CN104933320A - Prescription drug property quantifying method and system based on weighted PageRank algorithm - Google Patents

Abstract

The invention provides a prescription drug property quantifying method and system based on a weighted PageRank algorithm. The quantifying method and system are beneficial for improving the accuracy of a drug property PR value. The method includes the steps that drug properties of all drugs in a prescription are decomposed to determine the drug properties of the prescription and weights of the drug properties; the drug properties of the prescription serve as drug property nodes in a network link structure, and according to interaction among the different drug properties, a network link structure graph among the drug property nodes is built; according to the determined weights of the drug properties, the weighted PageRank algorithm is used for iterating the drug property nodes to determine the drug property PR value. The prescription drug property quantifying method and system are suitable for the technical field of drug property quantifying.

Description

A kind of recipe drug attribute quantivative approach based on weighting PageRank algorithm and system

Technical field

The present invention relates to medicine attribute quantification technical field, refer to a kind of recipe drug attribute quantivative approach based on weighting PageRank algorithm and system especially.

Background technology

Prescription is formed by multi-medicament compatibility, and each herbal medicine has corresponding medicine attribute, and the four property five tastes of such as medicine are returned through attributes such as, effects.The prescription action effect of the traditional Chinese medical science is interacted by medicine each in prescription or medicine attribute and jointly produces, maximum in order to embody the effect of which kind of medicine to prescription effect in prescription, need to quantize the action effect of each medicine, and then utilize medicine quantized result to quantize the medicine attribute of medicine, different pharmaceutical attribute shared significance level in prescription can be embodied.In the recipe drug of reality, the actual dose of each often medicine provided.List can not judge the drug effect size of medicine from actual dose size or be called contribution degree size, but based on relative dose (drug dose of different nature can compare under same standard), can determine that in prescription, each medicine is relative to the weight size of prescription.Tradition recipe drug and attribute quantification more completely construct the mathematical model of relative dose intensity by the actual dose of the Chinese medicine in prescription and its minimax conventional amount used value ratio, significant in the research of recipe drug Evaluating Quantitative System.But, traditional recipe drug quantization method is all provide static mathematical model when comparatively meeting drug medication rule, certain the medicine quantized result gone out by model inference is often only based on actual dose and the minimax amount of this medicine, and have ignored prescription effect is medicine or the interactional effect of medicine attribute.

Under tcm prescription gradually modern prerequisite, modern information technologies, as data mining technology is applied in tcm field each side gradually.Such as, based on the traditional Chinese medicine of cluster and Mining fuzzy association rules to dose-effect association analysis, be used for rough set theory and method excavating the analysis mining of correlationship between syndrome index and each medicine of prescription or Chinese medicinal formulae Compatibility Law.But yet there are no research data mining technology be applied in recipe drug quantification problem in prior art, traditional recipe drug attribute quantivative approach have ignored interactional effect between medicine attribute, the Relative drug property value obtained out of true.

Summary of the invention

The technical problem to be solved in the present invention is to provide a kind of recipe drug attribute quantivative approach based on weighting PageRank algorithm and system, have ignored the interphase interaction of medicine attribute with the traditional recipe drug attribute quantivative approach solved existing for prior art, cause the problem that the Relative drug property value degree of accuracy that obtains is low.

For solving the problems of the technologies described above, the embodiment of the present invention provides a kind of recipe drug attribute quantivative approach based on weighting PageRank algorithm, comprising:

The medicine attribute of medicine each in prescription is decomposed, determines the medicine attribute of prescription and the weights of described medicine attribute;

Using the medicine attribute in prescription as the medicine attribute node in link structure, according to the interaction between different pharmaceutical attribute, build the link structure figure between described medicine attribute node;

According to the weights of the medicine attribute determined, utilize weighting PageRank algorithm to carry out iteration to medicine attribute node, determine the PR value of medicine attribute.

Alternatively, the described medicine attribute by medicine each in prescription decomposes, and determines that the medicine attribute of prescription and the weights of described medicine attribute comprise:

Decomposed by the medicine attribute of medicine each in prescription, described medicine attribute comprises: the property of medicine, flavour of a drug and return through three class medicine attributes;

Determine the weights of the medicine attribute of every herbal medicine;

The weights of same medicine attribute in prescription are added, determine the medicine attribute of prescription and the weights of described medicine attribute.

