CN108009225A - Motif discovery and trend analysis based on technology policy text - Google Patents
Motif discovery and trend analysis based on technology policy text Download PDFInfo
- Publication number
- CN108009225A CN108009225A CN201711202951.7A CN201711202951A CN108009225A CN 108009225 A CN108009225 A CN 108009225A CN 201711202951 A CN201711202951 A CN 201711202951A CN 108009225 A CN108009225 A CN 108009225A
- Authority
- CN
- China
- Prior art keywords
- theme
- technology policy
- text
- msub
- mrow
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
- G06F16/355—Class or cluster creation or modification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
- G06F16/353—Clustering; Classification into predefined classes
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Machine Translation (AREA)
Abstract
The invention discloses a kind of motif discovery and trend analysis based on technology policy text, for the characteristics of technology policy text scale is big, length is long, policy text is modeled using LDA topic models, by text collection from the word space reflection of higher-dimension to the theme space of low-dimensional, solve the problems, such as that the text representation space higher-dimension that is easily produced when handling large-scale data is sparse, improve analysis efficiency;The invention also provides according to the similarity size that theme is included in technology policy, the cluster operation to technology policy is realized using k means algorithms, technology policy set is divided into different class clusters, for insufficient existing for k means algorithms, propose the k means innovatory algorithms based on community discovery, by the Centroid system of selection in community discovery and community, determine optimal number of clusters and initial cluster center, and the validity of improved method is proposed by experimental verification.
Description
Technical field
The present invention relates to technical field of information processing, more particularly to a kind of motif discovery based on technology policy text is with becoming
Method of potential analysis.
Background technology
The research emphasis of current technology policy, which concentrates on, to be researched and analysed China and formulates and implement which technology policy, which
Technology policy is that there are potential incidence relation, how further perfect between the hot spot policy in a certain period, which technology policy
The technology policy theory in China etc..Technology policy utilizes information as the specification criterion for instructing sciemtifec and technical sphere to develop in a healthy way
The processing of technique to high-efficiency and analysis policy text, have important practical value, and existing technology policy processing and analysis are basic
Using conventional Artificial Cognition and sorting technique, efficiency is low, heavy workload, and accuracy is poor, and therefore, exploitation design is a kind of
Trend analysis for technology policy text is this area urgent problem.
The content of the invention
The technical problems to be solved by the invention are to provide a kind of motif discovery based on technology policy text and trend point
Analysis method, is modeled policy text using LDA topic models, realize by text collection from the word space reflection of higher-dimension to
The theme space of low-dimensional, efficiently solves that the text representation space higher-dimension easily produced when handling large-scale data is sparse to ask
Technology policy set, is divided into different class clusters by topic, is determined number of clusters and the initial cluster center most having, is greatly improved section
The trend analysis efficiency of skill policy text and accuracy.
In order to solve the above technical problems, the technical solution used in the present invention is:It is a kind of to be based on technology policy text
Motif discovery and trend analysis, first according to the relation between technology policy theme and constructed theme storehouse, establish
Text mathematical model, is used for realization text message to the conversion process of digital information, then is associated point with the theme storehouse of structure
Analysis;Then motif discovery is realized to technology policy using improving clustering algorithm, establishes theme strength model, carry out topic distillation and
Theme trend analysis, it is therefore intended that the module of popular degree is established to the technology policy theme found, intensity is bigger, table
The attention rate of the bright theme is higher;Finally structure is used for the data mould for instructing the motif discovery and subject analysis for judging technology policy
Type, motif discovery and the trend analysis of technology policy text are carried out using constructed data model.
The foundation of the text mathematical model includes the following steps:
S1, select different theme quantity by artificial mode, and collection of document is carried out repeating to build using LDA models
Mould;
S2, calculate the corresponding puzzled angle value pre (D) of each modeling result;
Corresponding theme quantity is as optimal theme quantity when S3, selection puzzlement degree are minimum;
S4, after determining optimal theme quantity, be again modeled collection of document using LDA models, obtains optimal gather
Class result.
The foundation of the theme strength model includes:Time and region dual label for technology policy theme, calculates
Go out the corresponding subject importance IMP (ti) of technology policy, it is assumed that theme ti is the i-th theme in theme set T, theme ti by
N Feature Words composition, wherein ti={ w1, w2 ... wk } represent the category that theme ti is included《Technology policy theme dictionary》In k
Feature set of words, then the importance of theme ti be:
In formula, n represents the special testimony quantity included in theme ti, k represent to include in theme ti with《Scientific and technological descriptor
Storehouse》Matched Feature Words quantity, and p (wj | ti) represent that vocabulary wj appears in the probability in theme ti.
