CN105930406B - A kind of service recommendation method decomposed based on Poisson - Google Patents
A kind of service recommendation method decomposed based on Poisson Download PDFInfo
- Publication number
- CN105930406B CN105930406B CN201610237950.5A CN201610237950A CN105930406B CN 105930406 B CN105930406 B CN 105930406B CN 201610237950 A CN201610237950 A CN 201610237950A CN 105930406 B CN105930406 B CN 105930406B
- Authority
- CN
- China
- Prior art keywords
- service
- theme feature
- services composition
- theme
- recommendation
- 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
- 238000000034 method Methods 0.000 title claims abstract description 72
- 239000000203 mixture Substances 0.000 claims abstract description 46
- 238000009826 distribution Methods 0.000 claims abstract description 40
- 238000011156 evaluation Methods 0.000 claims abstract description 8
- 230000008569 process Effects 0.000 claims description 34
- 239000000284 extract Substances 0.000 claims description 11
- 238000004422 calculation algorithm Methods 0.000 claims description 10
- 230000004927 fusion Effects 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 4
- 238000000354 decomposition reaction Methods 0.000 claims description 2
- 238000012821 model calculation Methods 0.000 claims description 2
- 235000013399 edible fruits Nutrition 0.000 claims 1
- 230000002194 synthesizing effect Effects 0.000 abstract description 2
- 230000000694 effects Effects 0.000 description 7
- 230000009466 transformation Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000012067 mathematical method Methods 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
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
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
A kind of service recommendation method decomposed based on Poisson, the evaluation of record and user to Web service is called using the history of the description text of Web service, Web service, the theme distribution for respectively obtaining three services about Web, three theme distribution results are merged, the theme distribution as Web service；Using the issuing time information of existing Services Composition, the time series of Services Composition is generated.When developer proposes exploitation demand, the demand text that analysis developer proposes obtains the theme distribution of new demand servicing combination, and the joint probability distribution of " Services Composition-service " is calculated after synthesizing with the time series of the theme distribution of service, Services Composition.To obtain the Web service list of developer's demand, the sequence recommended from high to low is represented according to the sequence of probability value from big to small, the Web service list of recommendation is finally supplied to user.
Description
Technical field
The present invention is the service recommendation method of service-oriented combination and exploitation person a kind of.This method is used using the history of service
Record, description information and user decompose the theme feature of the service of extraction using Poisson, from subject layer to the comment information of service
The matching grating of service with Services Composition is excavated in face, final to propose Poisson service recommendation method.This method belongs to computer system
Modeling and data analysis field.
Background technique
As Service-Oriented Architecture Based (Service Oriented Architecture, hereinafter referred to as SOA) is widely applied,
Internet is undergoing the transformation by " data grid technology " to " service-centric ".A large amount of software suppliers on internet change
Become the rotating cylinder management mode of oneself, has been the mode for servicing (Software as a Service, abbreviation SaaS) with software, it will
The product of oneself is deployed on internet in the form of Web service.At the same time, developer utilizes SaaS moulds a large amount of on internet
The open service of formula develops oneself application, and suitable Web service is embedded in the program of oneself to reach quick exploitation and just
The purpose that victory uses.In the process, numerous Web service individuals is used in the form of dynamic combined by developer, forms
Services Composition or mashup, to generate increase in value.
The transformation of new software development model also brings new problem, and one side software supplier is use on the internet
Family provides all kinds of Web services of magnanimity, this serve individual Various Functions, service quality is also not quite similar, even similar
Web service also certainly exist subtle difference between them.This selects developer in the process of development properly
Web service process become very long and cumbersome, with the time develop Web service quantity also in rapid growth so that this
Process is more complicated.On the other hand, the functional requirement of developer's exploitation software application is often complicated and changeable, and user is difficult clearly
Clear statement needs the Web service of which speciality.
With an actual example come tool this problem.For example, developer will develop such a application: " fixed based on GPS
The social networking application of position, user can share geography information with the good friend in social networks ".In this segment description, developer
The Web service that demand adheres to three fields separately in fact is supported.First is that the relevant service in " map and positioning " field, to obtain geography
Information；Second is that the related service in " social network-i i-platform " field, to provide the interface of connection social network-i i-platform；Third is that " moving
The related service in dynamic terminal " field, because ' GPS positioning ' of user is likely to call the relevant field of some mobile terminals
Service.Then developer needs to find suitable Web service individual in these three fields, uses in the development plan of oneself.
