CN106126578B - A kind of web service recommendation method and device - Google Patents
A kind of web service recommendation method and device Download PDFInfo
- Publication number
- CN106126578B CN106126578B CN201610440017.8A CN201610440017A CN106126578B CN 106126578 B CN106126578 B CN 106126578B CN 201610440017 A CN201610440017 A CN 201610440017A CN 106126578 B CN106126578 B CN 106126578B
- Authority
- CN
- China
- Prior art keywords
- web service
- services composition
- vector
- moment
- services
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Abstract
The disclosure proposes a kind of web service recommendation method and device, utilize the information for the Web service that description text and the Services Composition of Web service, Services Composition use, the modeling of description text is completed by deep learning, the history use information of integrated service, description text, predefined auxiliary information train Web service recommendation model again, to realize the Web service recommendation of high accuracy.When Services Composition developer wishes that exploitation meets the Services Composition of some function, by the description text for submitting correlation function, one Services Composition is requested, the method automatically analyzes the functional character corresponding to it to the service in request, and then the Web service in service system is ranked up, Web service individual is recommended into Services Composition developer from high to low by the degree of agreeing with of Web service and the Services Composition, so that Services Composition developer be helped quickly and conveniently to complete the exploitation of Services Composition.
Description
Technical field
The disclosure belongs to computer system modeling and data analysis field, and in particular to a kind of web service recommendation method
And device.
Background technique
With service oriented computing (SOC, Service Oriented Computing), cloud computing (Cloud
Computing), continuing to bring out and being widely applied for the innovation modes such as networking (IoS, Internet of Service) of being engaged in, calculates
The trend of serviceization is fairly obvious.Services Oriented Achitecture (SOA, Service Oriented
Architecture) and business integration of the appearance between enterprises and enterprise of related protocol, specification provides convenient and fast pine
Coupled mode further pushes enterprise's production organization mode to change, to service-oriented enterprise (SOE, Service
Oriented Enterprise) transition.The business unit of itself is encapsulated into business service (Business by a large amount of enterprise
Service), and by internet realization enterprises different business unit and the business collaboration across business organization boundary with reality
Existing increase in value, so that the quantity of Web service on internet be made constantly to increase.In practical application, self-existent service is often
Granularity is smaller, is absorbed in limited single-minded function, it is difficult to meet the business demand of user's complexity.Therefore, it generally requires to service
Multiple services in system are combined use, to constitute the Services Composition (Service of increase in value
Compositions).Although above-mentioned state of development is encouraging, in reality operation, the Web of magnanimity takes Web service system
Business brings serious information overload, while the Web service individual of magnanimity provides abundant selection for user, also brings big
Measure unrelated or redundancy information.Therefore, how preferably utilize natural language semantic description information, for user request into
The effective service recommendation of row, is of great significance to the benign development of internet.
Although however, natural language be convenient for the mankind understanding and processing, also automatically processed to computer bring it is larger
Difficulty.For example, it is assumed that Services Composition developer wishes to develop a socially relevant application based on geographical location, about geography
Position, developer may use the nouns such as " GPS ", " geographical coordinate ", " position ", " orientation ", and about social activity, then may
Use the words such as " good friend ", " sharing ", " shared ".It is different from upper example it will be seen that being directed to same functional requirements
Services Composition developer may submit completely different keyword in user requests, and be corresponding to it, and the description of Web service is worked as
In it is also possible that describing identical function with diversified word.How to overcome synonym, ambiguity word problem, is to carry out standard
The Basic Problems of true service recommendation.
Summary of the invention
In order to better solve the above difficult point in Web service recommendation problem, the present disclosure proposes a kind of Web service recommendations
Method and device.
A kind of web service recommendation method, the method includes the following steps:
S100, for Services Composition involved in the request of Services Composition, extract the functional character of the Services Composition, obtain
Services Composition functional character vector is obtained, Z is denoted by;
S200, calculate each Web service from be published the moment to propose that the Services Composition request moment passed through when
Between length and each Web service from the moment is published to proposing the called number of Services Composition request moment institute;
Assuming that the sum of Web service is J, J is integer;By j-th of Web service from the moment is published to proposing Services Composition
Request elapsed time at moment length is denoted as uptj, j=1 ..., J;By j-th of Web service from the moment is published to proposing to take
The called number of business combination request moment institute is denoted as usgj, j=1 ..., J;
S300, for each Web service, utilize following formula to calculate pushing away for j-th of Web service Services Composition request relatively
Degree of recommending rj, j=1 ..., J:
rj=β1Zvj+β2uptj+β3usgj+β4
In formula:
β1, β2, β3, β4For known weight parameter;
vjFor known vector parameter relevant to j-th of Web service;
S400, Web service is exported according to its recommendation descending.
Preferably, the weight parameter β in the step S3001, β2, β3, β4And vector relevant to j-th of Web service
Parameter vj, j=1 ..., J obtained by following step:
S301, for each Web service, the function that each Web service is extracted from the description text of the Web service is special
Sign obtains Web service functional character vector;The corresponding Web service functional character vector of j-th of Web service is denoted as Yj, j=
1,…,J;
S302, prepare I Services Composition as sample, I is integer;
For each Services Composition in sample, its corresponding Services Composition is extracted from the description text of the Services Composition
Functional character vector;The corresponding Services Composition functional character vector of i-th of Services Composition is denoted as Xi, i ∈ { 1 ..., I };
S303, for each Services Composition in sample, calculate each Web service from being published the moment to the Services Composition
It is suggested elapsed time at moment length and each Web service and is suggested the moment from being published moment to the Services Composition
The called number of institute;
By j-th of Web service since be suggested being published constantly to i-th of Services Composition end passed through constantly when
Between length be denoted as uptij, i ∈ { 1 ..., I }, j ∈ { 1 ..., J };
J-th of Web service is suggested what cut-off constantly was called to i-th of Services Composition since being published constantly
Number is denoted as usgij, i ∈ { 1 ..., I }, j ∈ { 1 ..., J };
S304, following expressions is solved using optimal algorithmAnd then determine parameter beta1, β2, β3, β4And vj:
In formula:
rijRecommendation for j-th of Web service with respect to i-th of Services Composition;
λvFor regularization parameter.