Technology policy theme trend analysis includes the following steps:Become by calculating the subject importance in different document set
Change, analysis theme is with time and the changing rule of region, and similarity is converted into the corresponding master of calculating document between calculating document
Distance between topic distribution, using JS distance functions as the standard for weighing two Documents Similarity sizes;It is then determined that cluster numbers
Measure k;Finally determine initial cluster center.
It is using beneficial effect caused by above-mentioned technical proposal:The present invention is for technology policy text scale is big, a piece
The characteristics of width is long, is modeled policy text using LDA topic models, by text collection from the word space reflection of higher-dimension to low
The theme space of dimension, efficiently solves that the text representation space higher-dimension easily produced when handling large-scale data is sparse to ask
Topic, improves analysis efficiency;The invention also provides according to the similarity size that theme is included in technology policy, utilize k-means
Algorithm realizes the cluster operation to technology policy, and technology policy set is divided into different class clusters, is deposited for k-means algorithms
Deficiency, it is proposed that the k-means innovatory algorithms based on community discovery, are clicked by the centromere in community discovery and community
Selection method, determine optimal number of clusters and initial cluster center, and proposes the effective of improved method by experimental verification
Property.
Brief description of the drawings
Fig. 1 is puzzled degree and theme Figure of the quantitative relationship;
Fig. 2 is theme intensive analysis figure under time conditions;
Fig. 3 is theme intensive analysis figure under regional condition;
Fig. 4 is the F- metric comparison diagrams of two kinds of algorithms.
Embodiment
With reference to the attached drawing in the embodiment of the present invention, the technical solution in the embodiment of the present invention is carried out clear, complete
Ground describes, it is clear that described embodiment is only the part of the embodiment of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, those of ordinary skill in the art are obtained every other without making creative work
Embodiment, belongs to the scope of protection of the invention.
Many details are elaborated in the following description to facilitate a thorough understanding of the present invention, still the present invention can be with
Implemented using other different from other manner described here, those skilled in the art can be without prejudice to intension of the present invention
In the case of do similar popularization, therefore the present invention is from the limitation of following public specific embodiment.
As shown in Figure 1, the invention discloses a kind of motif discovery and trend analysis based on technology policy text, it is first
First according to the relation between technology policy theme and constructed theme storehouse, text mathematical model is established, is used for realization text envelope
The conversion process of digital information is ceased, then analysis is associated with the theme storehouse of structure;Then using improving clustering algorithm to section
Skill policy realizes motif discovery, establishes theme strength model, carries out topic distillation and theme trend analysis, it is therefore intended that to being sent out
Existing technology policy theme establishes the module of popular degree, and intensity is bigger, shows that the attention rate of the theme is higher;Last structure
Build for instructing to judge the motif discovery of technology policy and the data model of subject analysis, carried out using constructed data model
The motif discovery of technology policy text and trend analysis.
The foundation of the text mathematical model includes the following steps:
S1, select different theme quantity by artificial mode, and collection of document is carried out repeating to build using LDA models
Mould;
S2, calculate the corresponding puzzled angle value pre (D) of each modeling result;
Corresponding theme quantity is as optimal theme quantity when S3, selection puzzlement degree are minimum;
S4, after determining optimal theme quantity, be again modeled collection of document using LDA models, obtains optimal gather
Class result.
The foundation of the theme strength model includes:Time and region dual label for technology policy theme, calculates
Go out the corresponding subject importance IMP (ti) of technology policy, it is assumed that theme ti is the i-th theme in theme set T, theme ti by
N Feature Words composition, wherein ti={ w1, w2 ... wk } represent the category that theme ti is included《Technology policy theme dictionary》In k
Feature set of words, then the importance of theme ti be:
In formula, n represents the special testimony quantity included in theme ti, k represent to include in theme ti with《Scientific and technological descriptor
Storehouse》Matched Feature Words quantity, and p (wj | ti) represent that vocabulary wj appears in the probability in theme ti.
Technology policy theme trend analysis includes the following steps:Become by calculating the subject importance in different document set
Change, analysis theme is with time and the changing rule of region, and similarity is converted into the corresponding master of calculating document between calculating document
Distance between topic distribution, using JS distance functions as the standard for weighing two Documents Similarity sizes;It is then determined that cluster numbers
Measure k;Finally determine initial cluster center.
Concrete application process and verification are analyzed as follows:
First, technology policy motif discovery and analysis
LDA topic models quickly find text collection using the semantic relation inside statistical method analysis text collection
In potential subject information, motif discovery fangfa using LDA models to pretreated technology policy content carry out theme build
Mould, the quick subject information for finding to imply in technology policy set.
