CN108595595A - A kind of user knowledge requirement acquisition method calculated based on interactive differential evolution - Google Patents
A kind of user knowledge requirement acquisition method calculated based on interactive differential evolution Download PDFInfo
- Publication number
- CN108595595A CN108595595A CN201810355129.2A CN201810355129A CN108595595A CN 108595595 A CN108595595 A CN 108595595A CN 201810355129 A CN201810355129 A CN 201810355129A CN 108595595 A CN108595595 A CN 108595595A
- Authority
- CN
- China
- Prior art keywords
- user
- vector
- knowledge
- document
- population
- 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.)
- Granted
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention discloses a kind of user knowledge requirement acquisition method calculated based on interactive differential evolution,Convert knowledge document to vector,Build text vector knowledge base,Existing document is divided into N classes using K means methods and generates one at random from every class,It generates N number of initial population document and is pushed to user,Then the calculating of differential evolution operator is carried out to initial population document,The corresponding fitness function of document is determined by the click behavior of user wherein in selection operation,The filial generation of generation is pushed to user again,Fitness function is determined according to the behavior of user and carries out the calculating of evolutionary operator again,The text data that user clicks when reaching user satisfaction,It can be used to the knowledge requirement model of training user,It is fitted the knowledge requirement of user,The present invention can be in the case where lacking user knowledge document scores data set,Solve the problems, such as user's cold start-up and project cold start-up simultaneously.
Description
Technical field
The invention belongs to knowledge services fields, and in particular to a kind of user knowledge need calculated based on interactive differential evolution
Seek acquisition methods.
Background technology
In the case where most enterprises possess huge knowledge quantity and good knowledge resource, how supplying system is utilized
By knowledge active push to designer to improve the emphasis that enterprises production efficiency is current knowledge-intensive enterprise's concern.Knowing
Know in supplying system, the task and behavioural information according to user push product design personnel.So how will be suitable
The problem of knowledge document is pushed to new user and how new knowledge document (i.e. new projects) is pushed to the people of needs, that is, use
Family/project (knowledge document) cold start-up problem is the critical issue of knowledge supplying system development.Existing cold start-up solution
Known users score data is needed mostly, and can only solve the problems, such as one of user, project cold start-up.For product design field
Knowledge supplying system cold start-up for, due to the missing of user-knowledge document-score data, existing cold start-up solution party
Case is simultaneously not suitable for.
Invention content
In view of this, the present invention provides a kind of user knowledge demand acquisition sides calculated based on interactive differential evolution
Method, can be under conditions of lacking user-knowledge document-score data, while solving user's cold start-up and project cold start-up is asked
Topic.
Realize that technical scheme is as follows：
A kind of user knowledge requirement acquisition method calculated based on interactive differential evolution, is included the following steps：
Step 1: converting knowledge document to vector, structure vector data library；
Step 2: gathering the vector in vector data library for N classes, a vector is randomly selected from every class, obtains N number of kind
Group's vector, is denoted as { xi(g) }, wherein xi(g) indicate in population i-th of g generations to
Amount, i=1,2 ..., N；G=0,1,2 ...；
Step 3: the corresponding knowledge document of population vector is pushed to user, setting user satisfaction index is clicking rate,
If user is not less than 80% to the clicking rate of the knowledge document pushed, step 6 is executed, if clicking rate is less than 80%, executes step
Rapid four-step 5；
Step 4: obtaining new N number of vector { u into row variation, intersection respectively to N number of population vector in g generations successivelyi(g+
1)}；
Using greedy algorithm from obtained { uiAnd { x (g+1) }i(g) } selection obtains next-generation population at individual in；
Wherein, f (ui(g+1)) individual u is indicatedi(g+1) fitness value；
Step 5: calculating N number of g+1 successively for the vector in population vector and vector data library using cosine similarity algorithm
Similarity, choose the knowledge document corresponding to most similar N number of vector, interaction times cumulative 1 execute step 3；
Step 6: being exported current N number of population vector as user knowledge demand model.
Further, the knowledge document includes pdf documents, word document and txt documents.
Further, in step 1, knowledge document is converted by vector using doc2vec (text vector conversion) technology.
Further, in step 2, vector is gathered for N classes using K-means methods.
Advantageous effect：
The present invention is for the first time combined evolution algorithm to obtain user knowledge demand with interactive mode；According to the click of user
Behavior determines fitness function and carries out the calculating of evolutionary operator, the textual data that user clicks when reaching user satisfaction again
According to, you can it is used for the knowledge requirement model of training user, is fitted the knowledge requirement of user.The present invention is realized in missing user-
Under conditions of knowledge document-score data, while solving the problems, such as user's cold start-up and project cold start-up.