In this process, there are two the technical issues of developer face: one is that accurately can refine and summarize development plan
The field (generation referred to as theme in method) of required Web service；Another is to find in the Web service of the field magnanimity
The serve individual properly and definitely needed.The two processes obviously have certain difficulty, and greatly increase the development cycle and
Increase development cost.Therefore, in the case where this information on services overloads and lacks unified information Description standard, how to utilize
Semantic description information carries out effective service recommendation for Services Composition, so that user is efficiently carried out Services Composition, to internet
Benign development be of great significance to.
Summary of the invention
In order to targetedly solve above-mentioned technical problem, the invention proposes a kind of service recommendation sides decomposed based on Poisson
Method.When Services Composition developer proposes the functional requirement of new demand servicing combination, service theme feature is extracted using this method, to open
Originator provides related service list, the effective development time for shortening Services Composition, reduces the development cycle.This method has taken into account clothes
The speed and accuracy requirement that business is recommended, achieves preferable effect in real data set.
Present invention firstly provides a kind of service recommendation methods decomposed based on Poisson.The entire algorithm stream of the method for the present invention
Journey is made of: 1) model generating process two parts；2) service recommendation process
1. model generating process
Model generating process includes three sub-stages:
A) theme feature serviced extracts
I. service theme feature is extracted from the description text of service.The Technique Using Both Text information of service is obtained, and utilizes Poisson
Decompose (Poisson Factorization, PF) by the Technique Using Both Text information MAP of each service to regular length to
In amount, formalized description is carried out with theme feature distribution of the matrix method to service.
Ii. record is called to extract service theme feature from the history of service.Service is extracted from the calling record of service
Theme feature distribution.
Iii. service theme feature is extracted from the evaluation of user.Service is extracted using PF algorithm from service evaluation
Theme feature distribution.
B) the theme feature fusion serviced
Three theme features about service were generated in the upper stage respectively to be distributed, this stage uses CMF (Collective
Matrix factorization) algorithm by three theme features be distributed merge.
C) time series for having Services Composition generates
According to the issuing time sequence of Services Composition, Services Composition is grouped according to the method for time slice, generates service
Combined time series.
2. service recommendation process
Service recommendation process includes two sub-stages:
A) theme feature of Services Composition extracts
Acquisition demand text extracts the theme feature of destination service combination using PF algorithm.
B) service list is recommended
The theme feature of integrated service combination, the theme feature of service and the time serial message of Services Composition, with general
The size order of rate Distribution Value finally returns that service recommendation list as order standard.
In summary content, method proposed by the invention and existing web service recommendation method (such as Mashup-
description-based Collaborative Filtering(MDCF)、Time-aware Collaborative
Domain Regression (TCDR) etc.) it compares, the present invention sufficiently excavates service usage record, the description of service text, Yong Huping
A variety of data informations such as valence, ensure that the accuracy of recommendation；Next has used the PF mathematical method for being more suitable for long-tail data, guarantees
The real-time recommended, has a good application prospect.
Detailed description of the invention
The application has been described in detail service recommendation method by attached drawing, these descriptions are merely to illustrate of the invention
Content is not intended to limit the present invention.
Fig. 1 is the service recommendation method process decomposed in the present invention based on Poisson.
Fig. 2 is that model generating process illustrates in the present invention.
Fig. 3 is that service recommendation process illustrates in the present invention.
Specific embodiment
Further specific description is made to the present invention below in conjunction with attached drawing.The table 1 of the end of writing is related to according in the present invention
Number specific meaning, and table 2 is for describing parameter equation of the present invention.
Fig. 1 describes the service recommendation method process decomposed based on Poisson.This method process includes extracting going through for Web service
History calls record, the text description of service and the user comment information of service, the comprehensive theme distribution for generating service；According to
There is the issuing time section of Services Composition to generate the time series of Services Composition.When developer proposes exploitation demand, method is generated
The theme distribution of Services Composition, then the service recommendation column that suitable user requires are provided after synthesizing with the result of model generating process
Table.