Reciprocity technology policy motif discovery process can be described as:
Puzzlement degree assesses optimal theme quantity, in general by measuring predictive ability of the LDA models to collection of document
When puzzlement degree reaches minimum value, the effect of LDA modelings is best, and theme quantity at this time is optimal, and puzzlement degree pre (D) is calculated,
M represents the document number in D, and Ni represents the length of document di, and p (di) represents that LDA models produce the probability of document di.Using commenting
Valency puzzlement degree determines that the LDA modeling procedures of optimal theme quantity are as follows:
(1) different theme quantity is selected by artificial mode first, collection of document is repeated using LDA models
Modeling;
(2) the corresponding puzzled angle value pre (D) of each modeling result is calculated;
(3) corresponding theme quantity is as optimal theme quantity when selecting puzzlement degree minimum;
(4) after determining optimal theme quantity, collection of document is modeled again using LDA models, obtains optimal gather
Class result.Document-theme matrix D T, theme-characteristic item matrix TW and theme set T are obtained after the completion of LDA modelings.Using
LDA models are modeled technology policy set, document-theme matrix of acquisition.
Technology policy subject importance is measured;
Subject importance IMP (ti):Assuming that theme ti is the i-th theme in theme set T, theme ti is by n feature
Word forms, and wherein ti={ w1, w2 ... wk } represents the category that theme ti is included《Technology policy theme dictionary》In k Feature Words
Set.Then the importance of theme ti is:
In formula, n represents the special testimony quantity included in theme ti, k represent to include in theme ti with《Scientific and technological descriptor
Storehouse》Matched Feature Words quantity, and p (wj | ti) represent that vocabulary wj appears in the probability in theme ti.
2nd, technology policy theme trend analysis
1st, clustering algorithm
Input:Data set N to be clustered, number of clusters k.Output:K cluster C={ c1, c2 ... ck }
(1) k object is randomly choosed from the data acquisition system being made of N number of object as initial center;
(2) according to the average of each class cluster, the distance of each object and cluster center is calculated, will be each right by nearby principle
As being assigned in closest cluster;
(3) average of each class cluster is computed repeatedly, algorithm terminates when meeting end condition;Otherwise step (2) is returned to, it is false
If the sample size in data acquisition system is N, the dimension of each sample is M, and number of clusters is defined as K, then utilizes k-means algorithms
Carry out T iteration time complexity be 0 (T × K × N × M), it can be seen that the time complexity of k-means algorithms with
The increase of number of clusters and iterations and increase, k-means algorithm operational process is simple, has good retractility and can
Autgmentability, therefore can efficiently handle large-scale data acquisition system.
2nd, innovatory algorithm
Assuming that d1, d2 are two texts in text collection D, Djs (d1, d2) represents the similarity between d1 and d2, if
Determine δ and represent similarity threshold, if Djs (d1, d2)≤δ, then text d1 and d2 neighbours each other.
Defined according to neighbours, iterate to calculate similarity between document, the neighborhood of each document is determined, by finding document
Between neighborhood original document is converted into arbitrary node in neighbor networks, if there are neighborhood between two nodes,
Then between two nodes there are a line, and so on, show that complete neighborhood network G={ V, E }, wherein V represent section
Point set, E represent the set on side.
3rd, experimental verification and analysis
(1) technology policy motif discovery and filtering
1. Fig. 1 is shown:Select k to be equal to 10,20,30,40,50,60,70,80,90,100 successively and carry out theme modeling, and
Analyze variation tendency of the puzzlement degree in this ten modeling process.
2. Fig. 2 is shown:Model and have selected 9,24 and 32 3 themes of theme in 46 themes obtained, carry out theme intensity
Analysis.
3. Fig. 3 is shown:Policy execution scope then have selected " Hebei province ", " Henan Province " and " Tianjin " three regions, point
The intensity in different geographical in these three themes is not calculated.
Experimental analysis:
1. when theme quantity is equal to 80, puzzlement degree reaches minimum value, therefore selects k=80 as optimal number of topics
Amount, then using Gibbs model method iteration 1000 times, obtains probability distribution matrix DT, TW and theme set T.
2. ascendant trend is presented with the time in the intensity of theme 9,32 intensity of theme is gradually reduced with the time, and theme 24 exists
Decennary strength fluctuation is little.3. and be respectively 6 in Tick, at 7,8,14,16, network gross energy is smaller, network information flow
Behavior correspondingly almost no longer occurs.