Description of the drawings
Fig. 1 is the method for the present invention flow chart.
Specific implementation mode
The present invention will now be described in detail with reference to the accompanying drawings and examples.
As shown in Figure 1, the present invention provides a kind of user knowledge demand acquisition sides calculated based on interactive differential evolution
Method includes the following steps：
Step 1: being pre-processed to knowledge document using natural language processing technique, the knowledge document includes pdf texts
Shelves, word document and txt documents, the pretreatment include participle and removal stop words；Then doc2vec (text vectors are utilized
Conversion) technology converts pretreated knowledge document to vector.Knowledge document vector database is built, which is
The basic data of bottom layer treatment, knowledge document are the data being presented to the user.
Step 2: determine user single browsing items quantity N, setting interaction times initial value actNum=1.Wherein browse
Number of entries N refers to knowledge document number shown in the interface being presented to the user.Interaction times are defined as：Algorithm is often generated and is once pushed away
Recommend list, knowledge document be just pushed to user, user carries out a series of click to the knowledge document of recommendation, using difference into
Change the knowledge document that algorithm process user clicked and generates recommendation list again.Such algorithm push-user clicks
Process become primary interaction.Vector in vector data library is gathered using K-means methods for N classes, is taken out at random from every class
A vector is taken, N number of population vector is obtained, each vector is an individual, participates in subsequent evolution algorithm and calculates, is denoted as { xi
(g) }, wherein xi(g) i-th of vectorial, i=1,2 in g generations in population is indicated ..., N；G=0,1,2 ...；
Step 3: the corresponding knowledge document of population vector is pushed to user, setting user satisfaction index is clicking rate
CTR (Click-Through-Rate) executes step if user is not less than 80% to the clicking rate of the knowledge document pushed
Six, if clicking rate is less than 80%, execute step 4-step 5；
Step 4: carrying out differential evolution operator calculating.Its specific steps are：
Mutation operation is carried out respectively to N number of population vector in g generations successively and obtains intermediate { vi(g+1)}；
vi(g+1)=xr1(g)+F·xr2(g)-xr3(g)),i≠r1≠r2≠r3
Wherein, xr1(g) it is to wait for variation vector, r1=1,2 ..., N, x in g generationsr2(g) and xr3(g) it is g vectorial for population
In two different vectors, r2=1,2 ..., N, r3=1,2 ..., N, r1 ≠ r2 ≠ r3, F are zoom factor, are a constants；
To the institute directed quantity { x in populationiAnd the obtained intermediate { v of variation (g) }i(g+1) } crossover operation between progress individual
Obtain new N number of vector { ui(g+1)}；Wherein, ui(g+1) each element u inj,(g+1) it obtains according to the following formula, j=1,2 ...,
D, wherein D are vector element sum：
Wherein, CR is crossover probability, is setting value；vj,i(g+1) it is intermediate vi(g+1) j-th of element, xj,(g) it is
xi(g) j-th of element；If random number rand (0,1) is less than or equal to crossover probability CR, then the of i-th of g+1 generations vector
J element takes j-th of element of intermediate；Conversely, then taking j-th of element of i-th of vector in g generations；
Using greedy algorithm from obtained { uiAnd { x (g+1) }i(g) } selection obtains next-generation population at individual in；
Wherein, f (ui(g+1)) individual u is indicatedi(g+1) fitness value；
F (i) gets up the iterative process of evolution algorithm and user behavior informational linkage, user to the recommendation list of presentation into
Row is clicked, and clicking document by user corresponds to fitness value 1, and the corresponding fitness value of the text being not clicked on is 0.
Step 5: calculating N number of g+1 successively for the vector in population vector and vector data library using cosine similarity algorithm
Similarity, choose the knowledge document corresponding to most similar N number of vector, interaction times cumulative 1 execute step 3；
Step 6: being exported current N number of population vector as user knowledge demand model, and export current interaction time
Number.
One-step 6 of above-mentioned steps is the detailed process that user knowledge demand obtains, and can know which user knows
It is interested to know document.
In the present invention, population quantity N, zoom factor F and crossover probability CR can be tested according to real data collection, most
The parameter combination (population quantity N, zoom factor F and crossover probability CR) for reaching user satisfaction in small interaction times is best
Optimisation strategy value.