Method of the invention is made of two stages in order, and first stage is that model generates, and generates the master of Web service
The time series of topic distribution and Services Composition etc. is to be called；Second stage is service recommendation, is having developer to propose exploitation
When demand, demand is combined with the model calculation, provides service recommendation ranking results.
Fig. 2 describes model generating process.It is decomposed including application Poisson, calls record, clothes from the history of Web service
The text of business describes and the user comment information of service extracts the theme distribution of service, using Harmonious Matrix decomposition method,
In addition time serial message, finally obtains the theme distribution data of Web service.
Wherein model generating process includes three sub-stages:
A) theme feature serviced extracts.Service theme feature is extracted from the description text of service.Each Web service is being sent out
What developer can be described service function when cloth illustrates document.This stage obtains the Technique Using Both Text information of service, and benefit
(Poisson Factorization, PF) is decomposed by the Technique Using Both Text information MAP of each service to a fixed length with Poisson
On the vector of degree, formalized description is carried out with theme feature distribution of the matrix method to service.It calls and records from the history of service
Extract service theme feature.The outstanding Web service in part is called by existing Services Composition (mahsup), and algorithm is from calling
The theme feature distribution of such service is extracted in record.Service theme feature is extracted from the evaluation of user.User is browsing, is making
With some experience informations or opinion for service are often delivered after Web service, clothes equally are extracted using PF algorithm from evaluation
The theme feature of business is distributed.
B) three theme features about Web service are generated respectively in stage on the theme feature fusion serviced to be distributed, this
Stage is merged three feature distributions using CMF (Collective matrix factorization) algorithm.
C) time series for having Services Composition generates.According to the issuing time sequence of Services Composition, according to time slice
Method Services Composition is grouped, generate the time series of Services Composition.
Fig. 3 describes service recommendation process.It decomposes including application Poisson, is mentioned from the demand text that developer proposes
Take out the theme distribution of Services Composition (mashup), it is comprehensive by model generating process to Web service theme distribution number
According to finally providing the recommendation list of Web service.
Service recommendation process includes two sub-stages:
A) theme feature of Services Composition extracts.The demand text proposed using developer, extracts mesh using PF algorithm
Mark the theme feature of Services Composition.
B) service list is recommended.The theme feature of integrated service combination, the theme feature of service and Services Composition when
Between sequence information using the size order of probability distribution value as order standard finally return that Web service recommendation list.
Specific implementation steps are as follows
Step 0: coming into effect；
Complete model generating process in step 1~10
Step 1: empirically determined βw,k、μs,k、δs,k、∈s,kAnd ηm,kThe scale parameter (rte) of five Gamma distribution
And form parameter (shp).This step is the theme feature that word is arranged, the description text of service, the user's evaluation of service, service
History calls the basic parameter of record and the theme Gamma of fused service distribution, determines the initial shape of Gamma distribution
Shape.
Step 2: with the initial value of random value initialization Gamma distribution.The effect of this step is setting iterative initial value.
Step 3: according to table 2, the counting v of word is used for Web service descriptionsw, work as vswWhen > 0, more using iterative formula
New parameter.The effect of this step is to update the Poisson distribution of service describing word
Step 4: according to table 2, the counting c of word is used for the user comment of Web servicesw, work as cswWhen > 0, iteration is used
Formula undated parameter.The effect of this step is to update the Poisson distribution of user comment word
Step 5: according to table 2, the counting w of word is used for the description text of Services Compositionmw, work as wmwWhen > 0, using repeatedly
For formula undated parameter.The effect of this step is to update the description text word Poisson distribution of Services Composition
Step 6: according to table 2, calling the history of service to call record r Services Compositionms, work as rms> 0 uses iterative formula
Undated parameter.The effect of this step is the Poisson distribution for updating service history and calling
Step 7: according to table 2, using updated vsw、csw、wmwAnd rms, Gamma distribution is recalculated according to iterative formula
Parameter and.This step is the theme the iteration of feature.
Step 8: according to the calculated result of step 7, recalculating, that is, recalculate the theme Gamma of fused service
Distribution.This step is the theme feature fusion iteration of service.
Step 9: repeating step 1~8 until convergence.So far the theme feature for completing service extracts and the theme of service is special
Sign fusion.
Step 10: time series is calculated, is monthly divided into some time according to the time distribution order of Services Composition, it is right
In time period t, the publication month of Services Composition m is tm, corresponding time sequential value is Tm, formulaWherein λη、λtFor coefficient, ληIt can value 1, λtCan value 0.08, can appropriate adjustment.
tcurrentFor current time.This step is that the time series of existing Services Composition generates.
Complete service recommendation process in step 11~15.
Step 11: with the Services Composition Gamma of the model generating process scale parameter (rte) being distributed and form parameter
(shp) parameter of the Gamma distribution of the requirement documents proposed as initialization developer, with random value initialization new demand servicing combination
Gamma distribution.
Step 12: utilizing model generating process calculated result βw,k, work as wmwWhen > 0, updated according to table 2 using iterative formula
Parameter.This step is to calculate the Poisson distribution of the required Services Composition word of developer.
Step 13: using formulaUpdating developer needs
The parameter for asking document Gamma to be distributed.This step is the theme feature iteration of Services Composition.
Step 14: 12~13 are repeated, until convergence.So far the Services Composition feature extraction of user demand is completed.
Step 15: the new demand servicing combination m and service s of developer meets Poisson distributionBenefit
Use formulaIt calculates,The connection of " Services Composition-Web service " is obtained after being normalized
Probability distribution is closed, this value is supplied to developer as final service recommendation the results list, and numerical value is bigger to represent corresponding Web clothes
Business is more suitable for new Services Composition m.
For the ease of statement, symbol definition involved in specific steps is summarized as follows:
1 symbol definition table of table
2 parameter iteration formula of table
Claims (5)
1. a kind of service recommendation method decomposed based on Poisson, it is characterised in that the recommended method includes two processes: first
A process is model generating process, and it is to be called to generate the theme distribution of service and the time series of Services Composition etc.；Second
Process is service recommendation process, and in the demand of proposition, demand is combined with the model calculation, provides service recommendation sequence knot
Fruit；
First process the following steps are included:
A) the theme feature extraction step serviced,
B) the theme feature fusion steps serviced,
C) has the time series generation step of Services Composition；
Second process the following steps are included:
A) theme feature of Services Composition extracts,
B) service list is recommended；
B) the service list recommendation step includes: the theme feature of integrated service combination, the theme feature of service and service
Combined time serial message finally returns that Web service recommendation arranges using the size order of probability distribution value as order standard
Table.
2. service recommendation method according to claim 1, wherein a) the theme feature extraction step of service includes:
I. service theme feature is extracted from the description text of service,
Ii. record is called to extract service theme feature from the history of service,
Iii. service theme feature is extracted from the evaluation of user.
3. service recommendation method according to claim 1 or 2, wherein b) the theme feature fusion steps of service include: to make
Three theme feature distributions are merged with CMF (Collective matrix factorization) algorithm；
Three theme features refer to: the service theme feature extracted from the description text of service is called from the history of service
Record the service theme feature extracted, and the service theme feature extracted from the evaluation of user.
4. service recommendation method according to claim 1 or 2 wherein c) has the time series generation step of Services Composition
Include: the issuing time sequence according to Services Composition, Services Composition is grouped according to the method for time slice, generates Services Composition
Time series.