Preferably, further include following step before the step S301:
S300.1, the description text of the description text of Services Composition and Web service is subjected to vectorization processing, is serviced
Combine initial vector, Web service initial vector;
S300.2, noise addition processing is carried out to the Services Composition initial vector, Web service initial vector, is taken
Business combination noise vector, Web service noise vector;
S300.3, one is established by L layers of deep neural network model, L is integer;By Services Composition noise vector, Web
Input of the noise vector as the model is serviced, using corresponding Services Composition initial vector, Web service initial vector as the mould
The anticipated output of type;
When the calculating of model output and anticipated output error are more than or equal to a given threshold value, return step
S300.2;Otherwise, step S300.4 is executed;
S300.4, the parameter for retaining the model;Also, by Services Composition noise vector the deep neural network [L/
2] the corresponding output of layer is used as the Services Composition functional character vector;By Web service noise vector in the deep neural network
The corresponding output of [L/2] layer be used as Web service functional character vector, wherein [L/2] indicates to be rounded L/2.
Preferably, the vectorization processing in the step S300.1 includes the following steps:
S300.1a, bag of words (bag-of-word) model is established;
S300.1b, convert the description text of the description text of Services Composition or Web service to the bag of words to
Amount form obtains Services Composition bag of words vector or Web service bag of words vector;
S300.1c, place is weighted using TF-IDF algorithm to Services Composition bag of words vector or Web service bag of words vector
Reason, obtains corresponding Services Composition weighing vector or Web service weighing vector;
S300.1d, maximum value normalization is carried out to Services Composition weighing vector or Web service weighing vector, to obtain
The corresponding Services Composition initial vector or Web service initial vector.
Preferably, the noise addition processing in the step S300.2 includes the following steps:
Noise level is arranged in S300.2a;
S300.2b at random will be in Services Composition initial vector, Web service initial vector by probability of the noise level
Element is set to 0, to obtain Services Composition noise vector, Web service noise vector.
The method is particularly designed for the background of Web service recommendation problem with characteristic, fully considers Web service
The coupled relation of text and usage record and the evolution properties of Web service system are described in recommendation problem, make full use of depth refreshing
The characteristic of effective feature representation can be extracted through network, carry out Web service recommendation.This method achieves in real data set
Preferable effect improves the accuracy and diversity of service recommendation.
Method of disclosure is using deep neural network model to the description text of Web service and Services Composition developer
User requests text to model, and is mapped to vector space;Using the history use information of service, excavate user's request with
Match pattern before different services;And history use information, the description text, predefined auxiliary information of integrated service,
Triplicity is got up, by the Web service in service system according to being arranged from high to low using probability by service combination and exploitation person
Sequence is recommended.
According to the above method, a kind of Web service recommendation device is realized, described device includes the first extraction unit, first
Computing unit, the second computing unit, third computing unit, the first output unit;
First extraction unit: for each Services Composition, the clothes are extracted from the description text of the Services Composition
The combined functional character of business, obtains Services Composition functional character vector;And export the corresponding Services Composition function of the Services Composition
Feature vector;
First computing unit: it for each Services Composition, calculates and exports each Web service from being published the moment
Elapsed time at moment length is suggested to the Services Composition;
Second computing unit: it for each Services Composition, calculates and exports each Web service from being published the moment
The number that the moment is called is suggested to a Services Composition;
The third computing unit: it for each Services Composition, is calculated in conjunction with first extraction unit, first single
The output of member, the second computing unit, calculates recommendation of each Web service with respect to the Services Composition;
First output unit: for each Services Composition, according to the recommendation that third computing unit calculates, descending
Export Web service;
The calculating of the recommendation is as follows:
Assuming that the sum of Web service is J, J is integer;It is defeated after one Services Composition is inputted first extraction unit
Services Composition functional character vector out is Z;After the Services Composition is inputted the first computing unit, j-th of Web service of output
Corresponding time span is uptj, j=1 ..., J;After the Services Composition is inputted the second computing unit, j-th of Web of output
Servicing corresponding number is usgj, j=1 ..., J;The then recommendation r of the relatively described Services Composition request of j-th of Web servicej's
Calculating formula are as follows:
rj=β1Zvj+β2uptj+β3usgj+β4
In formula:
β1, β2, β3, β4For known weight parameter;
V is known vector parameter relevant to the Web service.
Preferably, first extraction unit, further includes: each Web service to input, from the description of Web service
The functional character of each Web service is extracted in text, obtains Web service functional character vector;And it is corresponding to export the Web service
Web service functional character vector;
Parameter beta in the third computing unit1, β2, β3, β4And v, it is determined by the first training unit;
First training unit: being defeated with the output of the first extraction unit, the first computing unit, the second computing unit
Enter;Services Composition functional character vector corresponding to the Services Composition is obtained by the output of the first extraction unit, it is assumed that has I
The corresponding Services Composition functional character vector of i-th of Services Composition is denoted as X for training by a Services Compositioni, i=1 ..., I,
I is integer;Web service functional character vector corresponding to the Web service is obtained by the output of the first extraction unit, it is assumed that
The sum of Web service is J, and J is integer, and the corresponding Web service functional character vector of j-th of Web service is denoted as Yj, j=
1 ..., J, J are integer;Moment to the Services Composition is published certainly by each Web service of the first computing unit acquisition to be suggested
J-th of Web service is suggested the moment to i-th of Services Composition since being published constantly by elapsed time at moment length
Cut-off elapsed time length is denoted as uptij;Each Web service, which is obtained, by the second computing unit is published the moment to one certainly
A Services Composition is suggested the number being called at the moment, and j-th of Web service is serviced since being published constantly to i-th
Combination is suggested the number that cut-off is called constantly and is denoted as usgij;
Utilize the Services Composition functional character vector X of acquisitioni, i=1 ..., I, Web service functional character vector Yj, j=
1 ..., being published moment to the Services Composition is suggested elapsed time at moment length upt certainly for J, each Web serviceij, i=
Being published moment a to Services Composition is suggested what the moment was called certainly for 1 ..., I, j=1 ..., J and each Web service
Number usgij, i=1 ..., I, j=1 ..., J, use optimal algorithm solve following expressionsAnd then determine simultaneously output parameter
β1, β2, β3, β4And vj:
In formula:
rijRecommendation for j-th of Web service with respect to i-th of Services Composition;
λvFor regularization parameter.