Alternatively, describedly determine that the weights of the medicine attribute of every herbal medicine comprise:

Utilize traditional recipe drug attribute quantivative approach that the actual dose of herbal medicine every in prescription is standardized as relative dose and normalization;

Relative dose value assignment after normalization is given the weights of medicine attribute as described medicine attribute of its correspondence;

Wherein, the mathematical model actual dose of herbal medicine every in prescription being standardized as relative dose is as follows:

If then X=kx (x-2b);

If $M<\sqrt{10}m,$ Then $X=\left\{\begin{array}{c}kx(x-2b),x>m\\ \frac{10x}{m},x\&le;m\end{array}\right.;$

If M>=10m, then

In formula, $k=\frac{10(10m-M)}{Mm(M-m)},b=\frac{10m^2-M^2}{2(10m-M)},$ X is the relative dose of medicine, and x is the actual dose of medicine, and M is medicine research on maximum utilized quantity, and m is medicine minimum amount.

Alternatively, the weights of the described medicine attribute according to determining, utilize weighting PageRank algorithm to carry out iteration to medicine attribute node, determine that the PR value of medicine attribute comprises:

Quantize medicine attribute node;

Iteration is carried out, until PR value tends towards stability by the PR value of PR value computing formula to each medicine attribute quantized of medicine attribute;

Wherein, the PR value computing formula of medicine attribute is as follows:

$PR\left(p\right)=(1-d)weight+d\underset{i=1}{\overset{n}{\&Sigma;}}\frac{PR\left(T_i\right)}{C\left(T_i\right)}$

In formula, PR (p) represents the PR value of medicine attribute node p, and n is the number of the other drug attribute node pointing to medicine attribute p, T _i(i=1,2 ..., n) represent the other drug attribute pointing to medicine attribute p, PR (T _i) represent medicine attribute T _ipR value, C (T _i) be medicine attribute T _ithe number of links outwards pointed out, weight is the weights of medicine attribute p, and d is boundary in (0,1) interval attenuation coefficient, is generally 0.85.

Alternatively, the weights of the described medicine attribute according to determining, utilize weighting PageRank algorithm to carry out iteration to medicine attribute node, comprise after determining the PR value of medicine attribute:

Size according to the PR value of the medicine attribute determined sorts, and determines the prescription important attribute after the acting in conjunction of described medicine attribute.

The embodiment of the present invention also provides a kind of recipe drug attribute quantitative system based on weighting PageRank algorithm, comprising:

Recipe drug attribute determining unit: for being decomposed by the medicine attribute of medicine each in prescription, determines the medicine attribute of prescription and the weights of described medicine attribute;

Link structure determining unit: for using the medicine attribute in prescription as the medicine attribute node in link structure, according to the interaction between different pharmaceutical attribute, build the link structure figure between described medicine attribute node;

PR value determining unit: for the weights according to the medicine attribute determined, utilizes weighting PageRank algorithm to carry out iteration to medicine attribute node, determines the PR value of medicine attribute.

Alternatively, described recipe drug attribute determining unit comprises:

Every herbal medicine medicine attribute determination module: for being decomposed by the medicine attribute of medicine each in prescription, described medicine attribute comprises: the property of medicine, flavour of a drug and return through three class medicine attributes;

Every herbal medicine medicine attribute weights determination module, for determining the weights of the medicine attribute of every herbal medicine;

Recipe drug attribute determination module: for being added by the weights of same medicine attribute in prescription, determines the medicine attribute of prescription and the weights of described medicine attribute.

Alternatively, described every herbal medicine medicine attribute weights determination module comprises:

Normalization submodule: the actual dose of herbal medicine every in prescription is standardized as relative dose and normalization for utilizing traditional recipe drug attribute quantivative approach;

Every herbal medicine medicine attribute weights determination submodule: for the relative dose value assignment after normalization being given the weights of medicine attribute as described medicine attribute of its correspondence;

Wherein, the mathematical model actual dose of herbal medicine every in prescription being standardized as relative dose is as follows:

If then X=kx (x-2b);

If $M<\sqrt{10}m,$ Then $X=\left\{\begin{array}{c}kx(x-2b),x>m\\ \frac{10x}{m},x\&le;m\end{array}\right.;$

If M>=10m, then

In formula, $k=\frac{10(10m-M)}{Mm(M-m)},b=\frac{10m^2-M^2}{2(10m-M)},$ X is the relative dose of medicine, and x is the actual dose of medicine, and M is medicine research on maximum utilized quantity, and m is medicine minimum amount.

Alternatively, described PR value determining unit comprises:

Quantization modules: for quantizing medicine attribute node;

PR value determination module: carry out iteration, until PR value tends towards stability for the PR value of PR value computing formula to each medicine attribute quantized by medicine attribute;

Wherein, the PR value computing formula of medicine attribute is as follows:

$PR\left(p\right)=(1-d)weight+d\underset{i=1}{\overset{n}{\&Sigma;}}\frac{PR\left(T_i\right)}{C\left(T_i\right)}$

In formula, PR (p) represents the PR value of medicine attribute node p, and n is the number of the other drug attribute node pointing to medicine attribute p, T _i(i=1,2 ..., n) represent the other drug attribute pointing to medicine attribute p, PR (T _i) represent medicine attribute T _ipR value, C (T _i) be medicine attribute T _ithe number of links outwards pointed out, weight is the weights of medicine attribute p, and d is boundary in (0,1) interval attenuation coefficient, is generally 0.85.