3. in the technology policy of Hebei province's issue, the maximum intensity of high technology industry special topic, the scientific and technological political affairs of Tianjin issue
Plan lays particular emphasis on finance support special topic, and the technology policy in Henan Province is then mainly distributed on finance support and high technology industry aspect.
(2) technology policy Subject Clustering compares
1. Fig. 4 is represented:The comparative result of clustering algorithm and traditional clustering algorithm after improvement.
Experimental analysis:Due in improved k-means clustering algorithms, optimizing initial cluster center, therefore cluster knot
Fruit is stablized relatively, and the accuracy rate of cluster result and F- metrics are all apparently higher than traditional k-means algorithms, effectively lifting
The clustering result quality of k-means algorithms
In short, the present invention is directed to the characteristics of technology policy text scale is big, length is long, using LDA topic models to policy
Text is modeled, and by text collection from the word space reflection of higher-dimension to the theme space of low-dimensional, is efficiently solved big in processing
The problem of text representation space higher-dimension easily produced during scale data is sparse, improves analysis efficiency;The invention also provides according to
According to the similarity size that theme is included in technology policy, the cluster operation to technology policy is realized using k-means algorithms, by section
Skill policy sets are divided into different class clusters, for insufficient existing for k-means algorithms, it is proposed that the k- based on community discovery
Means innovatory algorithms, by the Centroid system of selection in community discovery and community, determine optimal number of clusters and initial
Cluster centre, and the validity of improved method is proposed by experimental verification.
Claims (4)
1. a kind of motif discovery and trend analysis based on technology policy text, it is characterised in that:First according to scientific and technological political affairs
Relation between plan theme and constructed theme storehouse, establishes text mathematical model, is used for realization text message to digital information
Conversion process, then be associated analysis with the theme storehouse of structure;Then using improve clustering algorithm to technology policy realize lead
Topic is found, establishes theme strength model, carries out topic distillation and theme trend analysis, it is therefore intended that to the technology policy found
Theme establishes the module of popular degree, and intensity is bigger, shows that the attention rate of the theme is higher;Finally build and sentence for guidance
The motif discovery of disconnected technology policy and the data model of subject analysis, technology policy text is carried out using constructed data model
Motif discovery and trend analysis.
2. motif discovery and trend analysis according to claim 1 based on technology policy text, it is characterised in that:
The foundation of the text mathematical model includes the following steps:
S1, select different theme quantity by artificial mode, and collection of document is carried out using LDA models to repeat modeling;
S2, calculate the corresponding puzzled angle value pre (D) of each modeling result;
Corresponding theme quantity is as optimal theme quantity when S3, selection puzzlement degree are minimum;
S4, after determining optimal theme quantity, be again modeled collection of document using LDA models, obtains optimal cluster knot
Fruit.
3. motif discovery and trend analysis according to claim 1 based on technology policy text, it is characterised in that:
The foundation of the theme strength model includes:Time and region dual label for technology policy theme, calculates scientific and technological political affairs
The corresponding subject importance IMP (ti) of plan, it is assumed that theme ti is the i-th theme in theme set T, and theme ti is by n feature
Word forms, and wherein ti={ w1, w2 ... wk } represents the category that theme ti is included《Technology policy theme dictionary》In k Feature Words
Set, then the importance of theme ti is:
<mrow>
<mi>T</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>t</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<msub>
<mi>d</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>t</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<msub>
<mi>d</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<msub>
<mi>t</mi>
<mn>2</mn>
</msub>
<mrow>
<msubsup>
<mo>&Sigma;</mo>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</msubsup>
<msub>
<mi>t</mi>
<mi>k</mi>
</msub>
</mrow>
</mfrac>
</mrow>
In formula, n represents the special testimony quantity included in theme ti, k represent to include in theme ti with《Scientific and technological theme dictionary》
The Feature Words quantity matched somebody with somebody, and p (wj | ti) represent that vocabulary wj appears in the probability in theme ti.