, can be by its knowledge requirement of this method quick obtaining when new user enters system, its knowledge requirement mould of training
Suitable knowledge is pushed to new user by type in subsequent recommendation process, to solve the problems, such as new user's cold start-up.Work as new projects
When into system, new projects' (i.e. new knowledge document) can be judged for which using trained user knowledge demand model
Needed for a little users, and new knowledge is pushed to suitable user, to solve the problems, such as new projects' cold start-up.
In conclusion the above is merely preferred embodiments of the present invention, being not intended to limit the scope of the present invention.
All within the spirits and principles of the present invention, any modification, equivalent replacement, improvement and so on should be included in the present invention's
Within protection domain.
Claims (4)
1. a kind of user knowledge requirement acquisition method calculated based on interactive differential evolution, which is characterized in that including following step
Suddenly：
Step 1: converting knowledge document to vector, structure vector data library；
Step 2: gathering the vector in vector data library for N classes, a vector is randomly selected from every class, obtain N number of population to
Amount, is denoted as { xi(g) }, wherein xi(g) i-th of vectorial, i=1,2 in g generations in population is indicated ..., N；G=0,1,2 ...；
Step 3: the corresponding knowledge document of population vector is pushed to user, setting user satisfaction index is clicking rate, if with
Family is not less than 80% to the clicking rate of the knowledge document pushed, executes step 6, if clicking rate is less than 80%, executes step
Four-step 5；
Step 4: obtaining new N number of vector { u into row variation, intersection respectively to N number of population vector in g generations successivelyi(g+1)}；
Using greedy algorithm from obtained { uiAnd { x (g+1) }i(g) } selection obtains next-generation population at individual in；
Wherein, f (ui(g+1)) individual u is indicatedi(g+1) fitness value；
Step 5: calculating phases of N number of g+1 for population vector and the vector in vector data library successively using cosine similarity algorithm
Like degree, the knowledge document corresponding to most similar N number of vector is chosen, interaction times cumulative 1 execute step 3；
Step 6: being exported current N number of population vector as user knowledge demand model.
2. a kind of user knowledge requirement acquisition method calculated based on interactive differential evolution as described in claim 1, special
Sign is that the knowledge document includes pdf documents, word document and txt documents.
3. a kind of user knowledge requirement acquisition method calculated based on interactive differential evolution as described in claim 1, special
Sign is, in step 1, knowledge document is converted to vector using doc2vec technologies.
4. a kind of user knowledge requirement acquisition method calculated based on interactive differential evolution as described in claim 1, special
Sign is, in step 2, is gathered vector for N classes using K-means methods.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810355129.2A CN108595595B (en) | 2018-04-19 | 2018-04-19 | User knowledge demand acquisition method based on interactive differential evolution calculation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810355129.2A CN108595595B (en) | 2018-04-19 | 2018-04-19 | User knowledge demand acquisition method based on interactive differential evolution calculation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108595595A true CN108595595A (en) | 2018-09-28 |
| CN108595595B CN108595595B (en) | 2020-06-16 |
Family
ID=63613296
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810355129.2A Active CN108595595B (en) | 2018-04-19 | 2018-04-19 | User knowledge demand acquisition method based on interactive differential evolution calculation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108595595B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112948417A (en) * | 2021-01-29 | 2021-06-11 | 上海维外科技有限公司 | Sample knowledge polymerization degree calculation method and system based on selection interaction and storage medium |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103984721A (en) * | 2014-05-13 | 2014-08-13 | 中国矿业大学 | Personalized book searching method based on interactive evolutionary optimization |
| CN106326351A (en) * | 2016-08-05 | 2017-01-11 | 天津大学 | Recommendation system cold start solving method based on user feedback |
| CN106897914A (en) * | 2017-01-25 | 2017-06-27 | 浙江大学 | A kind of Method of Commodity Recommendation and system based on topic model |
| US20170186102A1 (en) * | 2015-12-29 | 2017-06-29 | Linkedin Corporation | Network-based publications using feature engineering |
| CN106940801A (en) * | 2016-01-04 | 2017-07-11 | 中国科学院声学研究所 | A kind of deeply for Wide Area Network learns commending system and method |