Preferably, first extraction unit is one L layers of deep neural network model, and L is integer, which uses
Parameter pass through the second training unit obtain;First extraction unit using Services Composition, Web service as input;And it should
[L/2] layer of model exports the output as the first extraction unit, wherein [L/2] indicates to be rounded L/2;
Second training unit is one L layers of deep neural network model, and L is integer, by Services Composition noise to
Amount, input of the Web service noise vector as the model make corresponding Services Composition initial vector, Web service initial vector
For the anticipated output of the model;When the calculating output of the model threshold value given less than one with anticipated output error, will obtain
The model parameter output obtained;
The Services Composition noise vector, Web service noise vector are by will be at the beginning of Services Composition initial vector, Web service
Beginning vector input noise adds processing unit and carries out noise addition processing acquisition;
The Services Composition initial vector, Web service initial vector are by by the description text of Services Composition and Web service
Description text input vectorization processing unit carry out vectorization processing obtain.
Preferably, the noise adds processing unit: according to the noise level of setting, using the noise level as probability with
Element in Services Composition initial vector, Web service initial vector is set to 0 by machine, to obtain Services Composition noise vector, Web
Service noise vector.
Preferably, the vectorization processing unit: the description text of the description text of Services Composition or Web service is converted
For the vector form of bag of words, Services Composition bag of words vector or Web service bag of words vector are obtained;To the Services Composition word of acquisition
Bag vector or Web service bag of words vector using TF-IDF algorithm are weighted processing, obtain corresponding Services Composition weighing vector
Or Web service weighing vector;Services Composition weighing vector or Web service weighing vector to acquisition carry out maximum value normalization,
To obtain the corresponding Services Composition initial vector or Web service initial vector.
Detailed description of the invention
Fig. 1 is the web service recommendation method flow chart in an embodiment of the present disclosure;
Fig. 2 is the probability graph model schematic diagram of the Web service functional character extraction model in an embodiment of the present disclosure;
Fig. 3 is the probability graph model schematic diagram of the Web service recommendation model in an embodiment of the present disclosure;
Fig. 4 is the Web service recommendation apparatus structure schematic diagram in an embodiment of the present disclosure.
Specific embodiment
The embodiment of method of disclosure is introduced with reference to the accompanying drawing.
A kind of web service recommendation method, the method includes the following steps:
S100, for Services Composition involved in the request of Services Composition, mentioned from the description text of the Services Composition
The functional character of the Services Composition is taken, Services Composition functional character vector is obtained, is denoted by Z;
S200, calculate each Web service from be published the moment to propose that the Services Composition request moment passed through when
Between length and each Web service from the moment is published to proposing the called number of Services Composition request moment institute;
Assuming that the sum of Web service is J, J is integer;By j-th of Web service from the moment is published to proposing Services Composition
Request elapsed time at moment length is denoted as uptj, j=1 ..., J;By j-th of Web service from the moment is published to proposing to take
The called number of business combination request moment institute is denoted as usgj, j=1 ..., J;
S300, for each Web service, utilize following formula to calculate pushing away for j-th of Web service Services Composition request relatively
Degree of recommending rj, j=1 ..., J:
rj=β1Zvj+β2uptj+β3usgj+β4
In formula:
β1, β2, β3, β4For known weight parameter;
vjFor known vector parameter relevant to j-th of Web service;
S400, Web service is exported according to its recommendation descending.
Above method step combination Fig. 1 is seen:
One Services Composition is requested, the above method by Services Composition request in the functional character of Services Composition use
Vector indicates, obtains corresponding Services Composition functional character vector, is denoted by Z.
Meanwhile the above method calculates existing each Web service in Web service system and believes with respect to the auxiliary of the Services Composition
Breath, comprising:
(1) j-th of Web service, will from the moment is published to Services Composition request elapsed time at moment length is proposed
It is denoted as uptj, j=1 ..., J;
(2) j-th of Web service are remembered from the moment is published to the called number of Services Composition request moment institute is proposed
Make usgj, j=1 ..., J;
The Services Composition functional character vector Z and auxiliary information of acquisition are substituted into recommendation calculation formula, can be obtained
Each Web service is with respect to the recommendation that the Services Composition is requested, and from high to low according to recommendation, successively by Web service recommendation
Give Services Composition requestor.
One Services Composition is requested, in order to obtain accurate Web service recommendation, above-mentioned recommendation is calculated public
Formula needs to select suitable, appropriate weight parameter β1, β2, β3, β4And vector parameter vj, j=1 ..., J, and weight parameter
β1, β2, β3, β4And vector parameter vj, j=1 ..., J are not that all Web service systems are the same, it is also desirable to
It is determined according to different Web service systems.Therefore, it is necessary to history use information based on Web service, description text, pre-
The auxiliary information of definition, establishes Services Composition and the matching relationship before different Web services, and the matching relationship passes through recommendation
Degree is to embody, and with respect to one Services Composition recommendation of a Web service is high, then the matching degree of the Web service and the Services Composition
It is high.Preferably, the weight parameter β in the step S3001, β2, β3, β4And vector parameter relevant to j-th of Web service
vj, j=1 ..., J obtained by following step:
S301, for each Web service, the function that each Web service is extracted from the description text of the Web service is special
Sign obtains Web service functional character vector;The corresponding Web service functional character vector of j-th of Web service is denoted as Yj, j=
1,…,J;
S302, prepare I Services Composition as sample, I is integer;
For each Services Composition in sample, its corresponding Services Composition is extracted from the description text of the Services Composition
Functional character vector;The corresponding Services Composition functional character vector of i-th of Services Composition is denoted as Xi, i ∈ { 1 ..., I };
S303, for each Services Composition in sample, calculate each Web service from being published the moment to the Services Composition
It is suggested elapsed time at moment length and each Web service and is suggested the moment from being published moment to the Services Composition
The called number of institute;
By j-th of Web service since be suggested being published constantly to i-th of Services Composition end passed through constantly when
Between length be denoted as uptij, i ∈ { 1 ..., I }, j ∈ { 1 ..., J };
J-th of Web service is suggested what cut-off constantly was called to i-th of Services Composition since being published constantly
Number is denoted as usgij, i ∈ { 1 ..., I }, j ∈ { 1 ..., J };
S304, following expressions is solved using optimal algorithmAnd then determine parameter beta1, β2, β3, β4And vj:
In formula:
rijRecommendation for j-th of Web service with respect to i-th of Services Composition;
λvFor regularization parameter.