5. service recommendation method according to claim 1 or 2, wherein the theme feature of a) Services Composition extracts packet
It includes: obtaining demand text, the theme feature of destination service combination is extracted using Poisson decomposition algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610237950.5A CN105930406B (en) | 2016-04-15 | 2016-04-15 | A kind of service recommendation method decomposed based on Poisson |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610237950.5A CN105930406B (en) | 2016-04-15 | 2016-04-15 | A kind of service recommendation method decomposed based on Poisson |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105930406A CN105930406A (en) | 2016-09-07 |
| CN105930406B true CN105930406B (en) | 2019-03-22 |
Family
ID=56838272
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610237950.5A Active CN105930406B (en) | 2016-04-15 | 2016-04-15 | A kind of service recommendation method decomposed based on Poisson |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105930406B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102750360A (en) * | 2012-06-12 | 2012-10-24 | 清华大学 | Mining method of computer data for recommendation systems |
| CN103473291A (en) * | 2013-09-02 | 2013-12-25 | 中国科学院软件研究所 | Personalized service recommendation system and method based on latent semantic probability models |
| CN103544623A (en) * | 2013-11-06 | 2014-01-29 | 武汉大学 | Web service recommendation method based on user preference feature modeling |
| CN103577414A (en) * | 2012-07-20 | 2014-02-12 | 富士通株式会社 | Data processing method and device |
| CN103761254A (en) * | 2013-12-26 | 2014-04-30 | 清华大学 | Method for matching and recommending service themes in various fields |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6687696B2 (en) * | 2000-07-26 | 2004-02-03 | Recommind Inc. | System and method for personalized search, information filtering, and for generating recommendations utilizing statistical latent class models |
| CN103927354A (en) * | 2014-04-11 | 2014-07-16 | 百度在线网络技术（北京）有限公司 | Interactive searching and recommending method and device |
-
2016
- 2016-04-15 CN CN201610237950.5A patent/CN105930406B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102750360A (en) * | 2012-06-12 | 2012-10-24 | 清华大学 | Mining method of computer data for recommendation systems |
| CN103577414A (en) * | 2012-07-20 | 2014-02-12 | 富士通株式会社 | Data processing method and device |
| CN103473291A (en) * | 2013-09-02 | 2013-12-25 | 中国科学院软件研究所 | Personalized service recommendation system and method based on latent semantic probability models |
| CN103544623A (en) * | 2013-11-06 | 2014-01-29 | 武汉大学 | Web service recommendation method based on user preference feature modeling |
| CN103761254A (en) * | 2013-12-26 | 2014-04-30 | 清华大学 | Method for matching and recommending service themes in various fields |
Non-Patent Citations (1)
| Title |
|---|
| Probabilistic Factor Models for Web Site Recommendation;Hao Ma等;《Sigir》;20110728;摘要及第1、4节 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105930406A (en) | 2016-09-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105224377B (en) | A kind of method and device by metadata automatic generating software project code file | |
| CN105592150B (en) | Method, device and system for acquiring push data | |
| CN104038517A (en) | Information pushing method based on group relation and server | |
| EP2051193A1 (en) | A system and method for generating the descriptive information of the communication user | |
| US10354080B2 (en) | Facilitating offline or other contemporaneous editing of tabular data | |
| CN104834695A (en) | Activity recommendation method based on user interest degree and geographic position | |
| CN101673276B (en) | Method and device for dynamically generating data template | |
| CN104036415A (en) | Platform promotion method based on analysis of user behaviors | |
| CN105447253B (en) | Integration method of three-dimensional process data | |
| CN110516815A (en) | The characteristic processing method, apparatus and electronic equipment of artificial intelligence recommended models | |
| CN109145047A (en) | Configuration method, data processing equipment and the storage medium of user tag portrait | |
| CN105653535B (en) | A kind of media resource recommended method | |
| CN112182393B (en) | Product personalized customization system based on extension design | |
| CN108053486A (en) | Construction method and system based on virtual technology | |
| CN103593393A (en) | Social circle digging method and device based on microblog interactive relationships | |
| CN106557410A (en) | User behavior analysis method and apparatus based on artificial intelligence | |
| CN108536718A (en) | A kind of method and system for the IT application in management realized based on input and output semantization | |
| CN110532487A (en) | The generation method and device of label | |
| CN113722289A (en) | Method, device, electronic equipment and medium for constructing data service | |
| Beuche | Using pure: variants across the product line lifecycle | |
| CN114035793A (en) | Page generation method, page generation device, equipment and storage medium | |
| CN104981803B (en) | The method and system of vision enhancement type calendar is generated from calendar pre-existing on electronic device | |
| CN114595673A (en) | Collaborative creation method of digital works, task release method and device | |
| CN105930406B (en) | A kind of service recommendation method decomposed based on Poisson | |
| CN104267967B (en) | The generation method and device of the program code of software |
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 |