Alternatively, described system also comprises:

Prescription important attribute determining unit: the size for the PR value according to the medicine attribute determined sorts, determines the prescription important attribute after the acting in conjunction of described medicine attribute.

The beneficial effect of technique scheme of the present invention is as follows:

In such scheme, by the medicine attribute of medicine each in prescription is decomposed, determine the medicine attribute of prescription and the weights of described medicine attribute, and using the medicine attribute in prescription as the medicine attribute node in link structure, according to the interaction between different pharmaceutical attribute, build the link structure figure between described medicine attribute node; Again according to the weights of the medicine attribute determined, utilize weighting PageRank algorithm to carry out iteration to medicine attribute node, determine the PR value of medicine attribute.Like this, the present invention not only considers the action effect between the medicine attribute of every herbal medicine self, interaction between the medicine attribute considering different pharmaceutical in prescription according to link structure again, perfect traditional recipe drug attribute quantivative approach ignores interactional deficiency between medicine attribute; And the PR value of the medicine attribute utilizing the PageRank algorithm of weighting to determine, this PR value can sort to prescription important attribute more accurately, for tcm prescription medicine and the quantitative examination of medicine attribute open a new thinking, advance the paces of TCM Modernization, have great importance.

Accompanying drawing explanation

The method flow diagram one of the recipe drug attribute quantivative approach based on weighting PageRank algorithm that Fig. 1 provides for the embodiment of the present invention;

The structural drawing that interlinks of the web page joint that Fig. 2 provides for the embodiment of the present invention;

The link structure schematic diagram that the medicine attribute node that Fig. 3 provides for the embodiment of the present invention builds;

The method flow diagram two of the recipe drug attribute quantivative approach based on weighting PageRank algorithm that Fig. 4 provides for the embodiment of the present invention;

The structural representation of the recipe drug attribute quantitative system based on weighting PageRank algorithm that Fig. 5 provides for the embodiment of the present invention.

Embodiment

For making the technical problem to be solved in the present invention, technical scheme and advantage clearly, be described in detail below in conjunction with the accompanying drawings and the specific embodiments.

The present invention is directed to the interphase interaction that existing traditional recipe drug attribute quantivative approach have ignored medicine attribute, cause the problem that the Relative drug property value degree of accuracy that obtains is low, a kind of recipe drug attribute quantivative approach based on weighting PageRank algorithm and system are provided.

Embodiment one

Shown in Fig. 1, a kind of recipe drug attribute quantivative approach based on weighting PageRank algorithm that the embodiment of the present invention provides, comprising:

S1: decomposed by the medicine attribute of medicine each in prescription, determines the medicine attribute of prescription and the weights of described medicine attribute;

S2: using the medicine attribute in prescription as the medicine attribute node in link structure, according to the interaction between different pharmaceutical attribute, builds the link structure figure between described medicine attribute node;

S3: according to the weights of the medicine attribute determined, utilizes weighting PageRank algorithm to carry out iteration to medicine attribute node, determines the PR value of medicine attribute.

The recipe drug attribute quantivative approach based on weighting PageRank algorithm described in the embodiment of the present invention, by the medicine attribute of medicine each in prescription is decomposed, determine the medicine attribute of prescription and the weights of described medicine attribute, and using the medicine attribute in prescription as the medicine attribute node in link structure, according to the interaction between different pharmaceutical attribute, build the link structure figure between described medicine attribute node; Again according to the weights of the medicine attribute determined, utilize weighting PageRank algorithm to carry out iteration to medicine attribute node, determine the PR value of medicine attribute.Like this, the present invention not only considers the action effect between the medicine attribute of every herbal medicine self, interaction between the medicine attribute considering different pharmaceutical in prescription according to link structure again, perfect traditional recipe drug attribute quantivative approach ignores interactional deficiency between medicine attribute; And the PR value of the medicine attribute utilizing the PageRank algorithm of weighting to determine, this PR value can sort to prescription important attribute more accurately, for tcm prescription medicine and the quantitative examination of medicine attribute open a new thinking, advance the paces of TCM Modernization, have great importance.

In order to better understand page rank (PageRank) algorithm, first PageRank algorithm is simply introduced:

In the embodiment of the present invention, the thought of PageRank algorithm is the analysis based on replica detection, and namely the importance (PageRank value or PR value) of a webpage is that webpage quantity by pointing to it is weighed, and quantity is more much more important; On the other hand, the importance pointing to its webpage is higher, and this webpage is more important, and namely also very important by the webpage of important web page interlinkage, webpage is more important, and its rank also can be more forward.

Shown in Fig. 2, webpage B is pointed in a link of webpage A, webpage B just obtains importance score value PR (A)/C (A) that webpage A provides, if the importance of webpage A self is larger, it is supplied to importance score value PR (the A)/C (A) of webpage B also can be larger.Therefore, the importance size of webpage B depends on the importance of self and points to its other Web page importances, is depicted as the PR value result of calculation after the effect of interlinking of four webpages referring to table 1.The linking relationship on internet between webpage can be reflected to a certain extent by PageRank algorithm, effectively can excavate important high-quality webpage from the link structure between webpage.

Table 1: the PR value after four webpage effects of interlinking

PR value iterative computation	PR(A)	PR(B)	PR(C)	PR(D)
					1	1	1	1	1
2	0.575	1.28333	0.85833	0.43333
					…	…	…	…	…
12	0.26327	0.29784	0.21649	0.22239
					13	0.26324	0.29786	0.21649	0.22239

Shown in Fig. 3, in the embodiment of the present invention, by using for reference PageRank theory of algorithm and link structure thereof, assuming that whole medicine attributes of medicine self are mutually related, medicine attribute between recipe drug is interactional, the interaction relationship that can obtain between recipe drug attribute is a kind of netted link structure, and prescription action effect is the result after of whole medicine attribute in prescription interacts.According to the mutual relationship in prescription between different pharmaceutical attribute, using the whole medicine attributes in prescription as the medicine attribute node in link structure, the link structure figure between described medicine attribute node can be built.Recycling weighting PageRank algorithm simulates the interaction relationship between recipe drug attribute, and in conjunction with traditional recipe drug attribute quantivative approach, the continuous iteration calculation of application weighting PageRank algorithm, until medicine attribute PR value reaches steady state (SS).

In the embodiment of the aforementioned recipe drug attribute quantivative approach based on weighting PageRank algorithm, alternatively, the described medicine attribute by medicine each in prescription decomposes, and determines that the medicine attribute of prescription and the weights of described medicine attribute comprise:

Decomposed by the medicine attribute of medicine each in prescription, described medicine attribute comprises: the property of medicine, flavour of a drug and return through three class medicine attributes;

Determine the weights of the medicine attribute of every herbal medicine;

The weights of same medicine attribute in prescription are added, determine the medicine attribute of prescription and the weights of described medicine attribute.

Shown in Fig. 4, in the embodiment of the present invention, prescription, for treatise on Febrile Diseases the 12nd article of Guizhi decoction, extracts each medicine in prescription, makes the every herbal medicine in the prescription corresponding property of medicine, flavour of a drug, returns through three class medicine attributes.And the dosage unit of different pharmaceutical in prescription to be standardized into " gram " as unit, suppose, prescription Guizhi decoction formula comprise: cassia twig 15g, root of herbaceous peony 15g, honey-fried licorice root 10g, ginger 15g, date 9g; The medicine Attribute decomposition of Guizhi decoction is as follows:

Cassia twig: temperature, pungent, sweet, the heart, lung, bladder;

The root of herbaceous peony: be slightly cold, bitter, acid, liver, spleen;

Ginger: tepor, pungent, lung, spleen, stomach;

Honey-fried licorice root: flat, sweet, the heart, spleen, lung, stomach;

Date: temperature, sweet, spleen, stomach.

Then, analyzing and processing is carried out to medicine dose, utilize traditional recipe drug attribute quantivative approach that the actual dose of medicine is standardized into relative dose and normalization in the embodiment of the present invention, normalized for each medicine relative dose value assignment is given the weights of medicine attribute as described medicine attribute of its correspondence.

Be added by the weights of same medicine attribute in prescription, count the weights of whole medicine attribute and described medicine attribute in prescription, table 2 is the weights of whole medicine attribute and relative medicine attribute in prescription.

Table 2: the weights of whole medicine attribute and medicine attribute in prescription

In the embodiment of the aforementioned recipe drug attribute quantivative approach based on weighting PageRank algorithm, alternatively, describedly determine that the weights of the medicine attribute of every herbal medicine comprise:

Utilize traditional recipe drug attribute quantivative approach that the actual dose of herbal medicine every in prescription is standardized as relative dose and normalization;

Relative dose value assignment after normalization is given the weights of medicine attribute as described medicine attribute of its correspondence;

Wherein, the mathematical model actual dose of herbal medicine every in prescription being standardized as relative dose is as follows:

If then X=kx (x-2b);

If $M<\sqrt{10}m,$ Then $X=\left\{\begin{array}{c}kx(x-2b),x>m\\ \frac{10x}{m},x\&le;m\end{array}\right.;$

If M>=10m, then

In formula, $k=\frac{10(10m-M)}{Mm(M-m)},b=\frac{10m^2-M^2}{2(10m-M)},$ X is the relative dose of medicine, and x is the actual dose of medicine, and M is medicine research on maximum utilized quantity, and m is medicine minimum amount.

In the embodiment of the present invention, prescription is set about with the formula of Guizhi decoction, carries out pre-service to prescription Chinese traditional medicine dose, determines the relative dose of recipe drug.Because, in tcm prescription, each medicine has a dose proportioning of comparatively generally acknowledging, list can not judge the drug effect size of medicine from the actual dose size of medicine or be called contribution degree size, namely in same prescription, two or more medicine has identical actual dose, but they may be different to the contribution degree of prescription.This just needs to utilize the relative dose of prescription Chinese traditional medicine instead of actual dose under same standard, carry out the comparison of contribution degree size.

In the embodiment of the present invention, utilize traditional recipe drug attribute quantivative approach, relative to dose mathematics computing model formula, the actual dose of herbal medicine every in prescription is standardized as relative dose and normalization by recipe drug, wherein, described recipe drug is as follows relative to dose mathematics computing model:

If then X=kx (x-2b);

If $M<\sqrt{10}m,$ Then $X=\left\{\begin{array}{c}kx(x-2b),x>m\\ \frac{10x}{m},x\&le;m\end{array}\right.;$

If M>=10m, then

In formula, $k=\frac{10(10m-M)}{Mm(M-m)},b=\frac{10m^2-M^2}{2(10m-M)},$ X is the relative dose of medicine, and x is the actual dose of medicine, and M is medicine research on maximum utilized quantity, and m is medicine minimum amount.

Then, the relative dose value assignment after normalization is given the weights of medicine attribute as described medicine attribute of its correspondence.

In the embodiment of the aforementioned recipe drug attribute quantivative approach based on weighting PageRank algorithm, alternatively, the weights of the described medicine attribute according to determining, utilize weighting PageRank algorithm to carry out iteration to medicine attribute node, determine that the PR value of medicine attribute comprises:

Quantize medicine attribute node;

Iteration is carried out, until PR value tends towards stability by the PR value of PR value computing formula to each medicine attribute quantized of medicine attribute;

Wherein, the PR value computing formula of medicine attribute is as follows:

$PR\left(p\right)=(1-d)weight+d\underset{i=1}{\overset{n}{\&Sigma;}}\frac{PR\left(T_i\right)}{C\left(T_i\right)}$

In formula, PR (p) represents the PR value of medicine attribute node p, and n is the number of the other drug attribute node pointing to medicine attribute p, T _i(i=1,2 ..., n) represent the other drug attribute pointing to medicine attribute p, PR (T _i) represent medicine attribute T _ipR value, C (T _i) be medicine attribute T _ithe number of links outwards pointed out, weight is the weights of medicine attribute p, and d is boundary in (0,1) interval attenuation coefficient, is generally 0.85.

In the embodiment of the present invention, first quantize medicine attribute node, then carry out iteration, until PR value tends towards stability by the PR value of PR value computing formula to each medicine attribute quantized of medicine attribute; Wherein, the PR value computing formula of medicine attribute is as follows:

$PR\left(p\right)=(1-d)weight+d\underset{i=1}{\overset{n}{\&Sigma;}}\frac{PR\left(T_i\right)}{C\left(T_i\right)}$

In formula, PR (p) represents the PR value of medicine attribute node p, and n is the number of the other drug attribute node pointing to medicine attribute p, T _i(i=1,2 ..., n) represent the other drug attribute pointing to medicine attribute p, PR (T _i) represent medicine attribute T _ipR value, C (T _i) be medicine attribute T _ithe number of links outwards pointed out, weight is the weights of medicine attribute p, and d is boundary in (0,1) interval attenuation coefficient, is generally 0.85.

In the embodiment of the aforementioned recipe drug attribute quantivative approach based on weighting PageRank algorithm, alternatively, the weights of the described medicine attribute according to determining, utilize weighting PageRank algorithm to carry out iteration to medicine attribute node, comprise after determining the PR value of medicine attribute:

Size according to the PR value of the medicine attribute determined sorts, and determines the prescription important attribute after the acting in conjunction of described medicine attribute.

In the embodiment of the present invention, traditional recipe drug attribute quantivative approach is depicted as referring to table 3, the recipe drug attribute quantivative approach of non-weighting PageRank algorithm and the PR value of medicine attribute calculated based on the recipe drug attribute quantivative approach of weighting PageRank algorithm, three kinds of quantivative approachs all correctly effectively can judge the prescription three class important attribute (property of medicine, flavour of a drug and return through), but after carrying out comparative analysis by the PR value of the medicine attribute determined three kinds of quantivative approachs in correctness and degree of accuracy, show that the PR value of the medicine attribute that the recipe drug attribute quantivative approach based on the PageRank algorithm of weighting calculates can more accurately and sort to prescription important attribute accurately, three class important attribute features of prescription can be judged according to the important attribute after sequence, also can infer the drug effect of prescription further and cure mainly disease type, the PR value of the medicine attribute that is calculated by the recipe drug attribute quantivative approach of the PageRank algorithm based on weighting is more accurate.Application data digging technology of the present invention-solve recipe drug attribute quantitative problem in traditional Chinese medical science field based on the recipe drug attribute quantivative approach of weighting PageRank algorithm, for tcm prescription medicine and the quantitative examination of medicine attribute open a new thinking, advance the paces of TCM Modernization, have great importance in science of TCM formulas theoretical research.

Table 3: the PR value of the medicine attribute that three kinds of different basis weights methods are determined

In the embodiment of the present invention, set about from traditional Chinese medicine prescription Guizhi decoction formula, extract each medicine in prescription, each medicine is resolved into the property of medicine, flavour of a drug respectively, returns through three generic attributes, in conjunction with traditional recipe drug attribute quantivative approach and weighting PageRank theory of algorithm and link structure thereof, quantize medicine attribute node and iteration calculation is carried out, until quantized value reaches stable state to quantized value (that is: PR value).Like this, the present invention is not only perfect, and traditional recipe drug attribute quantivative approach ignores interactional deficiency between medicine attribute, and the PR value adopting weighting PageRank algorithm to calculate can sort to prescription important attribute more accurately, for tcm prescription medicine and the quantitative examination of medicine attribute open a new thinking.

Embodiment two

The present invention also provides a kind of embodiment of the recipe drug attribute quantitative system based on weighting PageRank algorithm, because the recipe drug attribute quantitative system based on weighting PageRank algorithm provided by the invention is corresponding with the embodiment of the aforementioned recipe drug attribute quantivative approach based on weighting PageRank algorithm, object of the present invention should can be realized by the process step performed in said method embodiment based on the recipe drug attribute quantitative system of weighting PageRank algorithm, therefore above-mentioned based on the explanation explanation in the recipe drug attribute quantivative approach embodiment of weighting PageRank algorithm, also the embodiment of the recipe drug attribute quantitative system based on weighting PageRank algorithm provided by the invention is applicable to, to repeat no more in embodiment below the present invention.

Shown in Fig. 5, the embodiment of the present invention also provides a kind of recipe drug attribute quantitative system based on weighting PageRank algorithm, comprising:

Recipe drug attribute determining unit: for being decomposed by the medicine attribute of medicine each in prescription, determines the medicine attribute of prescription and the weights of described medicine attribute;

Link structure determining unit: for using the medicine attribute in prescription as the medicine attribute node in link structure, according to the interaction between different pharmaceutical attribute, build the link structure figure between described medicine attribute node;

PR value determining unit: for the weights according to the medicine attribute determined, utilizes weighting PageRank algorithm to carry out iteration to medicine attribute node, determines the PR value of medicine attribute.

The recipe drug attribute quantitative system based on weighting PageRank algorithm described in the embodiment of the present invention, by the medicine attribute of medicine each in prescription is decomposed, determine the medicine attribute of prescription and the weights of described medicine attribute, and using the medicine attribute in prescription as the medicine attribute node in link structure, according to the interaction between different pharmaceutical attribute, build the link structure figure between described medicine attribute node; Again according to the weights of the medicine attribute determined, utilize weighting PageRank algorithm to carry out iteration to medicine attribute node, determine the PR value of medicine attribute.Like this, the present invention not only considers the action effect between the medicine attribute of every herbal medicine self, interaction between the medicine attribute considering different pharmaceutical in prescription according to link structure again, perfect traditional recipe drug attribute quantivative approach ignores interactional deficiency between medicine attribute; And the PR value of the medicine attribute utilizing the PageRank algorithm of weighting to determine, this PR value can sort to prescription important attribute more accurately, for tcm prescription medicine and the quantitative examination of medicine attribute open a new thinking, advance the paces of TCM Modernization, have great importance.

In the embodiment of the aforementioned recipe drug attribute quantitative system based on weighting PageRank algorithm, alternatively, described recipe drug attribute determining unit comprises:

Every herbal medicine medicine attribute determination module: for being decomposed by the medicine attribute of medicine each in prescription, described medicine attribute comprises: the property of medicine, flavour of a drug and return through three class medicine attributes;

Every herbal medicine medicine attribute weights determination module, for determining the weights of the medicine attribute of every herbal medicine;

Recipe drug attribute determination module: for being added by the weights of same medicine attribute in prescription, determines the medicine attribute of prescription and the weights of described medicine attribute.

In the embodiment of the aforementioned recipe drug attribute quantitative system based on weighting PageRank algorithm, alternatively, described every herbal medicine medicine attribute weights determination module comprises:

Normalization submodule: the actual dose of herbal medicine every in prescription is standardized as relative dose and normalization for utilizing traditional recipe drug attribute quantivative approach;

Every herbal medicine medicine attribute weights determination submodule: for the relative dose value assignment after normalization being given the weights of medicine attribute as described medicine attribute of its correspondence;

Wherein, the mathematical model actual dose of herbal medicine every in prescription being standardized as relative dose is as follows:

If then X=kx (x-2b);

If $M<\sqrt{10}m,$ Then $X=\left\{\begin{array}{c}kx(x-2b),x>m\\ \frac{10x}{m},x\&le;m\end{array}\right.;$

If M>=10m, then

In formula, $k=\frac{10(10m-M)}{Mm(M-m)},b=\frac{10m^2-M^2}{2(10m-M)},$ X is the relative dose of medicine, and x is the actual dose of medicine, and M is medicine research on maximum utilized quantity, and m is medicine minimum amount.

In the embodiment of the aforementioned recipe drug attribute quantitative system based on weighting PageRank algorithm, alternatively, described PR value determining unit comprises:

Quantization modules: for quantizing medicine attribute node;

PR value determination module: carry out iteration, until PR value tends towards stability for the PR value of PR value computing formula to each medicine attribute quantized by medicine attribute;

Wherein, the PR value computing formula of medicine attribute is as follows:

$PR\left(p\right)=(1-d)weight+d\underset{i=1}{\overset{n}{\&Sigma;}}\frac{PR\left(T_i\right)}{C\left(T_i\right)}$

In formula, PR (p) represents the PR value of medicine attribute node p, and n is the number of the other drug attribute node pointing to medicine attribute p, T _i(i=1,2 ..., n) represent the other drug attribute pointing to medicine attribute p, PR (T _i) represent medicine attribute T _ipR value, C (T _i) be medicine attribute T _ithe number of links outwards pointed out, weight is the weights of medicine attribute p, and d is boundary in (0,1) interval attenuation coefficient, is generally 0.85.

In the embodiment of the aforementioned recipe drug attribute quantitative system based on weighting PageRank algorithm, alternatively, described system also comprises:

Prescription important attribute determining unit: the size for the PR value according to the medicine attribute determined sorts, determines the prescription important attribute after the acting in conjunction of described medicine attribute.

The above is the preferred embodiment of the present invention; it should be pointed out that for those skilled in the art, under the prerequisite not departing from principle of the present invention; can also make some improvements and modifications, these improvements and modifications also should be considered as protection scope of the present invention.

Claims (10)

1., based on a recipe drug attribute quantivative approach for weighting PageRank algorithm, it is characterized in that, comprising:

The medicine attribute of medicine each in prescription is decomposed, determines the medicine attribute of prescription and the weights of described medicine attribute;

Using the medicine attribute in prescription as the medicine attribute node in link structure, according to the interaction between different pharmaceutical attribute, build the link structure figure between described medicine attribute node;

According to the weights of the medicine attribute determined, utilize weighting PageRank algorithm to carry out iteration to medicine attribute node, determine the PR value of medicine attribute.

2. method according to claim 1, is characterized in that, the described medicine attribute by medicine each in prescription decomposes, and determines that the medicine attribute of prescription and the weights of described medicine attribute comprise:

Decomposed by the medicine attribute of medicine each in prescription, described medicine attribute comprises: the property of medicine, flavour of a drug and return through three class medicine attributes;

Determine the weights of the medicine attribute of every herbal medicine;

The weights of same medicine attribute in prescription are added, determine the medicine attribute of prescription and the weights of described medicine attribute.

3. method according to claim 2, is characterized in that, describedly determines that the weights of the medicine attribute of every herbal medicine comprise:

Utilize traditional recipe drug attribute quantivative approach that the actual dose of herbal medicine every in prescription is standardized as relative dose and normalization;

Relative dose value assignment after normalization is given the weights of medicine attribute as described medicine attribute of its correspondence;

Wherein, the mathematical model actual dose of herbal medicine every in prescription being standardized as relative dose is as follows:

If then X=kx (x-2b);

If $M<\sqrt{10}m,$ Then $X=\left\{\begin{array}{c}kx(x-2b),x>m\\ \frac{10x}{m},x\&le;m\end{array}\right.;$

If M>=10m, then

In formula, $k=\frac{10(10m-M)}{Mm(M-m)},b=\frac{10m^2-M^2}{2(10m-M)},$ X is the relative dose of medicine, and x is the actual dose of medicine, and M is medicine research on maximum utilized quantity, and m is medicine minimum amount.

4. method according to claim 1, is characterized in that, the weights of the described medicine attribute according to determining, utilizes weighting PageRank algorithm to carry out iteration to medicine attribute node, determines that the PR value of medicine attribute comprises:

Quantize medicine attribute node;

Iteration is carried out, until PR value tends towards stability by the PR value of PR value computing formula to each medicine attribute quantized of medicine attribute;

Wherein, the PR value computing formula of medicine attribute is as follows:

$PR\left(p\right)=(1-d)weight+d\underset{i=1}{\overset{n}{\&Sigma;}}\frac{PR\left(T_i\right)}{C\left(T_i\right)}$

In formula, PR (p) represents the PR value of medicine attribute node p, and n is the number of the other drug attribute node pointing to medicine attribute p, T _i(i=1,2 ..., n) represent the other drug attribute pointing to medicine attribute p, PR (T _i) represent medicine attribute T _ipR value, C (T _i) be medicine attribute T _ithe number of links outwards pointed out, weight is the weights of medicine attribute p, and d is boundary in (0,1) interval attenuation coefficient, is generally 0.85.

5. method according to claim 1, is characterized in that, the weights of the described medicine attribute according to determining, utilizes weighting PageRank algorithm to carry out iteration to medicine attribute node, comprises after determining the PR value of medicine attribute:

Size according to the PR value of the medicine attribute determined sorts, and determines the prescription important attribute after the acting in conjunction of described medicine attribute.

6., based on a recipe drug attribute quantitative system for weighting PageRank algorithm, it is characterized in that, comprising:

Recipe drug attribute determining unit: for being decomposed by the medicine attribute of medicine each in prescription, determines the medicine attribute of prescription and the weights of described medicine attribute;

Link structure determining unit: for using the medicine attribute in prescription as the medicine attribute node in link structure, according to the interaction between different pharmaceutical attribute, build the link structure figure between described medicine attribute node;

PR value determining unit: for the weights according to the medicine attribute determined, utilizes weighting PageRank algorithm to carry out iteration to medicine attribute node, determines the PR value of medicine attribute.

7. system according to claim 6, is characterized in that, described recipe drug attribute determining unit comprises:

Every herbal medicine medicine attribute determination module: for being decomposed by the medicine attribute of medicine each in prescription, described medicine attribute comprises: the property of medicine, flavour of a drug and return through three class medicine attributes;

Every herbal medicine medicine attribute weights determination module, for determining the weights of the medicine attribute of every herbal medicine;

Recipe drug attribute determination module: for being added by the weights of same medicine attribute in prescription, determines the medicine attribute of prescription and the weights of described medicine attribute.

8. system according to claim 7, is characterized in that, described every herbal medicine medicine attribute weights determination module comprises:

Normalization submodule: the actual dose of herbal medicine every in prescription is standardized as relative dose and normalization for utilizing traditional recipe drug attribute quantivative approach;

Every herbal medicine medicine attribute weights determination submodule: for the relative dose value assignment after normalization being given the weights of medicine attribute as described medicine attribute of its correspondence;

Wherein, the mathematical model actual dose of herbal medicine every in prescription being standardized as relative dose is as follows:

If then X=kx (x-2b);

If $M<\sqrt{10}m,$ Then $X=\left\{\begin{array}{c}kx(x-2b),x>m\\ \frac{10x}{m},x\&le;m\end{array}\right.;$

If M>=10m, then

In formula, $k=\frac{10(10m-M)}{Mm(M-m)},b=\frac{10m^2-M^2}{2(10m-M)},$ X is the relative dose of medicine, and x is the actual dose of medicine, and M is medicine research on maximum utilized quantity, and m is medicine minimum amount.

9. system according to claim 6, is characterized in that, described PR value determining unit comprises:

Quantization modules: for quantizing medicine attribute node;

PR value determination module: carry out iteration, until PR value tends towards stability for the PR value of PR value computing formula to each medicine attribute quantized by medicine attribute;

Wherein, the PR value computing formula of medicine attribute is as follows:

$PR\left(p\right)=(1-d)weight+d\underset{i=1}{\overset{n}{\&Sigma;}}\frac{PR\left(T_i\right)}{C\left(T_i\right)}$

In formula, PR (p) represents the PR value of medicine attribute node p, and n is the number of the other drug attribute node pointing to medicine attribute p, T _i(i=1,2 ..., n) represent the other drug attribute pointing to medicine attribute p, PR (T _i) represent medicine attribute T _ipR value, C (T _i) be medicine attribute T _ithe number of links outwards pointed out, weight is the weights of medicine attribute p, and d is boundary in (0,1) interval attenuation coefficient, is generally 0.85.

10. system according to claim 6, is characterized in that, also comprises:

Prescription important attribute determining unit: the size for the PR value according to the medicine attribute determined sorts, determines the prescription important attribute after the acting in conjunction of described medicine attribute.