Above-mentioned steps S301-S304 can be considered the establishment process of a Web service recommendation model, and step S100-S400
It can be considered the application process of a Web service recommendation model.In step S301-S304, by a certain amount of Services Composition sample
Matching degree between each Web service in Web service system, by using a loss functionIt establishes, passes through
It is solved using optimal algorithm, can determine suitable, appropriate weight parameter β1, β2, β3, β4And vector parameter vj, j=1 ...,
J.The optimal algorithm such as uses gradient descent method.Particularly, in loss functionIn, regularization parameter λvIt is preferable to use values
It is 0.0001.
Fig. 2 is the probability graph model schematic diagram for illustrating Web service recommendation model, in the figure, XL/2For Services Composition function
Feature vector, YL/2For Web service functional character vector, s is the auxiliary information for describing Web service, and information above takes Web
It is all given for recommended models of being engaged in.According to information above, the hidden vector model v of Web service is constructed, the training of model is passed through
Determine the parameter of model v.
When Services Composition developer wishes that exploitation meets the Services Composition of some function, by submitting retouching for correlation function
Text, i.e. the description text of Services Composition are stated, the description text that the method can automatically analyze Services Composition is serviced
Combined functional character, and then the Web service in service system is ranked up, by agreeing with for Web service and the Services Composition
Degree Web service is successively recommended from high to low to Services Composition issue request developer, thus help developer quickly,
Easily complete the exploitation of Services Composition.
The corresponding Services Composition functional character vector of the request is obtained from a Services Composition request for convenience, and
In order to facilitate the Web service functional character vector for obtaining a Web service.Preferably, under further including before the step S301
State step:
S300.1, the description text of the description text of Services Composition and Web service is subjected to vectorization processing, is serviced
Combine initial vector, Web service initial vector;
S300.2, noise addition processing is carried out to the Services Composition initial vector, Web service initial vector, is taken
Business combination noise vector, Web service noise vector;
S300.3, one is established by L layers of deep neural network model, L is integer;By Services Composition noise vector, Web
Input of the noise vector as the model is serviced, using corresponding Services Composition initial vector, Web service initial vector as the mould
The anticipated output of type;
When the calculating of model output and anticipated output error are more than or equal to a given threshold value, return step
S300.2;Otherwise, step S300.4 is executed;
S300.4, the parameter for retaining the model;Also, by Services Composition noise vector the deep neural network [L/
2] the corresponding output of layer is used as the Services Composition functional character vector;By Web service noise vector in the deep neural network
The corresponding output of [L/2] layer be used as Web service functional character vector, wherein [L/2] indicates to be rounded L/2.
One Services Composition is requested, it is first determined Services Composition in the request, and then obtain retouching for the Services Composition
State text.
Above-mentioned steps S300.1-S300.4 can be considered as the training process that a Web service function extracts model, by step
For S300.3 it is found that it is a deep neural network model that the Web service function, which extracts model, which retouches Services Composition
The description text for stating text and Web service is modeled, and is mapped to the vector space of functional character, makes full use of depth
Study can extract the characteristic of effective feature representation.
The training process of the model is as shown in Figure 3.In the figure, the description text of Web service is subjected to vectorization processing
The Web service initial vector obtained afterwards, uses XcIt indicates;It will be to Web service initial vector XcIt is obtained after carrying out noise addition processing
Web service noise vector, uses X0It indicates.By X0As the input of deep neural network model, by XcAs deep neural network mould
The anticipated output of type, the training deep neural network.
The deep neural network model can be considered a maximized loss function:
The parameter for wherein needing to optimize is the relevant parameter W of deep neural network+={ Wl,bl, l=1,2 ..., L }, L is
The number of plies of deep neural network, fr(·,W+) use neural network parameter W+Calculate input reconstruct as a result, i.e. neural network L
The output of layer.In training, BP algorithm can be used.
When the calculating output of the deep neural network model threshold value given less than one with anticipated output error, stop instruction
Practice, and by output as trained model in the bottleneck layer of the deep neural network (Bottleneck Layer) output.
Specifically, [L/2] layer of the deep neural network can be appointed as bottleneck layer, wherein [L/2] indicates to be rounded L/2.
It is preferable to use stackings to denoise automatic coding machine model for the deep neural network model, but other depths also can be used
Spend neural network model.Text can be modeled unsupervisedly using stacking denoising automatic coding machine model, using more
It is convenient, and effect is good enough.And by way of adding noise, stacking denoising automatic coding machine model can simulate real world
When one service/service combination of middle description, and for the uncertainty of word, stacking denoising automatic coding machine training has Shandong
Stick.
After the Web service function extraction model is trained to, retain the parameters of the model.One is issued in user
After Services Composition request, the description text input Web service function of Services Composition involved in the request is extracted into model, is obtained
, in the vector of bottleneck layer output, which is that the Services Composition requests corresponding Services Composition functional character vector for it.
In one embodiment, the specific steps of the vectorization processing in step S300.1 are disclosed, comprising:
S300.1a, bag of words (bag-of-word) model is established;
S300.1b, convert the description text of the description text of Services Composition or Web service to the bag of words to
Amount form obtains Services Composition bag of words vector or Web service bag of words vector;
S300.1c, place is weighted using TF-IDF algorithm to Services Composition bag of words vector or Web service bag of words vector
Reason, obtains corresponding Services Composition weighing vector or Web service weighing vector;
S300.1d, maximum value normalization is carried out to Services Composition weighing vector or Web service weighing vector, to obtain
The corresponding Services Composition initial vector or Web service initial vector.
Wherein, the type of word includes service description language (sdl), service type, label, agreement in the bag of words.
In one embodiment, the specific method of the noise addition processing in step S300.2 is disclosed, comprising:
Noise level is arranged in S300.2a;
S300.2b at random will be in Services Composition initial vector, Web service initial vector by probability of the noise level
Element is set to 0, to obtain Services Composition noise vector, Web service noise vector.
Wherein, noise level is denoted as nl, then nl meets: 0≤nl≤1.In this embodiment, element is set to 0,
Think that the element is noise, it is also possible to set other values for noise.
According to disclosed method, realize a kind of Web service recommendation device, described device include the first extraction unit,
First computing unit, the second computing unit, third computing unit, the first output unit;
First extraction unit: for the description text of each Services Composition, the function of extracting the Services Composition is special
Sign obtains Services Composition functional character vector;And export the corresponding Services Composition functional character vector of the Services Composition;
First computing unit: it for each Services Composition, calculates and exports each Web service from being published the moment
Elapsed time at moment length is suggested to the Services Composition;
Second computing unit: it for each Services Composition, calculates and exports each Web service from being published the moment
The number that the moment is called is suggested to a Services Composition;
The third computing unit: it for each Services Composition, is calculated in conjunction with first extraction unit, first single
The output of member, the second computing unit, calculates recommendation of each Web service with respect to the Services Composition;
First output unit: for each Services Composition, according to the recommendation that third computing unit calculates, descending
Export Web service;
The calculating of the recommendation is as follows:
Assuming that the sum of Web service is J, J is integer;It is defeated after one Services Composition is inputted first extraction unit
Services Composition functional character vector out is Z;After the Services Composition is inputted the first computing unit, j-th of Web service of output
Corresponding time span is uptj, j=1 ..., J;After the Services Composition is inputted the second computing unit, j-th of Web of output
Servicing corresponding number is usgj, j=1 ..., J;The then recommendation r of the relatively described Services Composition request of j-th of Web servicej's
Calculating formula are as follows:
rj=β1Zvj+β2uptj+β3usgj+β4
In formula:
β1, β2, β3, β4For known weight parameter;
V is known vector parameter relevant to the Web service.
Preferably, first extraction unit, further includes: each Web service to input, from the description of Web service
The functional character of each Web service is extracted in text, obtains Web service functional character vector;And it is corresponding to export the Web service
Web service functional character vector;
Parameter beta in the third computing unit1, β2, β3, β4And v, it is determined by the first training unit;
First training unit: being defeated with the output of the first extraction unit, the first computing unit, the second computing unit
Enter;Services Composition functional character vector corresponding to the Services Composition is obtained by the output of the first extraction unit, it is assumed that has I
The corresponding Services Composition functional character vector of i-th of Services Composition is denoted as X for training by a Services Compositioni, i=1 ..., I,
I is integer;Web service functional character vector corresponding to the Web service is obtained by the output of the first extraction unit, it is assumed that
The sum of Web service is J, and J is integer, and the corresponding Web service functional character vector of j-th of Web service is denoted as Yj, j=
1 ..., J, J are integer;Moment to the Services Composition is published certainly by each Web service of the first computing unit acquisition to be suggested
J-th of Web service is suggested the moment to i-th of Services Composition since being published constantly by elapsed time at moment length
Cut-off elapsed time length is denoted as uptij;Each Web service, which is obtained, by the second computing unit is published the moment to one certainly
A Services Composition is suggested the number being called at the moment, and j-th of Web service is serviced since being published constantly to i-th
Combination is suggested the number that cut-off is called constantly and is denoted as usgij;
Utilize the Services Composition functional character vector X of acquisitioni, i=1 ..., I, Web service functional character vector Yj, j=
1 ..., being published moment to the Services Composition is suggested elapsed time at moment length upt certainly for J, each Web serviceij, i=
Being published moment a to Services Composition is suggested what the moment was called certainly for 1 ..., I, j=1 ..., J and each Web service
Number usgij, i=1 ..., I, j=1 ..., J, use optimal algorithm solve following expressionsAnd then determine simultaneously output parameter
β1, β2, β3, β4And vj:
In formula:
rijRecommendation for j-th of Web service with respect to i-th of Services Composition;
λvFor regularization parameter.When in use, regularization parameter λvIt is preferably arranged to 0.0001.
Preferably, first extraction unit is one L layers of deep neural network model, and L is integer, which uses
Parameter pass through the second training unit obtain;First extraction unit using Services Composition, Web service as input;And it should
[L/2] layer of model exports the output as the first extraction unit, wherein [L/2] indicates to be rounded L/2;
Second training unit is one L layers of deep neural network model, and L is integer, by Services Composition noise to
Amount, input of the Web service noise vector as the model make corresponding Services Composition initial vector, Web service initial vector
For the anticipated output of the model;When the calculating output of the model threshold value given less than one with anticipated output error, will obtain
The model parameter output obtained;
The Services Composition noise vector, Web service noise vector are by will be at the beginning of Services Composition initial vector, Web service
Beginning vector input noise adds processing unit and carries out noise addition processing acquisition;
The Services Composition initial vector, Web service initial vector are by by the description text of Services Composition and Web service
Description text input vectorization processing unit carry out vectorization processing obtain.
Preferably, the noise adds processing unit: according to the noise level of setting, using the noise level as probability with
Element in Services Composition initial vector, Web service initial vector is set to 0 by machine, to obtain Services Composition noise vector, Web
Service noise vector.
Preferably, the vectorization processing unit: the description text of the description text of Services Composition or Web service is converted
For the vector form of bag of words, Services Composition bag of words vector or Web service bag of words vector are obtained.Word in the bag of words
Type includes service description language (sdl), service type, label, agreement.To the Services Composition bag of words vector or Web service bag of words of acquisition
Vector is weighted processing using TF-IDF algorithm, obtains corresponding Services Composition weighing vector or Web service weighing vector;It is right
The Services Composition weighing vector or Web service weighing vector of acquisition carry out maximum value normalization, to obtain the corresponding clothes
Business combination initial vector or Web service initial vector.
Following example combination Fig. 4 describes a kind of usage mode of Web service recommendation device.When a user issues one
When a Services Composition is requested, user oneself is provided or Web service system automatically generates the description text an of Services Composition.It connects
The vectorization processing unit for described device of getting off carries out vectorization processing based on description text of the bag of words to the Services Composition,
And export the Services Composition and request corresponding Services Composition initial vector to the first extraction unit, the first extraction unit exports the clothes
Business combination requests corresponding Services Composition functional character vector to third computing unit.Third computing unit is calculated in conjunction with first
The auxiliary information that unit and the second computing unit calculate is calculated for the Web in the Web service system under Services Composition request
The recommendation of service.First output unit is according to the recommendation of Web service by Web service recommendation to user.For auxiliary information,
Assuming that there is J Web service in Web service system, then auxiliary information includes:
(1) j-th of Web service is long from elapsed time at the time of being published moment to user's sending Services Composition request
Degree, is denoted by uptij, i=1 ..., I, j=1 ..., J;
Institute's called number at the time of (2) j-th of Web service are published moment to user's sending Services Composition request certainly,
It is denoted by usgij, i=1 ..., I, j=1 ..., J.
From fig. 4, it can be seen that third computing unit is a part of Web service recommendation model, the first extraction unit is Web
A part of service function Feature Selection Model, they are to carry out corresponding position using the model parameter respectively having had determined that
Reason.In the case where not knowing these model parameters, or by described device be applied to a new Web service system when, Web
The first training unit can be used to obtain its model parameter in service recommendation model;And Web service functional character extraction model can
Its model parameter is obtained to use the second training unit.
The training that model is extracted for Web service functional character, prepares several history Services Compositions as training sample, obtains
Take they description text and they use Web service information.By the description text of Services Composition and retouching for Web service
It states text and carries out vectorization processing according to bag of words, obtain Services Composition initial vector and Web service initial vector.Go forward side by side one
After Services Composition initial vector and Web service initial vector input noise addition processing unit are carried out noise addition processing by step,
Obtain Services Composition noise vector and Web service noise vector.
The Web service functional character extracts model and belongs to deep neural network model, by the Services Composition noise of acquisition
Vector sum Web service noise vector extracts the training input data of model as Web service functional character, will be used to be serviced
Combine the Services Composition initial vector of noise vector, the Web service initial vector conduct pair for obtaining Web service noise vector
The anticipated output answered, Lai Xunlian Web service functional character extract model.When the expection that Web service functional character extracts model is defeated
When out and calculating the error threshold value given less than one of output, then show to have obtained the Web service function that can be used special
Sign extracts model.
Trained Web service functional character is extracted parameter possessed by model to export to the first extraction unit, the list
Member is one and the identical model of the second training unit.Meanwhile [L/2] layer that Web service functional character extracts model being exported
The output of model is extracted as trained Web service functional character, wherein [L/2] indicates to be rounded L/2.That is: group will be serviced
Output of the noise vector in [L/2] layer is closed as Services Composition functional character vector, by Web service noise vector in [L/2] layer
Output as Web service functional character vector.
For Services Composition functional character vector, the Web service functional character vector exported in the second training unit, will make
For the input training sample of the first training unit, to obtain the parameter of Web service recommendation model.Meanwhile first training unit also
It needs to calculate auxiliary information using the first computing unit and the second computing unit.Assuming that the Services Composition as training sample has I
A, then auxiliary information includes:
Certainly being published moment to i-th of Services Composition is suggested elapsed time at moment length for (1) j-th of Web service,
It is denoted by uptij, i=1 ..., I, j=1 ..., J;
(2) j-th of Web service are suggested moment called number from moment to i-th of Services Composition is published, by it
It is denoted as usgij, i=1 ..., I, j=1 ..., J.
When the trained parameter for searching out Web service recommendation model of the first training unit, the first training unit is defeated parameter
Third computing unit is given out.
A kind of web service recommendation method of the disclosure and device are described in detail above, it is used herein
Specific case is expounded the principle and embodiment of the disclosure, and the above embodiments are only used to help understand originally
The core concept of disclosed method, apparatus;At the same time, for those skilled in the art is having according to the thought of the disclosure
It has change in body embodiment and application range to point out, in conclusion the content of the present specification should not be construed as to the application
Limitation.
Claims (8)
1. a kind of web service recommendation method, which is characterized in that the method includes the following steps:
S100, for Services Composition involved in the request of Services Composition, extracting from the description text of the Services Composition should
The functional character of Services Composition obtains Services Composition functional character vector, is denoted by Z;
S200, each Web service is calculated from the moment is published to proposing that the elapsed time at Services Composition request moment is long
Degree and each Web service are from the moment is published to the called number of proposition Services Composition request moment institute;
If the sum of Web service is J, J is integer;By j-th of Web service from the moment is published to when proposing Services Composition request
It carves elapsed time length and is denoted as uptj, j=1 ..., J;By j-th of Web service from the moment is published to proposing Services Composition
The called number of request moment institute is denoted as usgj, j=1 ..., J;
S300, for each Web service, utilize following formula to calculate the recommendation of the relatively described Services Composition request of j-th of Web service
rj, j=1 ..., J:
rj=β1Zvj+β2uptj+β3usgj+β4
In formula:
β1, β2, β3, β4For weight parameter;
vjFor the relevant vector parameter of j-th of Web service;
S400, Web service is exported according to its recommendation descending;
Weight parameter β in the step S3001, β2, β3, β4And vector parameter v relevant to j-th of Web servicej, j=
1 ..., J are obtained by following step:
S301, for each Web service, the functional character of each Web service is extracted from the description text of the Web service, is obtained
Obtain Web service functional character vector;The corresponding Web service functional character vector of j-th of Web service is denoted as Yj, j=1 ...,
J;
S302, prepare I Services Composition as sample, I is integer;
For each Services Composition in sample, its corresponding Services Composition function is extracted from the description text of the Services Composition
Feature vector;The corresponding Services Composition functional character vector of i-th of Services Composition is denoted as Xi, i ∈ { 1 ..., I };
S303, for each Services Composition in sample, calculate each Web service and mentioned from being published moment to the Services Composition
Being published moment to the Services Composition is suggested moment institute's quilt certainly for elapsed time at moment length and each Web service out
Call number;
J-th of Web service is long since cut-off elapsed time constantly is suggested to i-th of Services Composition being published constantly
Degree is denoted as uptij, i ∈ { 1 ..., I }, j ∈ { 1 ..., J };
J-th of Web service is suggested the number that cut-off is called constantly to i-th of Services Composition since being published constantly
It is denoted as usgij, i ∈ { 1 ..., I }, j ∈ { 1 ..., J };
S304, following expressions is solved using optimal algorithmAnd then determine parameter beta1, β2, β3, β4And vj:
In formula:
rijRecommendation for j-th of Web service with respect to i-th of Services Composition;
λvFor regularization parameter.