| CN107169031A (en) * | 2017-04-17 | 2017-09-15 | 广东工业大学 | It is a kind of to recommend method based on the picture materials that depthmeter reaches |
| CN107256508A (en) * | 2017-05-27 | 2017-10-17 | 上海交通大学 | Commercial product recommending system and its method based on Novel Temporal Scenario |
| CN107391713A (en) * | 2017-07-29 | 2017-11-24 | 内蒙古工业大学 | A kind of method and system for solving the problems, such as cold start-up in collaborative filtering recommending technology |
| CN107563867A (en) * | 2017-09-06 | 2018-01-09 | 天津大学 | A kind of commending system cold start-up method based on multi-arm fruit machine confidence upper limit |
-
2018
- 2018-04-19 CN CN201810355129.2A patent/CN108595595B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103984721A (en) * | 2014-05-13 | 2014-08-13 | 中国矿业大学 | Personalized book searching method based on interactive evolutionary optimization |
| US20170186102A1 (en) * | 2015-12-29 | 2017-06-29 | Linkedin Corporation | Network-based publications using feature engineering |
| CN106940801A (en) * | 2016-01-04 | 2017-07-11 | 中国科学院声学研究所 | A kind of deeply for Wide Area Network learns commending system and method |
| CN106326351A (en) * | 2016-08-05 | 2017-01-11 | 天津大学 | Recommendation system cold start solving method based on user feedback |
| CN106897914A (en) * | 2017-01-25 | 2017-06-27 | 浙江大学 | A kind of Method of Commodity Recommendation and system based on topic model |
| CN107169031A (en) * | 2017-04-17 | 2017-09-15 | 广东工业大学 | It is a kind of to recommend method based on the picture materials that depthmeter reaches |
| CN107256508A (en) * | 2017-05-27 | 2017-10-17 | 上海交通大学 | Commercial product recommending system and its method based on Novel Temporal Scenario |
| CN107391713A (en) * | 2017-07-29 | 2017-11-24 | 内蒙古工业大学 | A kind of method and system for solving the problems, such as cold start-up in collaborative filtering recommending technology |
| CN107563867A (en) * | 2017-09-06 | 2018-01-09 | 天津大学 | A kind of commending system cold start-up method based on multi-arm fruit machine confidence upper limit |
Non-Patent Citations (3)
| Title |
|---|
| 张春美: "《差分进化算法理论与应用》", 31 May 2014 * |
| 邱奇波: "《基于特征表示的推荐算法实践与研究》", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 郝国生: "《交互式进化计算基础理论与应用》", 30 April 2016 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112948417A (en) * | 2021-01-29 | 2021-06-11 | 上海维外科技有限公司 | Sample knowledge polymerization degree calculation method and system based on selection interaction and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108595595B (en) | 2020-06-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liang et al. | Dynamic clustering of streaming short documents | |
| Paliwal et al. | Semantics-based automated service discovery | |
| CN109508414B (en) | Synonym mining method and device | |
| CN110674407B (en) | Hybrid recommendation method based on graph convolution neural network | |
| CN103729359B (en) | A kind of method and system recommending search word | |
| Yang et al. | Venue recommendation: Submitting your paper with style | |
| CN103593425B (en) | Preference-based intelligent retrieval method and system | |
| CN109753602B (en) | Cross-social network user identity recognition method and system based on machine learning | |
| CN112966091B (en) | Knowledge map recommendation system fusing entity information and heat | |
| CN101382939B (en) | Web page text individuation search method based on eyeball tracking | |
| CN106547864A (en) | A kind of Personalized search based on query expansion | |
| Yankov et al. | Evaluation of explore-exploit policies in multi-result ranking systems | |
| CN110083683A (en) | Entity Semantics mask method based on random walk | |
| CN105677828A (en) | User information processing method based on big data | |
| CN110737837A (en) | Scientific research collaborator recommendation method based on multi-dimensional features under research gate platform | |
| CN103095849A (en) | A method and a system of spervised web service finding based on attribution forecast and error correction of quality of service (QoS) | |
| Yang et al. | Design and application of handicraft recommendation system based on improved hybrid algorithm | |
| CN104572915A (en) | User event relevance calculation method based on content environment enhancement | |
| He et al. | UP-TreeRec: Building dynamic user profiles tree for news recommendation | |
| Lu et al. | Recommender system based on scarce information mining | |
| CN108595595A (en) | A kind of user knowledge requirement acquisition method calculated based on interactive differential evolution | |
| Song et al. | QIVISE: a quantum-inspired interactive video search engine in VBS2023 | |
| Zhang et al. | Personalized book recommender system based on Chinese library classification | |
| Jiang et al. | A vertical news recommendation system: CCNS—An example from Chinese campus news reading system | |
| CN104615685A (en) | Hot degree evaluating method for network topic |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |