group of feature class A (feature class B corresponds to b feature group or feature class C corresponds to c feature group) is a computing unit, carries out above-mentioned steps 106 In probability value calculate step, to obtain each feature in the DRG group and a feature group (b feature group or c feature group) Correlation probabilities value between group.And to obtain the correlation probabilities in the DRG group and n feature group between each feature group Value, needs above-mentioned steps 104-106 in triplicate.A DRG group is being completed to relevant feature group (for example, above-mentioned n A feature group) between correlation probabilities value calculate after, can be with same step, (i.e. to all DRG groups in the library DRG Above-mentioned multiple DRG groups) to the correlation probabilities between relevant feature group (i.e. above-mentioned multiple DRG groups corresponding all feature groups) Value is calculated, as the foundation for setting up this feature probability net in the following steps 107.

In addition, it is necessary to explanation, after the completion of above-mentioned calculating step, can give up the concept of feature class, it directly will be every A feature group merely obtains incidence relation between a feature group and a DRG group and corresponding as a separate unit Weight, and then carry out the following steps 107 described in characteristic probability network establishment step.Similarly, in above-mentioned steps 102 When obtaining the total energy value of each DRG group according to all target signatures wait be grouped in case history, it is necessary first to according to each target Feature class belonging to feature positions the feature group belonging to it, then equally gives up the concept of feature class later, determines each Correlation probabilities value between feature group (this feature group includes either objective feature) and each DRG group, and sum, with To above-mentioned total energy value.

Step 107, the phase between above-mentioned multiple DRG groups and the corresponding all feature groups of above-mentioned multiple DRG groups is being got After closing property probability value, using above-mentioned multiple DRG groups and the corresponding all feature groups of above-mentioned multiple DRG groups as node, with above-mentioned more Incidence relation between a DRG group and the corresponding all feature groups of above-mentioned multiple DRG groups is side, with above-mentioned multiple DRG groups and on The weight that the correlation probabilities value between the corresponding all feature groups of multiple DRG groups is the side is stated, this feature probability net is established.

Illustratively, this feature probability net can be expressed as G (V, E, W).In this feature probability net G (V, E, W), V It can indicate are as follows: V=V_DRG∪V_f, wherein V_DRGIndicate the set of above-mentioned multiple DRG group nodes, V_fIndicate all feature group nodes Set；E can be indicated are as follows: E={ e_ij|i∈V_f,j∈V_DRG, wherein i V_fIn any feature group node, j V_DRGIn Any DRG group node, e_ijFor the side (i.e. the incidence relation of the two) of connection features group node i and DRG group node j；W can be with table It is shown as W={ ρ_j,p(i)|i∈V_f,j∈V_DRG, wherein ρ_j,p(i) indicate that feature group node i and DRG group at feature class p are saved Correlation probabilities value between point j.

Fig. 3 is the flow chart for implementing a kind of total energy value acquisition methods exemplified according to Fig.2, as shown in figure 3, on Stating step 102 may include:

Step 1021, according to the determination of this feature probability net, there are all first spies of incidence relation with the first DRG group Sign group.

Step 1022, determining multiple second to match with above-mentioned all target signatures in above-mentioned all fisrt feature groups Feature group.

Illustratively, the target signature that feature Value Types in above-mentioned all target signatures are discrete type, the step are directed to 1022 may include: step 10221, determine second feature class belonging to the target signature；Step 10222, above-mentioned all The multiple third feature groups for corresponding to the second feature class are determined in one feature group；Step 10223, in above-mentioned multiple third feature It is determining in group to have feature group locating for the feature of same characteristic features value with the target signature, as what is matched with the fisrt feature Second feature group；Step 10224, the determining second feature group to match with each target signature, as above-mentioned multiple second Feature group.For example, being directed to the target signature that characteristic value is " redness ", it is necessary first to determine belonging to target signature " redness " the Two feature classes " symptom "；Hereafter, in above-mentioned all fisrt feature groups (all feature groups i.e. in this feature probability net) really Surely correspond to multiple third feature groups of the second feature class, the corresponding feature class of above-mentioned multiple third feature groups is all " disease Shape "；It and then in feature class is to determine have the institute of feature of the characteristic value for " redness " in multiple third feature groups of " symptom " There is feature group, as the second feature group；Above-mentioned steps 10221 to 10223 are finally repeated, determine other target signature phases Matched multiple feature groups, as above-mentioned multiple second feature groups.

Alternatively, being directed to the target signature that feature Value Types in above-mentioned all target signatures are continuous type, the step 1022 It may include: step 10225, determine third feature class belonging to the target signature；Step 10226, special above-mentioned all first The multiple fourth feature groups for corresponding to the third feature class are determined in sign group, the third is special according to above-mentioned multiple fourth feature groups Multiple value intervals that maximum eigenvalue and minimal eigenvalue of the corresponding all features of sign class etc. separate；Step 10227, upper It states and determines the first value interval locating for the target signature in multiple value intervals；Step 10228, in above-mentioned multiple fourth feature Feature group corresponding with first value interval is determined in group, as the second feature group to match with the target signature；Step Rapid 10229, the determining second feature group to match with each target signature, as above-mentioned multiple second feature groups.For example, needle It is the target signature of " 6000 " for characteristic value, it is necessary first to determine third feature class belonging to target signature " 6000 " " in hospital Expense "；Hereafter, it determines to correspond in above-mentioned all fisrt feature groups (all feature groups i.e. in this feature probability net) and be somebody's turn to do Multiple fourth feature groups of third feature class, above-mentioned multiple fourth feature groups correspond to multiple the taking of third feature class " hospitalization cost " It is worth section；And then value interval locating for characteristic value " 6000 " is determined in above-mentioned multiple value intervals, and by the value area Between corresponding feature group as the second feature group；Finally repeat above-mentioned steps 10225 to 10228, determining and other mesh Multiple feature groups that mark feature matches, as above-mentioned multiple second feature groups.

Step 1023, it is determined between the first DRG group and above-mentioned multiple second feature groups according to this feature probability net The summation of multiple correlation probabilities values corresponds to the total energy value of the case history to be grouped as the first DRG group.

Illustratively, DRG group and the initialization energy value formula (2) of feature group can indicate are as follows:

Wherein, V_fFor the set of all feature group nodes in DRG system, V '_fFor above-mentioned all mesh wait be grouped in case history Mark the set of the corresponding feature group of feature, the meaning of the formula (2) are as follows: the set of all feature group nodes and above-mentioned disease to be grouped The intersection of sets of the corresponding feature group of all target signatures in going through concentrates the feature group node for including, if initialization energy value is 1；For exist only in it is above-mentioned be grouped case history and should the corresponding feature group node of the feature group of one of case history to be grouped, if Initialize energy value be 0, actual effect be ignore when calculating the total energy value exist only in it is above-mentioned be grouped case history and be somebody's turn to do The feature group of one of case history to be grouped.

Illustratively, it is based on the primary power value formula (2), the formula (3) for calculating the total energy value can indicate are as follows:

Wherein, Q_jIndicate the total energy value of j-th of DRG group, Q_iIndicate the initialization energy value of ith feature group, W_ijI-th The weight (correlation probabilities value) on side between a feature group and j-th of DRG group.

Fig. 4 is the flow chart for implementing a kind of feature partition method exemplified according to Fig.2, as shown in figure 4, when should When the feature Value Types of the corresponding all sample characteristics of fisrt feature class are discrete type, above-mentioned steps 105 may include:

Step 1051, the sample characteristics for having same characteristic features value in above-mentioned all sample characteristics are divided into a feature Group, to obtain above-mentioned multiple feature groups.

Step 1052, above-mentioned multiple feature groups are arranged so that in above-mentioned multiple feature groups sample characteristics quantity Distribution is preset in this.

Illustratively, which is normal distribution, due under the feature class comprising discrete type characteristic value, at random The quantity of the sample characteristics of arrangement will not be in normal distribution in most cases.Therefore, it needs in the step 1052 to upper It states multiple feature groups to be arranged, so that the quantity of its sample characteristics is in normal distribution.Specific arrangement mode can be with are as follows: more than Centered on the most feature group of quantity for stating the sample characteristics in multiple feature groups, by other feature groups with descending big sequence It is sequentially arranged in the left and right of the most feature group of quantity of above-mentioned sample characteristics, makes the quantity of its sample characteristics in normal distribution.

It step 1053, is above-mentioned multiple characteristic components with number according to the sequence of arrangement.

Illustratively, input variable of the number as above-mentioned probability density function is practical to indicate a feature group whole The location of in the normal distribution of body.

Fig. 5 is the flow chart for implementing another feature partition method exemplified according to Fig.2, as shown in figure 5, working as When the feature Value Types of the corresponding all sample characteristics of the fisrt feature class are continuous type, above-mentioned steps 105 may include:

Step 1054, the sample characteristics for having maximum eigenvalue in above-mentioned all sample characteristics are obtained and have minimal characteristic The sample characteristics of value.

Step 1055, multiple value intervals are separated between the maximum eigenvalue and the minimal eigenvalue etc..

Illustratively, the step 1055 may include: by the following group away from calculation formula (4) calculate value interval group away from:

Wherein, g_disExpression group is away from V_maxFor the maximum eigenvalue, V_minFor the minimal eigenvalue, g_numFor preset subregion Quantity, preferably 30.Hereafter available to arrive section [V_min+i*g_dis,V_min+(i+1)*g_dis], wherein i ∈ [0, g_num], i.e. i G is arrived for 0_numBetween any value interval serial number.

Step 1056, the sample characteristics that the same value interval is in above-mentioned all sample characteristics are divided to same Feature group, to obtain above-mentioned multiple feature groups.

It step 1057, is above-mentioned multiple spies according to the size of the endpoint value of the corresponding value interval of above-mentioned multiple feature groups Sign group distribution number.

Illustratively, different from the above-mentioned sample characteristics for having discrete type characteristic value, corresponding according to above-mentioned multiple feature groups Value interval endpoint value size above-mentioned multiple feature groups are arranged after, sample characteristics in above-mentioned multiple feature groups The distribution of quantity be inherently similar to normal distribution, without re-starting arrangement.It therefore, is herein directly to arrange Multiple characteristic components with number, the number is identical as the number effect in above-mentioned steps 1053.

In conclusion the disclosure can extract all target signatures wait be grouped in case history；According to the feature pre-established Probability net and above-mentioned all target signatures obtain the total energy value that the first DRG group corresponds to the case history to be grouped, this feature Probability net is using DRG group and feature group as node, using the incidence relation between DRG group and feature group as side, with DRG group and spy Correlation probabilities value between sign group is the network topology structure that the weight on the side is established, and the first DRG group is that the library DRG includes Any DRG group in multiple DRG groups, the total energy value are multiple spies belonging to the first DRG group and above-mentioned all target signatures The summation of multiple correlation probabilities values between sign group, each this feature group include to meet multiple features of same grouping condition； Determine that the DRG group for having the maximum total energy value is to be somebody's turn to do the corresponding target DRG group of case history to be grouped.It can be by according to case history In the network structure established of correlation of the feature of the discrete type for including and continuous type and DRG group case history is identified and is divided Group expands the scope of application of adaptive case history grouping while avoiding the step being manually grouped, and then improves case history grouping Efficiency and accuracy rate.

Fig. 6 is a kind of block diagram of the apparatus for grouping of case history shown according to an exemplary embodiment, as shown in fig. 6, using In above-mentioned DRG system, which includes:

Characteristic extracting module 610, for extracting all target signatures wait be grouped in case history；

Energy value obtains module 620, the characteristic probability network pre-established for basis and above-mentioned all target signatures, The total energy value that the first DRG group corresponds to the case history to be grouped is obtained, this feature probability net is using DRG group and feature group as section Point, using the incidence relation between DRG group and feature group as side, with the correlation probabilities value between DRG group and feature group for the side Weight establish network topology structure, the first DRG group be the library DRG include multiple DRG groups in any DRG group, this always Energy value is multiple correlation probabilities values between multiple feature groups belonging to the first DRG group and above-mentioned all target signatures Summation, each this feature group include to meet multiple features of same grouping condition；

Case history grouping module 630, for determining that the DRG group for having the maximum total energy value is to be somebody's turn to do case history to be grouped to correspond to Target DRG group.

Fig. 7 is the block diagram for implementing the apparatus for grouping of another case history exemplified according to Fig.6, as shown in fig. 7, should Each of library DRG DRG group correspond to it is multiple be grouped case history, each this has been grouped in case history comprising multiple features, above-mentioned more A feature corresponds to multiple feature classes, the device 600 further include:

Sample acquisition module 640 obtains that the 2nd DRG group is corresponding multiple to be grouped disease for being directed to fisrt feature class Belong to all sample characteristics of the fisrt feature class in going through, the 2nd DRG group is any DRG group in the library DRG, first spy Levying class is any feature class in above-mentioned multiple feature classes；

Feature grouping module 650 is used for according to the corresponding grouping condition of the fisrt feature class, by above-mentioned all sample characteristics Multiple feature groups are divided into, so that the quantity of sample characteristics is in default distribution in above-mentioned multiple feature groups；

Correlation determining module 660, for the number according to sample characteristics in each feature group in above-mentioned multiple feature groups Amount and the default corresponding probability density function of distribution, obtain in the 2nd DRG group and above-mentioned multiple feature groups Correlation probabilities value between each feature group；

Network establishes module 670, for getting above-mentioned multiple DRG groups and the corresponding all spies of above-mentioned multiple DRG groups It is section with above-mentioned multiple DRG groups and the corresponding all feature groups of above-mentioned multiple DRG groups after correlation probabilities value between sign group Point, using the incidence relation between above-mentioned multiple DRG groups and the corresponding all feature groups of above-mentioned multiple DRG groups as side, with above-mentioned more Correlation probabilities value between a DRG group and the corresponding all feature groups of above-mentioned multiple DRG groups is the weight on the side, establishes the spy Levy probability net.

Fig. 8 is that a kind of energy value for implementing to exemplify according to Fig.7, obtains the block diagram of module, as shown in figure 8, the energy Value obtains module 620, comprising:

Fisrt feature group determines submodule 621, for there is pass with the first DRG group according to this feature probability net is determining All fisrt feature groups of connection relationship；

Second feature group determines submodule 622, for determining and above-mentioned all targets in above-mentioned all fisrt feature groups Multiple second feature groups that feature matches；

Energy value computational submodule 623, for determining the first DRG group and above-mentioned multiple the according to this feature probability net The summation of multiple correlation probabilities values between two feature groups corresponds to the total energy of the case history to be grouped as the first DRG group Magnitude.

Fig. 9 is the block diagram for implementing a kind of feature grouping module exemplified according to Fig.7, as shown in figure 9, this feature Feature Value Types are discrete type or continuous type, and the corresponding all features of each this feature class have same feature Value Types, the spy Levy grouping module 650, comprising:

Fisrt feature group acquisition submodule 651, the sample for will have same characteristic features value in above-mentioned all sample characteristics Feature is divided into a feature group, to obtain above-mentioned multiple feature groups；

Feature group arranges submodule 652, for arranging above-mentioned multiple feature groups, so that in above-mentioned multiple feature groups The quantity of sample characteristics is in the default distribution；

Fisrt feature group # submodule 653, for being above-mentioned multiple characteristic components with number according to the sequence of arrangement.

Figure 10 is the block diagram for implementing another feature grouping module exemplified according to Fig.7, as shown in Figure 10, the spy Levy grouping module 650, comprising:

Characteristic value acquisition submodule 654, for obtaining the sample spy for having maximum eigenvalue in above-mentioned all sample characteristics It seeks peace and has the sample characteristics of minimal eigenvalue；

Interval division submodule 655, for separating multiple values between the maximum eigenvalue and the minimal eigenvalue etc. Section；

Second feature group acquisition submodule 656, for the same value interval will to be in above-mentioned all sample characteristics Sample characteristics are divided to the same feature group, to obtain above-mentioned multiple feature groups；

Second feature group # submodule 657, for the number of endpoint according to the corresponding value interval of above-mentioned multiple feature groups The size of value is above-mentioned multiple characteristic components with number.

Optionally, which is normal distribution, which is used for:

By input variable of the number as normpdf of each feature group, with obtain this second The correlation probabilities value between each feature group in DRG group and above-mentioned multiple feature groups；Wherein, the normal distribution probability density Function includes:

Optionally, which determines submodule 622, is used for:

Determine second feature class belonging to the target signature；

The multiple third feature groups for corresponding to the second feature class are determined in above-mentioned all fisrt feature groups；

It is determining in above-mentioned multiple third feature groups to have feature locating for the feature of same characteristic features value with the target signature Group, as the second feature group to match with the fisrt feature；

The determining second feature group to match with each target signature, as above-mentioned multiple second feature groups.

Optionally, it is directed to the target signature that feature Value Types in above-mentioned all target signatures are continuous type, second spy Sign group determines submodule 622, is used for:

Determine third feature class belonging to the target signature；

The multiple fourth feature groups for corresponding to the third feature class are determined in above-mentioned all fisrt feature groups, it is above-mentioned multiple The maximum eigenvalue and minimal eigenvalue etc. of the corresponding all features of the third feature class separate more according to fourth feature group A value interval；

The first value interval locating for the target signature is determined in above-mentioned multiple value intervals；

Feature group corresponding with first value interval is determined in above-mentioned multiple fourth feature groups, as special with the target Levy the second feature group to match；

About the device in above-described embodiment, wherein modules execute the concrete mode of operation in related this method Embodiment in be described in detail, no detailed explanation will be given here.

Figure 11 is the block diagram of a kind of electronic equipment 1100 shown according to an exemplary embodiment.As shown in figure 11, the electricity Sub- equipment 1100 may include: processor 1101, memory 1102, multimedia component 1103, input/output (I/O) interface 1104 and communication component 1105.

Wherein, processor 1101 is used to control the integrated operation of the electronic equipment 1100, to complete point of above-mentioned case history All or part of the steps in group method.Memory 1102 is for storing various types of data to support in the electronic equipment 1100 operation, these data for example may include any application or method for operating on the electronic equipment 1100 Instruction and the relevant data of application program, such as contact data, the message of transmitting-receiving, picture, audio, video etc..It should Memory 1102 can be realized by any kind of volatibility or non-volatile memory device or their combination, such as static state Random access memory (Static Random Access Memory, abbreviation SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, abbreviation EEPROM), erasable programmable Read-only memory (Erasable Programmable Read-Only Memory, abbreviation EPROM), programmable read only memory (Programmable Read-Only Memory, abbreviation PROM), and read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, disk or CD.Multimedia component 1103 may include screen and audio component.Wherein Screen for example can be touch screen, and audio component is used for output and/or input audio signal.For example, audio component may include One microphone, microphone is for receiving external audio signal.The received audio signal can be further stored in storage Device 1102 is sent by communication component 1105.Audio component further includes at least one loudspeaker, is used for output audio signal.I/ O Interface 1104 provides interface between processor 1101 and other interface modules, other above-mentioned interface modules can be keyboard, mouse Mark, button etc..These buttons can be virtual push button or entity button.Communication component 1105 for the electronic equipment 1100 with Wired or wireless communication is carried out between other equipment.Wireless communication, such as Wi-Fi, bluetooth, near-field communication (Near Field Communication, abbreviation NFC), 2G, 3G or 4G or they one or more of combination, therefore corresponding communication Component 1105 may include: Wi-Fi module, bluetooth module, NFC module.

In one exemplary embodiment, electronic equipment 1100 can be by one or more application specific integrated circuit (Application Specific Integrated Circuit, abbreviation ASIC), digital signal processor (Digital Signal Processor, abbreviation DSP), digital signal processing appts (Digital Signal Processing Device, Abbreviation DSPD), programmable logic device (Programmable Logic Device, abbreviation PLD), field programmable gate array (Field Programmable Gate Array, abbreviation FPGA), controller, microcontroller, microprocessor or other electronics member Part is realized, for executing the group technology of above-mentioned case history.

In a further exemplary embodiment, a kind of computer readable storage medium including program instruction, example are additionally provided It such as include the memory 1102 of program instruction, above procedure instruction can be executed by the processor 1101 of electronic equipment 1100 to complete The group technology of above-mentioned case history.

The preferred embodiment of the disclosure is described in detail in conjunction with attached drawing above, still, the disclosure is not limited to above-mentioned reality The detail in mode is applied, in the range of the technology design of the disclosure, those skilled in the art are considering specification and practice After the disclosure, it is readily apparent that other embodiments of the disclosure, belongs to the protection scope of the disclosure.

It is further to note that specific technical features described in the above specific embodiments, in not lance In the case where shield, it can be combined in any appropriate way.Simultaneously between a variety of different embodiments of the disclosure Any combination can also be carried out, as long as it, without prejudice to the thought of the disclosure, equally should be considered as disclosure disclosure of that. The disclosure is not limited to the precision architecture being described above out, and the scope of the present disclosure is only limited by the attached claims System.

Claims

1. a kind of group technology of case history, which is characterized in that the described method includes:

Extract all target signatures wait be grouped in case history；

According to the characteristic probability network and all target signatures pre-established, obtain the first DRG group correspond to described in Be grouped the total energy value of case history, the characteristic probability network be using DRG group and feature group as node, with DRG group and feature group it Between incidence relation be side, with the correlation probabilities value between DRG group and feature group be the side weight establish network open up Flutter structure, the first DRG group is any DRG group in multiple DRG groups that the library DRG includes, and the total energy value is described the The summation of multiple correlation probabilities values between multiple feature groups belonging to one DRG group and all target signatures, Mei Gesuo Stating feature group includes to meet multiple features of same grouping condition；

2. the method according to claim 1, wherein each of described library DRG DRG group correspond to it is multiple It is grouped case history, each described be grouped includes multiple features in case history, and the multiple feature corresponds to multiple feature classes, mentions described Before taking all target signatures wait be grouped in case history, the method also includes:

It is directed to fisrt feature class, the 2nd DRG group corresponding multiple be grouped is obtained and belongs to the fisrt feature class in case history All sample characteristics, the 2nd DRG group are any DRG group in the library DRG, and the fisrt feature class is the multiple spy Levy any feature class in class；

It is corresponding according to the quantity of sample characteristics in each feature group in the multiple feature group and the default distribution Probability density function, obtain the 2nd DRG group and the correlation between each feature group in the multiple feature group be general Rate value；

Get the correlation probabilities value between the multiple DRG group and the corresponding all feature groups of the multiple DRG group it Afterwards, using the multiple DRG group and the corresponding all feature groups of the multiple DRG group as node, with the multiple DRG group and described Incidence relation between the corresponding all feature groups of multiple DRG groups is side, with the multiple DRG group and the multiple DRG group pair The correlation probabilities value between all feature groups answered is the weight on the side, establishes the characteristic probability network.

3. the method according to claim 1, wherein characteristic probability network that the basis pre-establishes and institute All target signatures are stated, the total energy value that the first DRG group corresponds to the case history to be grouped is obtained, comprising:

Multiple correlations between the first DRG group and the multiple second feature group are determined according to the characteristic probability network The summation of probability value corresponds to the total energy value of the case history to be grouped as the first DRG group.

4. according to the method described in claim 2, it is characterized in that, the feature Value Types of the feature are for discrete type or continuously Type, the corresponding all features of each feature class have same feature Value Types, when the fisrt feature class is corresponding all It is described according to the corresponding grouping condition of the fisrt feature class when feature Value Types of sample characteristics are discrete type, by the institute There are sample characteristics to be divided into multiple feature groups, so that the quantity of sample characteristics is in default distribution in the multiple feature group, Include:

The sample characteristics for having same characteristic features value in all sample characteristics are divided into a feature group, it is described more to obtain A feature group；

The multiple feature group is arranged, so that the quantity of sample characteristics is in the default distribution in the multiple feature group State；

5. according to the method described in claim 4, it is characterized in that, when the corresponding all sample characteristics of the fisrt feature class It is described according to the corresponding grouping condition of the fisrt feature class when feature Value Types are continuous type, by all sample characteristics Multiple feature groups are divided into, so that the quantity of sample characteristics is in default distribution in the multiple feature group, comprising:

It obtains and has the sample characteristics of maximum eigenvalue in all sample characteristics and have the sample characteristics of minimal eigenvalue；

The sample characteristics that the same value interval is in all sample characteristics are divided to the same feature group, to obtain Take the multiple feature group；

It is the multiple characteristic component with number according to the size of the endpoint value of the corresponding value interval of the multiple feature group.

6. according to the method described in claim 5, it is characterized in that, the default distribution is normal distribution, the basis The quantity of target signature and the corresponding probability of the default distribution are close in each feature group in the multiple feature group Function is spent, the correlation probabilities value between the 2nd DRG group each feature group corresponding with the 2nd DRG group, packet are obtained It includes:

Input variable by the number of each feature group as normpdf, to obtain described second The correlation probabilities value between each feature group in DRG group and the multiple feature group；Wherein, the normal distribution probability is close Spending function includes:

Wherein, x is input variable, and σ is the standard deviation of the quantity of target signature in the multiple feature group, and μ is the multiple spy The average value of the quantity of sample characteristics in sign group.

7. a kind of apparatus for grouping of case history, which is characterized in that described device includes:

Energy value obtains module, for according to the characteristic probability network that pre-establishes and all target signatures, obtaining the One DRG group corresponds to the total energy value of the case history to be grouped, and the characteristic probability network is using DRG group and feature group as section Point is described with the correlation probabilities value between DRG group and feature group using the incidence relation between DRG group and feature group as side The network topology structure that the weight on side is established, the first DRG group are any DRG group in multiple DRG groups that the library DRG includes, Multiple correlations of the total energy value between the first DRG group and multiple feature groups belonging to all target signatures The summation of probability value, each feature group include to meet multiple features of same grouping condition；

Case history grouping module, for determining that the DRG group for having the maximum total energy value is that the case history to be grouped is corresponding Target DRG group.

8. device according to claim 7, which is characterized in that each of described library DRG DRG group correspond to it is multiple It is grouped case history, each described be grouped includes multiple features in case history, and the multiple feature corresponds to multiple feature classes, described device Further include:

Sample acquisition module obtains corresponding multiple be grouped in case history of the 2nd DRG group and belongs to for being directed to fisrt feature class All sample characteristics of the fisrt feature class, the 2nd DRG group are any DRG group in the library DRG, and described first is special Levying class is any feature class in the multiple feature class；

Feature grouping module, for according to the corresponding grouping condition of the fisrt feature class, all sample characteristics to be divided For multiple feature groups, so that the quantity of sample characteristics is in default distribution in the multiple feature group；

Correlation determining module, for the quantity according to sample characteristics in each feature group in the multiple feature group, and The corresponding probability density function of the default distribution, obtains each of the 2nd DRG group and the multiple feature group Correlation probabilities value between feature group；

Network establishes module, for getting between the multiple DRG group and the corresponding all feature groups of the multiple DRG group Correlation probabilities value after, using the multiple DRG group and the corresponding all feature groups of the multiple DRG group as node, with institute The incidence relation stated between multiple DRG groups and the corresponding all feature groups of the multiple DRG group is side, with the multiple DRG group Correlation probabilities value between the corresponding all feature groups of the multiple DRG group is the weight on the side, establishes the feature Probability net.

9. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program quilt The step of any one of claim 1-6 the method is realized when processor executes.

10. a kind of electronic equipment characterized by comprising

Memory is stored thereon with computer program；

Processor, for executing the computer program in the memory, to realize described in any one of claim 1-6 The step of method.