2. the method according to claim 1, wherein further including following step before the step S301:
S300.1, the description text of the description text of Services Composition and Web service is subjected to vectorization processing, obtains Services Composition
Initial vector, Web service initial vector;
S300.2, noise addition processing is carried out to the Services Composition initial vector, Web service initial vector, obtains service group
Close noise vector, Web service noise vector;
S300.3, one is established by L layers of deep neural network model, L is integer;By Services Composition noise vector, Web service
Input of the noise vector as the model, using corresponding Services Composition initial vector, Web service initial vector as the model
Anticipated output;
When the calculating of model output and anticipated output error are more than or equal to a given threshold value, return step S300.2;
Otherwise, step S300.4 is executed;
S300.4, the parameter for retaining the model;Also, it is Services Composition noise vector is defeated in the correspondence of [L/2] layer of the model
It is used as the Services Composition functional character vector out;Corresponding output by Web service noise vector in [L/2] layer of the model is made
For Web service functional character vector, wherein [L/2] indicates to be rounded L/2.
3. according to the method described in claim 2, it is characterized in that, the vectorization processing in the step S300.1 includes following
Step:
S300.1a, bag of words are established;
S300.1b, the vector shape for converting the description text of the description text of Services Composition or Web service to the bag of words
Formula obtains Services Composition bag of words vector or Web service bag of words vector;
S300.1c, processing is weighted using TF-IDF algorithm to Services Composition bag of words vector or Web service bag of words vector, obtained
Obtain corresponding Services Composition weighing vector or Web service weighing vector;
S300.1d, maximum value normalization is carried out to Services Composition weighing vector or Web service weighing vector, to obtain corresponding
The Services Composition initial vector or Web service initial vector.
4. according to the method described in claim 2, it is characterized in that, the noise addition in the step S300.2 is handled under including
State step:
Noise level is arranged in S300.2a;
S300.2b is probability at random by the element in Services Composition initial vector, Web service initial vector using the noise level
It is set to 0, to obtain Services Composition noise vector, Web service noise vector.
5. a kind of Web service recommendation device, it is characterised in that:
Described device includes the first extraction unit, the first computing unit, the second computing unit, third computing unit, the first output
Unit;
First extraction unit: for each Services Composition, the service group is extracted from the description text of the Services Composition
The functional character of conjunction obtains Services Composition functional character vector;And export the corresponding Services Composition functional character of the Services Composition
Vector;
First computing unit: it for each Services Composition, calculates and exports each Web service and extremely should from the moment is published
Services Composition is suggested elapsed time at moment length;
Second computing unit: it for each Services Composition, calculates and exports each Web service from being published the moment to one
A Services Composition is suggested the number being called at the moment;
The third computing unit: for each Services Composition, in conjunction with first extraction unit, the first computing unit,
The output of two computing units calculates recommendation of each Web service with respect to the Services Composition;
First output unit: for each Services Composition, according to the recommendation that third computing unit calculates, descending output
Web service;
The calculating of the recommendation is as follows:
If the sum of Web service is J, J is integer;After one Services Composition is inputted first extraction unit, the clothes of output
Business combination function feature vector is Z;After the Services Composition is inputted the first computing unit, j-th of Web service of output is corresponding
Time span is uptj, j=1 ..., J;After the Services Composition is inputted the second computing unit, j-th of Web service pair of output
The number answered is usgj, j=1 ..., J;The then recommendation r of the relatively described Services Composition request of j-th of Web servicejCalculating formula
Are as follows:
rj=β1Zvj+β2uptj+β3usgj+β4
In formula:
β1, β2, β3, β4For known weight parameter;
vjFor known vector parameter relevant to the Web service;
First extraction unit, further includes: each Web service to input is extracted every from the description text of Web service
The functional character of a Web service obtains Web service functional character vector;And export the corresponding Web service function of the Web service
Feature vector;
Parameter beta in the third computing unit1, β2, β3, β4And vj, determined by the first training unit;
First training unit: being input with the output of the first extraction unit, the first computing unit, the second computing unit;It is logical
The output for crossing the first extraction unit obtains Services Composition functional character vector corresponding to the Services Composition, is equipped with I service
The corresponding Services Composition functional character vector of i-th of Services Composition is denoted as X for training by combinationi, i=1 ..., I, I is whole
Number;Web service functional character vector corresponding to the Web service is obtained by the output of the first extraction unit, if Web service
Sum be J, J is integer, and the corresponding Web service functional character vector of j-th of Web service is denoted as Yj, j=1 ..., J, J is
Integer;Each Web service is obtained by the first computing unit to be passed through from being published moment to the Services Composition and be suggested the moment
Time span, by j-th of Web service since being published constantly to i-th of Services Composition be suggested constantly cut-off passed through
Time span be denoted as uptij;Each Web service, which is obtained, by the second computing unit is published the moment to a Services Composition certainly
It is suggested the number that the moment is called, j-th of Web service is suggested since being published constantly to i-th of Services Composition
The number that moment cut-off is called is denoted as usgij;
Utilize the Services Composition functional character vector X of acquisitioni, i=1 ..., I, Web service functional character vector Yj, j=1 ...,
J, being published moment a to Services Composition is suggested elapsed time at moment length upt certainly for each Web serviceij, i=
Being published moment a to Services Composition is suggested what the moment was called certainly for 1 ..., I, j=1 ..., J and each Web service
Number usgij, i=1 ..., I, j=1 ..., J, use optimal algorithm solve following expressionsAnd then determine simultaneously output parameter
β1, β2, β3, β4And vj:
In formula:
rijRecommendation for j-th of Web service with respect to i-th of Services Composition;
λvFor regularization parameter.
6. device according to claim 5, it is characterised in that:
First extraction unit is one L layers of deep neural network model, and L is integer, and the parameter which uses passes through
Second training unit obtains;First extraction unit using Services Composition, Web service as input;And by [L/2] of the model
Output of the layer output as the first extraction unit, wherein [L/2] indicates to be rounded L/2;
Second training unit is one L layers of deep neural network model, and L is integer, by Services Composition noise vector,
Input of the Web service noise vector as the model, using corresponding Services Composition initial vector, Web service initial vector as
The anticipated output of the model;When the calculating output of the model threshold value given less than one with anticipated output error, will obtain
Model parameter output;
The Services Composition noise vector, Web service noise vector by by Services Composition initial vector, Web service initially to
Input noise addition processing unit progress noise addition processing is measured to obtain;
The Services Composition initial vector, Web service initial vector are by by the description text of Services Composition and retouching for Web service
It states text input vectorization processing unit and carries out vectorization processing acquisition.
7. device according to claim 6, it is characterised in that:
The noise adds processing unit: according to the noise level of setting, will service group at random by probability of the noise level
Close initial vector, the element in Web service initial vector is set to 0, with obtain Services Composition noise vector, Web service noise to
Amount.
8. device according to claim 6, it is characterised in that:
The vectorization processing unit: bag of words are converted by the description text of the description text of Services Composition or Web service
Vector form, obtain Services Composition bag of words vector or Web service bag of words vector;Services Composition bag of words vector to acquisition or
Web service bag of words vector is weighted processing using TF-IDF algorithm, obtains corresponding Services Composition weighing vector or Web service
Weighing vector;Services Composition weighing vector or Web service weighing vector to acquisition carry out maximum value normalization, to obtain phase
The Services Composition initial vector or Web service initial vector answered.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610440017.8A CN106126578B (en) | 2016-06-17 | 2016-06-17 | A kind of web service recommendation method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610440017.8A CN106126578B (en) | 2016-06-17 | 2016-06-17 | A kind of web service recommendation method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106126578A CN106126578A (en) | 2016-11-16 |
CN106126578B true CN106126578B (en) | 2019-07-19 |
Family
ID=57469977
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610440017.8A Active CN106126578B (en) | 2016-06-17 | 2016-06-17 | A kind of web service recommendation method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106126578B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108280104B (en) | 2017-02-13 | 2020-06-02 | 腾讯科技(深圳)有限公司 | Method and device for extracting characteristic information of target object |
CN107491992B (en) * | 2017-08-25 | 2020-12-25 | 哈尔滨工业大学(威海) | Intelligent service recommendation method based on cloud computing |
CN108287763A (en) * | 2018-01-29 | 2018-07-17 | 中兴飞流信息科技有限公司 | Parameter exchange method, working node and parameter server system |
CN110135747B (en) * | 2019-05-21 | 2023-05-05 | 成都派沃特科技股份有限公司 | Flow customization method based on neural network |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1503548A1 (en) * | 2003-08-01 | 2005-02-02 | fg microtec GmbH | Distributed Quality of Service Management System |
CN103544623A (en) * | 2013-11-06 | 2014-01-29 | 武汉大学 | Web service recommendation method based on user preference feature modeling |
CN103699667A (en) * | 2013-12-24 | 2014-04-02 | 天津大学 | Web service multi-dimensional semantic model building method |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100257131A1 (en) * | 2007-12-28 | 2010-10-07 | Kun-Oh Kim | Apparatus and method for controlling hybrid motor |
CN102231757B (en) * | 2011-06-29 | 2013-11-06 | 浙江大学 | On-line service combination recommendation system and recommendation method thereof |
CN102521283B (en) * | 2011-11-28 | 2014-11-12 | 浙江大学 | Service composition recommendation method based on Bayes principle, and system for the same |
CN103207858B (en) * | 2012-01-11 | 2016-03-23 | 富士通株式会社 | For recommending the apparatus and method of Web service combination |
CN102902740B (en) * | 2012-09-12 | 2015-05-20 | 北京航空航天大学 | Recommendation method and device for Web service combination |
CN104008195B (en) * | 2014-06-13 | 2017-06-06 | 清华大学 | The service recommendation method of service-oriented combination |
-
2016
- 2016-06-17 CN CN201610440017.8A patent/CN106126578B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1503548A1 (en) * | 2003-08-01 | 2005-02-02 | fg microtec GmbH | Distributed Quality of Service Management System |
CN103544623A (en) * | 2013-11-06 | 2014-01-29 | 武汉大学 | Web service recommendation method based on user preference feature modeling |
CN103699667A (en) * | 2013-12-24 | 2014-04-02 | 天津大学 | Web service multi-dimensional semantic model building method |
Non-Patent Citations (1)
Title |
---|
"全局视角下的Web服务系统模型及推荐策略";刘轶等;《计算机集成制造系统》;20160131;第22卷(第1期);第3.3、4,2节 |
Also Published As
Publication number | Publication date |
---|---|
CN106126578A (en) | 2016-11-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20220222920A1 (en) | Content processing method and apparatus, computer device, and storage medium | |
CN106126578B (en) | A kind of web service recommendation method and device | |
CN103544623B (en) | A kind of Web service recommendation method based on user preference feature modeling | |
CN103325061B (en) | A kind of community discovery method and system | |
CN104598611B (en) | The method and system being ranked up to search entry | |
CN103761254B (en) | Method for matching and recommending service themes in various fields | |
CN109416826A (en) | The system and method for search and matching content on social networks relevant to individual | |
CN110097125A (en) | A kind of across a network account correlating method indicated based on insertion | |
CN106709767A (en) | Personalized recommendation method and system of OTA (online travel website) hotels | |
CN106021366A (en) | API (Application Programing Interface) tag recommendation method based on heterogeneous information | |
CN108573041A (en) | Probability matrix based on weighting trusting relationship decomposes recommendation method | |
CN106779053A (en) | The knowledge point of a kind of allowed for influencing factors and neutral net is known the real situation method | |
CN112507106B (en) | Deep learning model training method and device and FAQ similarity discrimination method | |
CN107239489A (en) | The prediction of network public-opinion and emulation mode in accident based on SOAR models | |
CN108920521A (en) | User's portrait-item recommendation system and method based on pseudo- ontology | |
CN108628967A (en) | A kind of e-learning group partition method generating network similarity based on study | |
CN109411093A (en) | A kind of intelligent medical treatment big data analysis processing method based on cloud computing | |
CN106651427A (en) | Data association method based on user behavior | |
CN110245310A (en) | A kind of behavior analysis method of object, device and storage medium | |
CN109190116A (en) | Semantic analytic method, system, electronic equipment and storage medium | |
CN106055714A (en) | Method for capturing cloud calculating data from RIA (Rich Internet Application) page | |
CN116957128A (en) | Service index prediction method, device, equipment and storage medium | |
CN112035567B (en) | Data processing method, device and computer readable storage medium | |
CN103645889B (en) | Dynamic software self-adaption generating method | |
CN106599117B (en) | A kind of virtual ontology modeling method in Tea Science field |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |