CN113343086B - Data-driven green design knowledge pushing method - Google Patents
Data-driven green design knowledge pushing method Download PDFInfo
- Publication number
- CN113343086B CN113343086B CN202110607982.0A CN202110607982A CN113343086B CN 113343086 B CN113343086 B CN 113343086B CN 202110607982 A CN202110607982 A CN 202110607982A CN 113343086 B CN113343086 B CN 113343086B
- Authority
- CN
- China
- Prior art keywords
- designer
- knowledge
- designers
- design knowledge
- design
- 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
- 238000013461 design Methods 0.000 title claims abstract description 178
- 238000000034 method Methods 0.000 title claims abstract description 47
- 239000011159 matrix material Substances 0.000 claims abstract description 40
- 238000005065 mining Methods 0.000 claims abstract description 8
- 238000012545 processing Methods 0.000 claims abstract description 4
- 230000006399 behavior Effects 0.000 claims description 46
- 230000009193 crawling Effects 0.000 claims description 25
- 238000001914 filtration Methods 0.000 claims description 16
- 238000004364 calculation method Methods 0.000 claims description 15
- BUGBHKTXTAQXES-UHFFFAOYSA-N Selenium Chemical compound [Se] BUGBHKTXTAQXES-UHFFFAOYSA-N 0.000 claims description 11
- 230000007613 environmental effect Effects 0.000 claims description 11
- 229910052711 selenium Inorganic materials 0.000 claims description 11
- 239000011669 selenium Substances 0.000 claims description 11
- 238000011156 evaluation Methods 0.000 claims description 4
- 238000012935 Averaging Methods 0.000 claims description 3
- 235000014510 cooky Nutrition 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 238000007726 management method Methods 0.000 claims description 3
- 238000012552 review Methods 0.000 claims description 3
- 238000012163 sequencing technique Methods 0.000 claims description 3
- 238000010276 construction Methods 0.000 description 7
- 239000013598 vector Substances 0.000 description 4
- 238000012356 Product development Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
Images
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
-
- 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
- G06F16/951—Indexing; Web crawling techniques
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
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
The invention discloses a data-driven green design knowledge pushing method, and belongs to the technical field of knowledge pushing. The invention relates to a data-driven green design knowledge pushing method which comprises the steps of mining green design knowledge, constructing a design knowledge set characteristic word weight matrix, constructing a designer dominant model, constructing a designer-knowledge behavior matrix, constructing a designer implicit model, constructing a designer demand model and pushing the green design knowledge to carry out various business processing; the problem that the recommended information is inaccurate due to lack of diversity and individuation in the knowledge recommendation process in the existing system is solved.
Description
Technical Field
The invention relates to the technical field of knowledge pushing, in particular to a data-driven green design knowledge pushing method.
Background
Enterprises are under various industrial pressures in the development process, and the product quality needs to be continuously improved under the constraint of laws and regulations, so that the continuously changing customer requirements are met, and the response time to the rapidly changing market is shortened. Developing new products and upgrading existing products are knowledge-intensive processes, necessary knowledge needs to be applied, shared and pushed, and the knowledge reuse efficiency in the product development process is improved. The novel green product meets the requirements of the era, attracts customers to purchase, and needs designers to take an important task of green design. Under the background of big data, how to establish effective contact between designers and green design knowledge enables the designers to effectively acquire interesting and required design knowledge, data-driven green design knowledge pushing facing to the early product design stage is realized, and the efficiency of the designers for carrying out green design on products by using knowledge is significant.
In recent years, the academic and industrial circles have been concerned with knowledge push systems for new product development and services to support developers to acquire resources more conveniently and quickly, implement parallel engineering and collaboration techniques, and shorten the lead time for new product development. Collaborative filtering is the most widely researched and used push technology, however, the traditional collaborative filtering method is difficult to reflect the difference of interest and preference of target designers and similar design people, and the recommendation information is inaccurate due to the lack of diversity and personalization in the knowledge recommendation process.
Disclosure of Invention
The invention provides a data-driven green design knowledge pushing method, which is used for carrying out various business processes by using green design knowledge mining, design knowledge set characteristic word weight matrix construction, designer explicit model construction, designer-knowledge behavior matrix construction, designer implicit model construction, designer demand model construction and green design knowledge pushing.
The existing system has the problem that the recommended information is inaccurate due to lack of diversity and individuation in the knowledge recommendation process.
In order to achieve the purpose, the invention adopts the following technical scheme:
a data-driven green design knowledge pushing method comprises the following steps:
step 1, green design is to treat the resource consumption and environmental influence problems caused by the product while considering the function, quality and cost of the product in the life cycle of the product, so that the product can not only ensure the use function of the product, but also consider that various environmental indexes meet the green design requirements and the negative total influence on the environment is minimized, fully consider the environmental attributes of the product, reduce the environmental emission of the product to the minimum as much as possible, and provide data support for knowledge push by mining green design knowledge and crawl from web pages;
the mining of green design knowledge mainly comprises crawling of thesis information, using a Selenium frame, inputting corresponding keyword entries in a program through simulating operation behaviors of a browser, performing fuzzy matching, combining Google browser driving, simulating a designer to click a website, analyzing obtained webpage contents, positioning key information, and then crawling webpage documents, wherein main contents crawled by the thesis list page comprise information of obtaining titles, links, publication time, authors and the like. Finally, the acquired information related to the literature is stored in a CSV format, so that designers can conveniently and quickly acquire various related thesis information in the research field, and the time consumed for searching the literature is reduced;
the document crawling process is shown in fig. 2, the article crawling is mainly divided into two parts, firstly, Cookies information is obtained through a Selenium frame, then, a page table is crawled rapidly, and after crawling is completed, the article information is stored in a CSV format file. And secondly, judging whether the entry is queried or not, using a Selenium frame simulator to perform page turning browsing, and finishing searching according to judgment conditions when the content cannot be searched according to fuzzy matching of the entry.