4. motif discovery and trend analysis according to claim 3 based on technology policy text, it is characterised in that:
Technology policy theme trend analysis includes the following steps:Changed by calculating the subject importance in different document set, analysis
Theme is with time and the changing rule of region, and similarity is converted between the corresponding theme distribution of calculating document between calculating document
Distance, using JS distance functions as weigh two Documents Similarity sizes standard;It is then determined that number of clusters k;Finally
Determine initial cluster center.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711202951.7A CN108009225A (en) | 2017-11-27 | 2017-11-27 | Motif discovery and trend analysis based on technology policy text |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711202951.7A CN108009225A (en) | 2017-11-27 | 2017-11-27 | Motif discovery and trend analysis based on technology policy text |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108009225A true CN108009225A (en) | 2018-05-08 |
Family
ID=62053735
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711202951.7A Pending CN108009225A (en) | 2017-11-27 | 2017-11-27 | Motif discovery and trend analysis based on technology policy text |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108009225A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109726402A (en) * | 2019-01-11 | 2019-05-07 | 中国电子科技集团公司第七研究所 | A kind of document subject matter word extraction method |
CN114417837A (en) * | 2022-01-19 | 2022-04-29 | 合肥工业大学 | Method for measuring popularity and frontier of science and technology big data based on subject evolution trend |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103793501A (en) * | 2014-01-20 | 2014-05-14 | 惠州学院 | Theme community discovery method based on social network |
US20150154305A1 (en) * | 2013-12-02 | 2015-06-04 | Qbase, LLC | Method of automated discovery of topics relatedness |
CN105335349A (en) * | 2015-08-26 | 2016-02-17 | 天津大学 | Time window based LDA microblog topic trend detection method and apparatus |
KR20160113826A (en) * | 2015-03-23 | 2016-10-04 | 목포대학교산학협력단 | A Method on Initial Seeds Selection of K-Means for Big Data Clustering |
-
2017
- 2017-11-27 CN CN201711202951.7A patent/CN108009225A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150154305A1 (en) * | 2013-12-02 | 2015-06-04 | Qbase, LLC | Method of automated discovery of topics relatedness |
CN103793501A (en) * | 2014-01-20 | 2014-05-14 | 惠州学院 | Theme community discovery method based on social network |
KR20160113826A (en) * | 2015-03-23 | 2016-10-04 | 목포대학교산학협력단 | A Method on Initial Seeds Selection of K-Means for Big Data Clustering |
CN105335349A (en) * | 2015-08-26 | 2016-02-17 | 天津大学 | Time window based LDA microblog topic trend detection method and apparatus |
Non-Patent Citations (1)
Title |
---|
李少博: ""基于主题的科技政策分析系统设计与实现"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109726402A (en) * | 2019-01-11 | 2019-05-07 | 中国电子科技集团公司第七研究所 | A kind of document subject matter word extraction method |
CN109726402B (en) * | 2019-01-11 | 2022-12-23 | 中国电子科技集团公司第七研究所 | Automatic extraction method for document subject term |
CN114417837A (en) * | 2022-01-19 | 2022-04-29 | 合肥工业大学 | Method for measuring popularity and frontier of science and technology big data based on subject evolution trend |
CN114417837B (en) * | 2022-01-19 | 2024-02-13 | 合肥工业大学 | Scientific and technological big data popularity and frontier measurement method based on subject evolution trend |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101916382B (en) | Method for recognizing image of plant leaf | |
CN106503148B (en) | A kind of table entity link method based on multiple knowledge base | |
CN104346481B (en) | A kind of community detection method based on dynamic synchronization model | |
CN111143567B (en) | Comment emotion analysis method based on improved neural network | |
CN111178077A (en) | Corpus generation method, corpus generation device and intelligent device | |
TWI464604B (en) | Data clustering method and device, data processing apparatus and image processing apparatus | |
CN105303450A (en) | Complex network community discovery method based on spectral clustering improved intersection | |
CN108717601A (en) | A kind of more innovative approach towards enterprise's problem are integrated and fusion method | |
CN101576903A (en) | Method for measuring similarity of documents | |
CN106997553A (en) | Multi-objective optimization-based commodity combination mode mining method | |
CN108009225A (en) | Motif discovery and trend analysis based on technology policy text | |
CN103869999B (en) | The method and device that candidate item caused by input method is ranked up | |
CN101697167A (en) | Clustering-decision tree based selection method of fine corn seeds | |
CN106599924A (en) | Classifier construction method based on three-way decision | |
CN106251241A (en) | A kind of feature based selects the LR Bagging algorithm improved | |
CN106611016B (en) | A kind of image search method based on decomposable word packet model | |
CN105631465A (en) | Density peak-based high-efficiency hierarchical clustering method | |
CN107451617A (en) | One kind figure transduction semisupervised classification method | |
CN116738551B (en) | Intelligent processing method for acquired data of BIM model | |
CN112085335A (en) | Improved random forest algorithm for power distribution network fault prediction | |
CN105354264A (en) | Locality-sensitive-hashing-based subject label fast endowing method | |
CN108509764A (en) | A kind of extinct plants and animal pedigree evolution analysis method based on genetic property yojan | |
CN110705650A (en) | Metal plate layout method based on deep learning | |
WO2021017638A2 (en) | Method for determining similarity of any two technology systems | |
KR20230122872A (en) | Transfer learning system and method for deep neural network |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180508 |