The paper crawling specifically comprises the following steps:
1) a keyword is input. Inputting related keyword entries needing to be captured in a program, such as a front-end module, and then starting a compiler to start working;
2) and waiting for data crawling. After the crawler is started, waiting for a program to connect with a knowledge network, and displaying a prompt box to start crawling the thesis, wherein the crawling progress of the document and the information condition of the related thesis can be checked;
3) and (6) downloading data. Reading related information such as title, author, abstract, source, citation, link and the like of each document, and storing the related information into a set CSV format file;
4) and (6) ending. And according to the judgment condition, when the documents are not matched, ending the program.
Step 2, giving design knowledge set Ks ═ Ks 1 ,ks 2 ,...ks n In order to effectively represent the design knowledge set, design knowledge features and outlines will be describedDefining the vocabulary including the design knowledge content as the characteristic words of Ks, and expressing the main characteristic word sequence as F ═ (F) 1 ,f 2 ,...,f j ,...f k ) Wherein f is j Representing design knowledge ks n J feature words are represented, k represents the number of the feature words, the design knowledge set mainly judges the interest preference of designers according to historical browsing records of the designers so as to push the needed contents to the designers, and the feature information of the designers reflects the similarity relation among the same designers to a great extent because the designers contain the feature information of the designers. Meanwhile, the characteristic attribute of the designer is stable and can be expressed as different characteristics among the designers;
by introducing the characteristic attributes of the designers, the characteristic attribute similarity of similar designers can be extracted, and the weights of irrelevant items of target items in similar items can be changed, so that a neighbor designer set can be calculated. The relationship between the designer and the design knowledge is represented by the following equation:
D=d 1 ∪d 2 ∪…∪d i (1)
d 1 ={ks 11 ,ks 12 ,...ks 1n } (2)
d i ={ks i1 ,ks i2 ,...ks in } (3)
wherein D represents the set of designers, ks in Indicating the nth design knowledge owned by the rear designer i.
For a given design knowledge set Ks ═ Ks 1 ,ks 2 ,...ks n And the characteristic word f j Text vectorization, feature word matrix sequence and design knowledge set ks i Corresponding to each other, ks i Can be expressed as: ks is the product of i =(w i1 ,w i2 ,...,w ij ,...,w ik ) Wherein w is ij Representation feature word f j Weight in ks, w ij Is in the range of [1,10 ]](ii) a When w is ij When 0, the feature word f is explained j Not in the design knowledge set ks; the mathematical formula for designing the weight matrix of the knowledge set feature words is as follows:
the constructed weight matrix of the characteristic words of the design knowledge set is the evaluation condition of the designer on the browsing knowledge for the first time, namely the dominant knowledge weight of the designer.
Designer explicit model definition: vectorizing a design knowledge text browsed by a designer, constructing a vectorized weight matrix, and finally forming an explicit interest model of the target designer; can be given by the formula DDM ═ (w 1) 1 ,w1 2 ,...,w1 j ,...,w1 k ) Is shown, w1 j Representing the weight of the sequence of feature words in the explicit model.
Preferably, the knowledge behavior matrix construction adopts a collaborative filtering method for mixing designer behaviors and design knowledge contents, so that the similarity of the designer can be expanded into two parts, namely behavior similarity and content similarity; and setting a corresponding weight lambda by using a linear combination method, and combining the behavior similarity between the designers and the content similarity between the designers to serve as a similarity calculation method between the two designers. The formula is as follows:
sim(d u ,d v )=(1-λ)sim b (d u ,d v )+λsim c (d u ,d v ) (5)
in the formula, lambda is a similarity parameter, and the value range is [0, 1 ]; when the value of lambda is (0, 1), the similarity between designers comprehensively considers the similarity of the content of the browsing design knowledge and the behavior of the browsing design knowledge;
designer set D ═ D 1 ,d 2 ,...,d i ,…,d k Denotes the ensemble set, where k denotes the total number of designers, and the knowledge set Ks ═ Ks 1 ,ks 2 ,…ks n And according to the record of the reading knowledge of the designers, the knowledge can be expressed as a designer knowledge behavior matrix, wherein the row of the matrix represents each designer, and the column of the matrix represents each designerEach item of design knowledge, the number 1 in the matrix indicates that the designer has read a certain knowledge ks n The designer d is represented by the number 0 i Not reading a certain knowledge ks n The designer-knowledge behavior matrix constructed is shown in table 1:
TABLE 1 designer-knowledge behavior matrix
Design knowledge ks 1 | Design knowledge ks 2 | ... | Design knowledge ks n | |
Designer d 1 | 0 | 1 | ... | 1 |
Designer d 2 | 1 | 0 | ... | 0 |
... | ... | ... | ... | 1 |
Designer d i | 0 | 0 | ... | 0 |
Designer implicit model definition: pushing the interest obtained from the dominant model of the similar designer to the target designer by using a collaborative filtering mechanism, and further obtaining a recessive interest model of the target designer through weighting processing;
can be given by the formula DIM ═ (w 2) 1 ,w2 2 ,…,w2 j ,…,w2 k ) Is shown, w2 j Representing the weight of the characteristic word sequence in the recessive model;
the designer's implicit interest model is different from the designer's explicit interest model, and cannot be retrieved through previous designer reviews or history. The collaborative filtering can push the interests of similar designers to target designers to initialize the implicit interests of the target designers;
and after the designer-knowledge behavior matrix is generated, calculating the content similarity of the target designer and all other designers by adopting the Jaccard similarity.
Definition 1: content similarity sim between designers c (d u ,d v )
Designer d u And d v The following formula:
because the reading knowledge sets of designers are more, in order to more accurately obtain a recessive interest model of a target designer, the content similarity between the designers is calculated by adopting an averaging method, namely, the design knowledge recently browsed by the similar designers is averaged according to the weight of the feature words, and the designer d u And d v The content similarity calculation formula of (2) is as follows:
definition 2: similarity of behavior sim between designers b (d u ,d v )
In the invention, the preference condition of the designer to the design knowledge is obtained according to the behavior record of the designer to the design knowledge, which is called the behavior similarity between the designers, and the target designer d t And other designers d i Is used as the similarity of behaviors of b (d t ,d i ) To represent;
in order to reflect the difference between hot knowledge and cold knowledge browsed by different designers, the invention adopts improved cosine similarity to calculate the similarity of the behavior of the designers:
wherein D (ks) i ) Indicating that the same knowledge set ks has been browsed i The number of the (c) component(s),andrepresenting the amount of design knowledge that designers read individually;
from the designer mixed similarity calculation equation (5), assume the designer's d u Group of similar designers is D u ={d v1 ,d v2 ,…,d vk }, designer d u With any designer d in the group of similar designers vi Has a mixed similarity of sim (d) u ,d vi ) Similarity designer d vi The dominant interest model is DDM vi =(w1 vi1 ,w1 vi2 ,…,w1 vij ,…,w1 vik ) Based on the above conditions, the designer d is obtained u Weight of feature words of implicit interest model:
the implicit interest model of the designer is obtained through calculation based on the dominant interest model of the designer, the weights calculated through the improved collaborative filtering method are ranked, and then the part of design knowledge with the highest weight in the interest model of the similar designer is pushed to the target designer to initialize the implicit interest of the target designer, and finally the diversity design knowledge is pushed to the designer.
Preferably, the designer requirement model is a final requirement model of the target designer obtained by combining the explicit interest model and the implicit interest model of the designer according to a certain rule. Can use formula DRM ═ (w 3) 1 ,w3 2 ,…,w3 j ,…,w3 k ) Is shown, w3 j Representing weights of sequences of feature words in a demand model;
After the dominant model and the recessive model of the target designer are obtained, in order to more accurately push the design knowledge of the self requirement to the designer, the invention selects the first K items of design knowledge with the front scoring values in the two models according to the design knowledge in the dominant model and the recessive model obtained by the designer, thereby not only meeting the requirement of the personalized knowledge of the designer, but also ensuring the diversity of the design knowledge of the designer.
According to the above theory, designer d u The dominant interest model of (c) is: DDM u =(w1 u1 ,w1 u2 ,…,w1 uj ,…,w1 uk ) Designer d u The implicit interest model of (c) is: DIM u =(w2 u1 ,w2 u2 ,...,w2 uj ,...,w2 uk ) Setting up a designer d u The demand model of (1) is DRM u =(w3 u1 ,w3 u2 ,...,w3 uj ,...,w3 uk ) The design knowledge set is Ks ═ Ks 1 ,ks 2 ,...ks n }. At designer d u In the demand model, according to DDM u And DIM u And (3) evaluating the design knowledge of the top-K item with higher degree, calculating to obtain the weight of the design knowledge in the actual requirement model of the designer, wherein the calculation formula is as follows:
w3 ui =max top-K w1 ui (11)
w3 uj =max top-K w2 uj (12)
w3 u ={w3 ui ,w3 uj } (13)
in formula (13), w3 u Represents selection w1 ui And w2 uj The first K items of design knowledge are pushed to the designer.
Preferably, the green design knowledge pushing is an active knowledge management method, mainly reflects in the interactivity and real-time performance with designers, and pushes knowledge related to design tasks to designers when designing products. The method specifically comprises the following steps:
1) the pushing process of the green product design knowledge is divided into two parts, namely a designer and a pushing system, and product design is carried out according to the designer end and the pushing system end.
2) At the designer's end, the designer receives a task list starting from the design requirements. After the designer allocates the design task, the design knowledge requirements which may be used in the task are matched with the knowledge resources. After finding the similarity, the designer can obtain the matched knowledge, and can improve the related design knowledge and finally store the related design knowledge in the knowledge base.
3) At a pushing system end, matching knowledge requirements of designers with knowledge resources, wherein the matching comprises feature extraction, knowledge matching, sequencing, knowledge storage and data updating of the designers;
the green design knowledge pushing flow is shown in fig. 3.
Compared with the prior art, the invention provides a data-driven green design knowledge pushing method, which has the following beneficial effects:
(1) from the perspective of accurately and quickly pushing green design knowledge to designers, in order to achieve complete pushing of the green design knowledge, an automated test framework of Python and Selenium browsers is used, Google browsers and corresponding browser drivers are installed to combine websites and crawler technologies, automation of document capturing is achieved, weighted feature vectors are used as representation models to determine how each feature contributes to classification, feature words are selected to represent importance of knowledge vectors, a recessive model, a dominant model and a demand model of the designers are built in combination with content pushing and collaborative filtering pushing, and similarity comparison is conducted on mined related knowledge and a designer demand model based on data driving, and personalized pushing of the green design knowledge is achieved.
Drawings
FIG. 1 is a schematic diagram of a data-driven green design knowledge push model framework proposed by the present invention;
FIG. 2 is a flow diagram of the crawling of green design knowledge proposed by the present invention;
fig. 3 is a flow chart of pushing green design knowledge proposed by the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments.
Example 1:
a data-driven green design knowledge pushing method comprises the following steps:
step 1, green design is to treat the resource consumption and environmental influence problems caused by the product while considering the function, quality and cost of the product in the life cycle of the product, so that the product can not only ensure the use function of the product, but also consider that various environmental indexes meet the green design requirements and the negative total influence on the environment is minimized, fully consider the environmental attributes of the product, reduce the environmental emission of the product to the minimum as much as possible, and provide data support for knowledge push by mining green design knowledge and crawl from web pages;
the mining of green design knowledge mainly comprises crawling of thesis information, using a Selenium frame, inputting corresponding keyword entries in a program through simulating operation behaviors of a browser, performing fuzzy matching, combining Google browser driving, simulating a designer to click a website, analyzing obtained webpage contents, positioning key information, and then crawling webpage documents, wherein main contents crawled by the thesis list page comprise information of obtaining titles, links, publication time, authors and the like. Finally, the acquired information related to the literature is stored in a CSV format, so that designers can conveniently and quickly acquire various related thesis information in the research field, and the time consumed for searching the literature is reduced;
the document crawling process is shown in fig. 2, the article crawling is mainly divided into two parts, firstly, Cookies information is obtained through a Selenium frame, then, a page table is crawled rapidly, and after crawling is completed, the article information is stored in a CSV format file. And secondly, judging whether the entry is queried or not, using a Selenium frame simulator to perform page turning browsing, and finishing searching according to judgment conditions when the content cannot be searched according to fuzzy matching of the entry.
The paper crawling specifically comprises the following steps:
1) a keyword is input. Inputting related keyword entries needing to be captured in a program, such as a front-end module, and then starting a compiler to start working;
2) and waiting for data crawling. After the crawler is started, waiting for a program to connect with a knowledge network, and displaying a prompt box to start crawling the thesis, wherein the crawling progress of the document and the information condition of the related thesis can be checked;
3) and (6) downloading data. Reading related information such as title, author, abstract, source, citation, link and the like of each document, and storing the related information into a set CSV format file;
4) and (6) ending. And according to the judgment condition, when the documents are not matched, ending the program.
Step 2, giving design knowledge set Ks ═ { Ks ═ Ks 1 ,ks 2 ,...ks n In order to effectively represent the design knowledge set, words which can describe the characteristics of the design knowledge and summarize the content of the design knowledge are defined as characteristic words of Ks, and the main characteristic word sequence is represented as F ═ F (F ═ F) 1 ,f 2 ,...,f j ,...f k ) Wherein f is j Representing design knowledge ks n J feature words are represented, k represents the number of the feature words, the design knowledge set mainly judges the interest preference of designers according to historical browsing records of the designers so as to push the needed contents to the designers, and the feature information of the designers reflects the similarity relation among the same designers to a great extent because the designers contain the feature information of the designers. Meanwhile, the characteristic attribute of the designer is stable and can be expressed as different characteristics among the designers;
by introducing the characteristic attributes of the designers, the characteristic attribute similarity of similar designers can be extracted, and the weights of irrelevant items of target items in similar items can be changed, so that a neighbor designer set can be calculated. The relationship between the designer and the design knowledge is represented by the following equation:
D=d 1 ∪d 2 ∪…∪d i (1)
d 1 ={ks 11 ,ks 12 ,...ks 1n } (2)
d i ={ks i1 ,ks i2 ,...ks in } (3)
wherein D represents the set of designers, ks in Indicating the nth design knowledge owned by the post designer i.
For a given design knowledge set Ks ═ Ks 1 ,ks 2 ,...ks n And the characteristic word f j Text vectorization, feature word matrix sequence and design knowledge set ks i Corresponding to each other, ks i Can be expressed as: ks is the product of i =(w i1 ,w i2 ,...,w ij ,...,w ik ) Wherein w is ij Representation feature word f j Weight in ks, w ij Is in the range of [1,10 ]](ii) a When w is ij When 0, the feature word f is explained j Not in the design knowledge set ks; the mathematical formula for designing the weight matrix of the knowledge set feature words is as follows:
the constructed weight matrix of the characteristic words of the design knowledge set is the evaluation condition of the designer on the browsing knowledge for the first time, namely the dominant knowledge weight of the designer.
Designer explicit model definition: vectorizing a design knowledge text browsed by a designer, constructing a vectorized weight matrix, and finally forming an explicit interest model of the target designer; can be given by the formula DDM ═ (w 1) 1 ,w1 2 ,...,w1 j ,...,w1 k ) Is shown, w1 j Representing the weight of the sequence of feature words in the explicit model.
The knowledge behavior matrix construction adopts a collaborative filtering method for mixing designer behaviors and design knowledge content, and can expand the similarity of designers into two parts, namely behavior similarity and content similarity; and setting a corresponding weight lambda by using a linear combination method, and combining the behavior similarity between the designers and the content similarity between the designers to serve as a similarity calculation method between the two designers. The formula is as follows:
sim(d u ,d v )=(1-λ)sim b (d u ,d v )+λsim c (d u ,d v ) (5)
in the formula, lambda is a similarity parameter, and the value range is [0, 1 ]; when the value of lambda is (0, 1), the similarity between designers comprehensively considers the similarity of the content of the browsing design knowledge and the behavior of the browsing design knowledge;
designer set D ═ D 1 ,d 2 ,...,d i ,...,d k Denotes the ensemble set, where k denotes the total number of designers, and the knowledge set Ks ═ Ks 1 ,ks 2 ,...ks n And according to the record of the reading knowledge of the designer, the knowledge can be expressed as a designer knowledge behavior matrix, wherein the row of the matrix represents each designer, the column of the matrix represents each design knowledge, and the number 1 in the matrix represents that the designer reads a certain knowledge ks n The designer d is represented by the number 0 i Not reading a certain knowledge ks n The designer-knowledge behavior matrix constructed is shown in table 1:
TABLE 1 designer-knowledge behavior matrix
Design knowledge ks 1 | Design knowledge ks 2 | ... | Design knowledge ks n | |
Designer d 1 | 0 | 1 | ... | 1 |
Designer d 2 | 1 | 0 | ... | 0 |
... | ... | ... | ... | 1 |
Designer d i | 0 | 0 | ... | 0 |
Designer implicit model definition: pushing the interest obtained from the dominant model of the similar designer to the target designer by using a collaborative filtering mechanism, and further obtaining a recessive interest model of the target designer through weighting processing;
can be given by the formula DIM ═ (w 2) 1 ,w2 2 ,...,w2 j ,...,w2 k ) Is shown, w2 j Representing the weight of the characteristic word sequence in the recessive model;
the designer's implicit interest model is different from the designer's explicit interest model, and cannot be retrieved through previous designer reviews or history. The collaborative filtering can push the interests of similar designers to target designers to initialize the implicit interests of the target designers;
and after the designer-knowledge behavior matrix is generated, calculating the content similarity of the target designer and all other designers by adopting the Jaccard similarity.
Definition 1: content similarity sim between designers c (d u ,d v )
Designer d u And d v The following formula:
because the reading knowledge sets of designers are more, in order to more accurately obtain the implicit interest model of the target designer, the content similarity between the designers is calculated by adopting an averaging method, namely, the design knowledge recently browsed by the similar designers is averaged according to the weight of the feature words, and the designer d u And d v The content similarity calculation formula of (2) is as follows:
definition 2: similarity of behavior sim between designers b (d u ,d v )
In the invention, the preference condition of the designer to the design knowledge is obtained according to the behavior record of the designer to the design knowledge, which is called the behavior similarity between the designers, and the target designer d t And other designers d i Is used as the similarity of behaviors of b (d t ,d i ) To represent;
in order to reflect the difference between hot knowledge and cold knowledge browsed by different designers, the invention adopts improved cosine similarity to calculate the similarity of the behavior of the designers:
wherein D (ks) i ) Indicating that the same knowledge set ks has been browsed i The number of the (c) component(s),andrepresenting the amount of design knowledge that designers read individually;
from the designer mixed similarity calculation equation (5), assume the designer's d u Group of similar designers is D u ={d v1 ,d v2 ,…,d vk }, designer d u With any designer d in the group of similar designers vi Has a mixed similarity of sim (d) u ,d vi ) Similarity designer d vi The dominant interest model is DDM vi =(w1 vi1 ,w1 vi2 ,...,w1 vij ,...,w1 vik ) Based on the above conditions, the designer d is obtained u Weight of feature words of implicit interest model:
the implicit interest model of the designer is obtained through calculation based on the dominant interest model of the designer, the weights calculated through the improved collaborative filtering method are ranked, and then the part of design knowledge with the highest weight in the interest model of the similar designer is pushed to the target designer to initialize the implicit interest of the target designer, and finally the diversity design knowledge is pushed to the designer.
The designer demand model is a final demand model of a target designer obtained by combining the explicit interest model and the implicit interest model of the designer according to a certain rule. Can use formula DRM ═ (w 3) 1 ,w3 2 ,...,w3 j ,...,w3 k ) Is shown, w3 j Representing the weight of the characteristic word sequence in the demand model;
after the dominant model and the recessive model of the target designer are obtained, in order to more accurately push the design knowledge of the self requirement to the designer, the invention selects the first K items of design knowledge with the front scoring values in the two models according to the design knowledge in the dominant model and the recessive model obtained by the designer, thereby not only meeting the requirement of the personalized knowledge of the designer, but also ensuring the diversity of the design knowledge of the designer.
According to the above theory, designer d u The dominant interest model of (c) is: DDM u =(w1 u1 ,w1 u2 ,...,w1 uj ,...,w1 uk ) Designer d u The implicit interest model of (c) is: DIM u =(w2 u1 ,w2 u2 ,...,w2 uj ,...,w2 uk ) Setting up the designer d u The demand model of (1) is DRM u =(w3 u1 ,w3 u2 ,...,w3 uj ,...,w3 uk ) The design knowledge set is Ks ═ Ks 1 ,ks 2 ,...ks n }. At designer d u In the demand model, according to DDM u And DIM u And (3) evaluating the design knowledge of the top-K item with higher evaluation, and calculating to obtain the weight of the design knowledge in the actual requirement model of the designer, wherein the calculation formula is as follows:
w3 ui =max top-K w1 ui (11)
w3 uj =max top-K w2 uj (12)
w3 u ={w3 ui ,w3 uj } (13)
in formula (13), w3 u Represents selection w1 ui And w2 uj The first K items of design knowledge are pushed to the designer.
The green design knowledge pushing is an active knowledge management method, mainly reflects in the interactivity and real-time performance between designers, and pushes the knowledge related to design tasks to the designers when the products are designed. The method specifically comprises the following steps:
1) the pushing process of the green product design knowledge is divided into two parts, namely a designer and a pushing system, and product design is carried out according to the designer end and the pushing system end.
2) At the designer's end, the designer receives a task list starting from the design requirements. After the designer allocates the design task, the design knowledge requirements which may be used in the task are matched with the knowledge resources. After finding the similarity, the designer can obtain the matched knowledge, and can improve the related design knowledge and finally store the related design knowledge in the knowledge base.
3) At a pushing system end, matching knowledge requirements of designers with knowledge resources, wherein the matching comprises feature extraction, knowledge matching, sequencing, knowledge storage and data updating of the designers;
the green design knowledge pushing flow is shown in fig. 3.
From the perspective of accurately and quickly pushing green design knowledge to designers, in order to achieve complete pushing of the green design knowledge, an automated test framework of Python and Selenium browsers is used, Google browsers and corresponding browser drivers are installed to combine websites and crawler technologies, automation of document capturing is achieved, weighted feature vectors are used as representation models to determine how each feature contributes to classification, feature words are selected to represent importance of knowledge vectors, a recessive model, a dominant model and a demand model of the designers are built in combination with content pushing and collaborative filtering pushing, and similarity comparison is conducted on mined related knowledge and a designer demand model based on data driving, and personalized pushing of the green design knowledge is achieved.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered as the technical solutions and the inventive concepts of the present invention within the technical scope of the present invention.
Claims (1)
1. A data-driven green design knowledge pushing method is characterized by comprising the following steps: the method comprises the following steps:
step 1, mining green design knowledge;
the method specifically comprises the following steps: the green design is that in the life cycle of the product, the problems of resource consumption and environmental influence caused by the product are processed while the function, quality and cost of the product are considered, so that the product not only can ensure the use function of the product, but also more environment indexes are considered to meet the green design requirement and minimize the total negative influence on the environment, the environmental attribute of the product is fully considered, the environmental emission of the product is reduced to the minimum as far as possible, the mining of green design knowledge provides data support for knowledge pushing, and the data support is crawled from a webpage;
the paper crawling specifically comprises the following steps:
1) inputting a keyword; inputting related keyword entries needing to be captured in a program, and then starting a compiler to start working;
2) waiting for data crawling; after the crawler is started, a program is waited for connection, a prompt box displays that a thesis starts to be crawled, and at the moment, the crawling progress of the document and the information condition of the related thesis can be checked;
obtaining Cookies information by using a Selenium frame, then quickly crawling a page table, and storing thesis information into a CSV format file after crawling is completed;
using a Selenium frame, simulating the operation behavior of a browser, inputting corresponding keyword entries in a program, performing fuzzy matching, combining with Google browser drive, simulating a designer to click a website, analyzing the acquired webpage content, positioning key information, and then crawling webpage documents, wherein the main content crawled by a paper list page comprises acquisition titles, links, publication time and author information; finally, storing the acquired information related to the literature in a CSV format; 3) downloading data; reading the title, author, abstract, source, citation and link related information of each document, and storing the information into a set CSV format file;
4) finishing; according to the judgment condition, when the document is not matched, ending the program;
judging whether the query of the vocabulary entry is finished, using a Selenium frame simulator to perform page turning browsing, and finishing the search according to the judgment condition when the content cannot be searched according to the fuzzy matching of the vocabulary entry;
step 2, constructing an explicit model of a designer;
the method specifically comprises the following steps: given a design knowledge set Ks ═ Ks 1 ,ks 2 ,...ks n In order to effectively represent the design knowledge set, words which can describe the characteristics of the design knowledge and summarize the content of the design knowledge are defined as characteristic words of Ks, and the main characteristic word sequence is represented as F ═ F (F ═ F) 1 ,f 2 ,...,f j ,...f k ) Wherein f is j Representing design knowledge ks n J feature words, k represents the number of the feature words, the design knowledge set mainly judges the interest preference of designers according to the historical browsing records of the designers so as to push the needed contents to the designers, and the feature information of the designers reflects the similarity relation among the designers as the designers contain the feature information of the designers; meanwhile, the characteristic attribute of the designer is stable and can be expressed as different characteristics among the designers;
by introducing the characteristic attributes of the designers, the characteristic attribute similarity among similar designers can be extracted, and the weights of irrelevant items of a target item in similar items can be changed, so that a neighbor designer set can be calculated; the relationship between the designer and the design knowledge is represented by the following equation:
D=d 1 ∪d 2 ∪…∪d i (1)
d 1 ={ks 11 ,ks 12 ,...ks 1n } (2)
d i ={ks i1 ,ks i2 ,...ks in } (3)
wherein D represents the set of designers, ks in Represents the nth design knowledge owned by the designer i;
for a given design knowledge set Ks ═ Ks 1 ,ks 2 ,...ks n And the characteristic word f j Text vectorization, feature word matrix sequence and design knowledge set ks i Corresponding to each other, ks i Can be expressed as: ks is the product of i =(w i1 ,w i2 ,...,w ij ,...,w ik ) Wherein w is ij Representation feature word f j Weight in ks, w ij Is in the range of [1,10 ]](ii) a When w is ij When 0, the feature word f is explained j Not in the design knowledge set ks; the mathematical formula for designing the weight matrix of the knowledge set feature words is as follows:
the constructed weight matrix of the characteristic words of the design knowledge set is the evaluation condition of the designer on the browsing knowledge for the first time, namely the dominant knowledge weight of the designer;
designer explicit model definition: vectorizing a design knowledge text browsed by a designer, constructing a vectorized weight matrix, and finally forming an explicit interest model of the target designer; can be given by the formula DDM ═ (w 1) 1 ,w1 2 ,...,w1 j ,...,w1 k ) Is shown, w1 j Indicating that a sequence of feature words is in dominant modeThe weight in the form;
step 3, constructing a designer-knowledge behavior matrix;
the method specifically comprises the following steps: designer set D ═ D 1 ,d 2 ,...,d i ,...,d k Denotes the ensemble set, where k denotes the total number of designers, and the knowledge set Ks ═ Ks 1 ,ks 2 ,...ks n And expressing the read knowledge record of the designer as a designer knowledge behavior matrix according to the read knowledge record of the designer, wherein the row of the matrix represents each designer, the column of the matrix represents each design knowledge, and the number 1 in the matrix represents that the designer reads a certain knowledge ks n The designer d is represented by the number 0 i Not reading a certain knowledge ks n The designer-knowledge behavior matrix constructed is as follows:
step 4, constructing a hidden model of a designer;
the method specifically comprises the following steps: designer implicit model definition: pushing the interest obtained from the dominant model of the similar designer to the target designer by using a collaborative filtering mechanism, and further obtaining a recessive interest model of the target designer through weighting processing;
can be given by the formula DIM ═ (w 2) 1 ,w2 2 ,...,w2 j ,...,w2 k ) Is shown, w2 j Representing the weight of the characteristic word sequence in the recessive model;
the designer's implicit interest model is different from the designer's explicit interest model, and cannot be retrieved through previous designer reviews or history; the collaborative filtering can push the interests of similar designers to target designers to initialize the implicit interests of the target designers;
the method comprises the steps that a collaborative filtering method for mixing designer behaviors and design knowledge contents is adopted to construct a designer-knowledge behavior matrix, and the similarity of the designer can be expanded into two parts, namely behavior similarity and content similarity; setting a corresponding weight lambda by using a linear combination method, and combining the behavior similarity between the designers and the content similarity between the designers to serve as a similarity calculation method between the two designers; the formula is as follows:
sim(d u ,d v )=(1-λ)sim b (d u ,d v )+λsim c (d u ,d v ) (5)
in the formula, lambda is a similarity parameter, and the value range is [0, 1 ]; when the value of lambda is (0, 1), the similarity between designers comprehensively considers the similarity of the content of the browsing design knowledge and the behavior of the browsing design knowledge;
after a designer-knowledge behavior matrix is generated, calculating the content similarity of a target designer and all other designers by adopting the Jaccard similarity;
definition 1: content similarity sim between designers c (d u ,d v )
Designer d u And d v The following formula:
because the reading knowledge sets of designers are more, in order to more accurately obtain the implicit interest model of the target designer, the content similarity between the designers is calculated by adopting an averaging method, namely, the design knowledge recently browsed by the similar designers is averaged according to the weight of the feature words, and the average value is setCounter d u And d v The content similarity calculation formula of (2) is as follows:
definition 2: similarity of behavior sim between designers b (d u ,d v )
Obtaining the preference of the designer to the design knowledge according to the behavior record of the designer to the design knowledge, called the behavior similarity between the designers, and the target designer d t And other designers d i Is used as the similarity of behaviors of b (d t ,d i ) To represent;
in order to reflect the difference between the hot knowledge and the cold knowledge of browsing among different designers, the behavior similarity between the designers is calculated by adopting the improved cosine similarity:
wherein D (ks) i ) Indicating that the same knowledge set ks has been browsed i The number of the (c) component(s),andrepresenting the amount of design knowledge that designers read individually;
according toDesigner hybrid similarity calculation equation (5), designer d u The group of similar designers is D u ={d v1 ,d v2 ,…,d vk }, designer d u With any designer d in the group of similar designers vi Has a mixed similarity of sim (d) u ,d vi ) Similarity designer d vi The dominant interest model is DDM vi =(w1 vi1 ,w1 vi2 ,…,w1 vij ,…,w1 vik ) From the above equations (1) - (9), the designer d is derived u Weight of feature words of implicit interest model:
the implicit interest models of the designers are obtained by calculation based on the explicit interest models of the designers, the weights calculated by the improved collaborative filtering method are sorted, and then the part of design knowledge with the highest weight in the interest models of the similar designers is pushed to the target designers to initialize the implicit interest of the target designers, so that the diversity design knowledge is finally pushed to the designers;
step 5, constructing a designer requirement model;
the method specifically comprises the following steps: the designer demand model is a final demand model of a target designer, which is obtained by combining an explicit interest model and a implicit interest model of the designer according to a certain rule; can use formula DRM ═ (w 3) 1 ,w3 2 ,…,w3 j ,…,w3 k ) Is shown, w3 j Representing the weight of the characteristic word sequence in the demand model;
after acquiring the dominant model and the recessive model of the target designer, in order to more accurately push the design knowledge of the self-demand to the designer, according to the design knowledge in the dominant model and the recessive model acquired by the designer, the front K items of design knowledge with the front scoring values in the two models are selected, so that the requirement of the self-personalized knowledge of the designer is met, and the diversity of the design knowledge of the designer can be ensured;
designer d u The dominant interest model of (c) is: DDM u =(w1 u1 ,w1 u2 ,…,w1 uj ,…,w1 uk ) Designer d u The implicit interest model of (c) is: DIM u =(w2 u1 ,w2 u2 ,…,w2 uj ,...,w2 uk ) Setting up a designer d u The demand model of (1) is DRM u =(w3 u1 ,w3 u2 ,...,w3 uj ,...,w3 uk ) The design knowledge set is Ks ═ Ks 1 ,ks 2 ,…ks n }; at designer d u In the demand model, according to DDM u And DIM u And (3) evaluating the design knowledge of the top-K item with higher degree, calculating to obtain the weight of the design knowledge in the actual requirement model of the designer, wherein the calculation formula is as follows:
w3 ui =max top-K w1 ui (11)
w3 uj =max top-K w2 uj (12)
w3 u ={w3 ui ,w3 uj } (13)
in formula (13), w3 u Represents selection w1 ui And w2 uj Pushing the first K items of design knowledge to a designer;
step 6, pushing green design knowledge;
the method specifically comprises the following steps: the green design knowledge pushing is an active knowledge management method, is mainly reflected in interactivity and real-time performance with designers, and pushes knowledge related to design tasks to the designers when the products are designed; the method specifically comprises the following steps:
1) dividing the pushing process of the green product design knowledge into two parts, namely a designer and a pushing system, and designing the product according to the designer end and the pushing system end;
2) at a designer end, starting from a design requirement, the designer receives a task list; after the design task is distributed, the designer matches the design knowledge requirements used in the task with knowledge resources; after the similarity is found, the designer can obtain the matched knowledge, can improve the related design knowledge and finally store the improved design knowledge in a knowledge base;
3) at the pushing system end, knowledge requirements of designers are matched with knowledge resources, and the knowledge requirements comprise feature extraction, knowledge matching, sequencing, knowledge storage and data updating of the designers.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110607982.0A CN113343086B (en) | 2021-06-01 | 2021-06-01 | Data-driven green design knowledge pushing method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110607982.0A CN113343086B (en) | 2021-06-01 | 2021-06-01 | Data-driven green design knowledge pushing method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113343086A CN113343086A (en) | 2021-09-03 |
CN113343086B true CN113343086B (en) | 2022-09-16 |
Family
ID=77474080
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110607982.0A Active CN113343086B (en) | 2021-06-01 | 2021-06-01 | Data-driven green design knowledge pushing method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113343086B (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103593336A (en) * | 2013-10-30 | 2014-02-19 | 中国运载火箭技术研究院 | Knowledge pushing system and method based on semantic analysis |
CN104899242A (en) * | 2015-03-10 | 2015-09-09 | 四川大学 | Mechanical product design two-dimensional knowledge pushing method based on design intent |
DE102018104824A1 (en) * | 2017-03-13 | 2018-09-13 | General Motors Llc | SYSTEMS, METHODS, AND APPARATUS FOR SEARCHING CONTENT USING HYBRID COLLABORATIVE FILTERS |
CN110148043A (en) * | 2019-03-01 | 2019-08-20 | 安徽省优质采科技发展有限责任公司 | The bid and purchase information recommendation system and recommended method of knowledge based map |
WO2020073528A1 (en) * | 2018-10-12 | 2020-04-16 | 平安科技(深圳)有限公司 | Session-based information push method and apparatus, computer device, and storage medium |
WO2021068610A1 (en) * | 2019-10-12 | 2021-04-15 | 平安国际智慧城市科技股份有限公司 | Resource recommendation method and apparatus, electronic device and storage medium |
-
2021
- 2021-06-01 CN CN202110607982.0A patent/CN113343086B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103593336A (en) * | 2013-10-30 | 2014-02-19 | 中国运载火箭技术研究院 | Knowledge pushing system and method based on semantic analysis |
CN104899242A (en) * | 2015-03-10 | 2015-09-09 | 四川大学 | Mechanical product design two-dimensional knowledge pushing method based on design intent |
DE102018104824A1 (en) * | 2017-03-13 | 2018-09-13 | General Motors Llc | SYSTEMS, METHODS, AND APPARATUS FOR SEARCHING CONTENT USING HYBRID COLLABORATIVE FILTERS |
WO2020073528A1 (en) * | 2018-10-12 | 2020-04-16 | 平安科技(深圳)有限公司 | Session-based information push method and apparatus, computer device, and storage medium |
CN110148043A (en) * | 2019-03-01 | 2019-08-20 | 安徽省优质采科技发展有限责任公司 | The bid and purchase information recommendation system and recommended method of knowledge based map |
WO2021068610A1 (en) * | 2019-10-12 | 2021-04-15 | 平安国际智慧城市科技股份有限公司 | Resource recommendation method and apparatus, electronic device and storage medium |
Non-Patent Citations (6)
Title |
---|
《Knowledge-based active push system for ecological design》;Lei Zhang 等;《Procedia CIRP》;20190131;第39-44页 * |
《基于属性映射的产品绿色设计方案优化方法》;张雷 等;《机械工程学报》;20190930;第153-161页 * |
基于用户偏好挖掘的电子商务协同过滤推荐算法研究;贺桂和;《情报科学》;20131205(第12期);第38-42页 * |
基于用户兴趣度量的知识发现服务精准推荐;丁梦晓等;《图书情报工作》;20190220(第03期);第21-29页 * |
基于设计人员需求的知识推送技术研究;刘同磊等;《电子设计工程》;20170120(第02期);第31-36页 * |
面向协同的产品设计知识推送研究;蒋翠清等;《中国机械工程》;20120825(第16期);第1972-1977页 * |
Also Published As
Publication number | Publication date |
---|---|
CN113343086A (en) | 2021-09-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111737495B (en) | Middle-high-end talent intelligent recommendation system and method based on domain self-classification | |
Aligon et al. | A collaborative filtering approach for recommending OLAP sessions | |
US8473473B2 (en) | Object oriented data and metadata based search | |
CN104933239A (en) | Hybrid model based personalized position information recommendation system and realization method therefor | |
CN111859160B (en) | Session sequence recommendation method and system based on graph neural network | |
CN106296312A (en) | Online education resource recommendation system based on social media | |
CN105488697A (en) | Potential customer mining method based on customer behavior characteristics | |
CN102063453A (en) | Method and device for searching based on demands of user | |
Lee et al. | Fundamentals of big data network analysis for research and industry | |
Sumiran | An overview of data mining techniques and their application in industrial engineering | |
CN104794242A (en) | Searching methods | |
Khatter et al. | An intelligent personalized web blog searching technique using fuzzy-based feedback recurrent neural network | |
CN108205572A (en) | A kind of searching method, device and equipment | |
US11941073B2 (en) | Generating and implementing keyword clusters | |
CN101751439A (en) | Image retrieval method based on hierarchical clustering | |
CN112989215B (en) | Sparse user behavior data-based knowledge graph enhanced recommendation system | |
Bibi et al. | Web search personalization using machine learning techniques | |
US20130332440A1 (en) | Refinements in Document Analysis | |
Hamad et al. | Knowledge-driven decision support system based on knowledge warehouse and data mining for market management | |
CN112487304B (en) | Method for establishing influence propagation model based on viewpoint vectorization | |
Kaur | Web content classification: a survey | |
Shah et al. | An agent based personalized intelligent E-learning | |
Yang et al. | A novel evolutionary method to search interesting association rules by keywords | |
CN113343086B (en) | Data-driven green design knowledge pushing method | |
CN112100395A (en) | Expert cooperation feasibility analysis method |
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 |