CN101388024B - Compression space high-efficiency search method based on complex network - Google Patents
Compression space high-efficiency search method based on complex network Download PDFInfo
- Publication number
- CN101388024B CN101388024B CN200810121364XA CN200810121364A CN101388024B CN 101388024 B CN101388024 B CN 101388024B CN 200810121364X A CN200810121364X A CN 200810121364XA CN 200810121364 A CN200810121364 A CN 200810121364A CN 101388024 B CN101388024 B CN 101388024B
- Authority
- CN
- China
- Prior art keywords
- user
- node
- comment
- commodity
- scoring
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 18
- 230000006835 compression Effects 0.000 title claims description 9
- 238000007906 compression Methods 0.000 title claims description 9
- 239000013598 vector Substances 0.000 claims description 22
- 238000011156 evaluation Methods 0.000 claims description 17
- 238000003012 network analysis Methods 0.000 claims description 4
- 230000009194 climbing Effects 0.000 abstract description 3
- 230000000694 effects Effects 0.000 abstract description 3
- 230000003213 activating effect Effects 0.000 abstract 2
- 230000000739 chaotic effect Effects 0.000 abstract 1
- 238000000691 measurement method Methods 0.000 abstract 1
- 238000005065 mining Methods 0.000 abstract 1
- 238000007781 pre-processing Methods 0.000 abstract 1
- 239000011159 matrix material Substances 0.000 description 5
- 230000004913 activation Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 241000196324 Embryophyta Species 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 230000002452 interceptive effect Effects 0.000 description 2
- 238000012552 review Methods 0.000 description 2
- 241000239290 Araneae Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to a high effect searching method for compressed space based on a complicate network, which aims at mining a core node with higher influence in a complicate network as an initiate active node, and then sequentially activating other nodes in the network according to the influence weights on the directed edge of the network, thereby furthest activating more nodes. The problem can be transformed into a problem of maximization of network overlay in the graph theory, which is proved as an NP difficulty in mathematics, and therefore, aiming to the characteristic that different parameter measurement methods only can detect a certain aspect of a complicate network in the complicated network, the invention provides a compressed space searching algorithm based on heuristic information. The compressed space searching algorithm comprises after preprocessing of common greed algorithm, hill climbing algorithm and high in-degree heuristic information algorithm, selecting three ordered optimal node sets from the global scope to be merged into a chaotic candidate node set, and adding suboptimum node sets of the three algorithms to form a candidate node universal set capable of complete enumeration in an effective time. The searching method compresses a huge, incompact original solution space with large amount of redundant information into another solution space which is concentrative, and is processible by a computer and provided with high heuristic information, thereby guaranteeing to find out a group of better solution in utmost.
Description
Technical field
The present invention relates to fields such as social network analysis and graph theory, particularly relate to a kind of compression space high-efficiency search method based on complex network.
Background technology
Along with increasing client begins online shopping, traditional target marketing method is inapplicable.Best marketing method is to allow client oneself carry out the recommendation and the marketing of commodity.Sociological result of study shows, and is often more similar between the often mutual each other user.Very similar probably between Lian Xi the user each other, and similar user tends to buy same product.At present the community network website is all the fashion, FaceBook for example, and Myspace, being widely current of social network sites such as Twitter makes online target marketing have the lot of data resource.Owing to utilize passing from mouth to mouth between the user, make and carry out the target marketing strategy more efficiently with quick by the virtual society network.
Carry out target marketing by community network, traditional method is by comparing different user calculates the user to the evaluation of same group of commodity similarity matrix, the similarity matrix that utilizes this method to obtain is very sparse, between most users without any incidence relation.In order to address this problem, the researcher has proposed a kind of method again, promptly utilize trusting relationship to carry out the expansion of community network, but because the trusting relationship on spider lines is a field independence, when certain user is added to another user in trust list or the list of friends, do not specify the field of being trusted, and the preference of different user and personal interest differ greatly, therefore only utilize trusting relationship expand can be very out of true.In order to solve the sparse problem of similarity matrix, and provide more accurate target marketing, the present invention proposes a kind of method of utilizing evaluating network to reduce the similarity matrix degree of rarefication, the similarity of will marking and evaluation similarity integrate reciprocal influence power effect between the computing node.Because the relation of evaluation is that the field is clear and definite, so the accuracy height of target marketing.
Investigate demonstration according to the global Nelson to 26,486 Internet users in 47 markets, 78% user thinks that other users' recommendation is the most reliable a kind of advertisement form.82% read the influence that decision-making that the user who estimates shows that they buy commodity directly is subjected to these evaluations.Therefore utilizing the user to estimate relation comes the influence power between the analysis user to concern it is very accurately.Evaluation of user has played important effect when client buys the commodity decision-making on the one hand, estimates individual preference and the interest that relation can reflect client well on the other hand.
For the core node of seeking in the network, usual way has to utilize node degree number, bee-line and greedy algorithm to wait to seek a maximum in the complex network and covers, and makes node as much as possible be activated.
Summary of the invention
The objective of the invention is to overcome the defective of above-mentioned technology, and provide a kind of compression space high-efficiency search method based on complex network, at in complex network, the characteristic of different parameter reflection network different aspects, and network coverage maximization is a np hard problem, different parameter measures can only be investigated characteristic in a certain respect, so the present invention proposes a kind of heuristic information compression stroke searching algorithm, can find out one group of more excellent separating most possibly.
The objective of the invention is to be achieved through the following technical solutions.This compression space high-efficiency search method based on complex network may further comprise the steps:
(1) grasps trusting relationship from each user at the personal homepage of commodity online comment website Epinions.com by web crawlers, the user is counted as the node among the figure, trusting relationship is counted as a directed edge, points to the user who is trusted from the user who trusts others;
(2) after two users mark to two or more identical goodses simultaneously, calculate these two users' commodity scoring similarity, when calculating two users' commodity scoring similarity, each user is regarded in the scoring of the commodity of common evaluation as an one-dimensional vector, utilize the distance between the vector to calculate two users similarity each other then, computing formula is as follows:
Wherein, Sim
p(p represents commodity for A, the B) scoring of the commodity between expression user A and user B similarity, and n represents the quantity of the common commodity of estimating of two users,
Represent the maximum possible distance of two vectors;
(3) often read comment that user C writes and the words of marking as a user D, can regard user D as an one-dimensional vector to the scoring of user C so, setting user C is a best result to the scoring of the own comment of being write, and we still utilize the distance between the vector to calculate two users comment scoring similarity each other.Computing formula is as follows:
Wherein, Sim
r(r represents comment for C, the D) scoring of the comment between expression user C and user D similarity, the quantity of the article that k representative of consumer C writes in some concrete fields,
Represent two maximum possible distances between the vector;
(4) selected a certain specific field, grasp out the user list of writing comment in this field and estimating this field comment then, choose a front N node, grasp each other trusting relationship of this N node, commodity scoring relation and comment then and estimate relation, be about to this N node in twos each other relation all excavate out;
(5) according to commodity scoring similarity and comment evaluation similarity trust network is expanded, form a complicated directed networks figure who comprises multiple heuristic information, the user is counted as the node in the digraph, commodity scoring relation or comment evaluation concern the limit of regarding as in the digraph, and commodity scoring similarity or comment are estimated the size of similarity as the weight on the directed edge;
(6) several different parameters according to social network analysis sort to the node in the whole directed networks, M both candidate nodes before selecting in the Bao Congsan kind algorithm ranking results, the value of M is to want the final core customer's that determines quantity, choose 3 * M both candidate nodes altogether, if interstitial content no show 3 * M that above-mentioned three kinds of situations are chosen, continue so to continue to add, make the number of both candidate nodes arrive 3 * M from M+1 node of three kinds of algorithms;
(7) in candidate's 3 * M node, choose M node core customer the most, total
Plant combination, draw the optimum solution of in 3 * M node, excavating M core node, just overall solution space more excellent separating.
Beneficial effect of the present invention: by greed, climb the mountain, the anticipating of in-degree ordering etc., this algorithm is chosen following heuristic information in global scope: will merge into candidate's nodal set of a chaos through three orderly optimum nodal sets that above-mentioned three kinds of algorithms are selected, and replenish the suboptimum nodal set that adds above-mentioned three kinds of algorithms, constitute candidate's node complete or collected works that in effective time, can enumerate fully.That is, with huge, loose, as to have a very big redundant information former solution space, being compressed into one can be by another solution space Computer Processing, that concentrate, that have high heuristic information.Can be described as and carried out the branch-and-bound on the global sense one time.Ensuingly enumerate entirely, solved above-mentioned three kinds of insoluble problems of general algorithm: promptly made full use of the interactive information between candidate's node, thereby avoided being absorbed in local optimum.
Description of drawings
Fig. 1 concerns synoptic diagram for user's community network among the present invention;
Embodiment
For making the purpose, technical solutions and advantages of the present invention clearer, the present invention is described in further detail below in conjunction with drawings and the specific embodiments:
Compression space high-efficiency search method based on complex network of the present invention, step is as follows:
(1) Epinions.com is a commodity online comment website, the user on the website, submit to commodity comment (review) and to these commodity give a mark (score), other user reads these comments and marking, and this feeding back evaluation (Rate), Epinions.com can also allow the user specify the trust network (Web of Trust) of oneself simultaneously.When certain user has read the shiploads of merchandise that another user write comment and felt helpful, he can add this user in the trust network of oneself to.Many researchers was often only at above-mentioned a kind of building network that concerns in the past, the researcher who for example has only studies trusting relationship, the researcher who has only pays close attention to commodity scoring similarity relation, only from the contact between two users of a side reflection, the information that is comprised is not comprehensive for each independent relation.The present invention has made up an online community network that comprises multiple relation, multiple relations such as trusting relationship, commodity scoring similarity and comment feedback similarity is merged, thereby generate a complicated community network that comprises a large amount of heuristic informations.
(2) trusting relationship can grasp at the personal homepage of Epinions.com from each user by web crawlers.The user is counted as the node among the figure, and trusting relationship is counted as a directed edge, points to the user who is trusted from the user who trusts others.
(3) after two users mark to two or more identical goodses simultaneously, just can calculate these two users' commodity scoring similarity.When calculating two users' commodity scoring similarity, each user is regarded in the scoring of the commodity of common evaluation as an one-dimensional vector, utilize the distance between the vector to calculate two users similarity each other then.Computing formula is as follows:
Wherein, Sim
p(p represents " product " (commodity) for A, the B) scoring of the commodity between expression user A and user B similarity, and n represents the quantity of the common commodity of estimating of two users,
Represent the maximum possible distance of two vectors, be best result when a user gives a mark to all common commodity of estimating, and another user is when giving a mark to minimum branch to all common commodity of estimating, the distance between two vectors reaches maximum so.
In the step (3) multiple relation is comprised that commodity scoring relation, comment evaluation relation and trusting relationship are incorporated in the same complex network, thereby reduced the degree of rarefication of user's similarity matrix effectively, can provide information more accurately for determining between the user influence power each other.
(4) often read comment that user C writes and the words of marking as a user D, can regard user D as an one-dimensional vector to the scoring of user C so, because the comment of commodity has reflected user's interest and individual character preference, thus we to set user C all be best result (5 minutes) to the scoring of own own the comment of being write.We still utilize the distance between the vector to calculate two users comment scoring similarity each other.Computing formula is as follows:
Wherein, Sim
r(r represents " review " (comment), the quantity of the article that k representative of consumer C writes in some concrete fields for C, the D) scoring of the comment between expression user C and user D similarity.
Represent two maximum possible distances between the vector, when the marking of the comment of user D being write as user C all was minimum branch, the distance between two vectors was maximum.
(5) selected a certain specific field, for example " books ", grasp out the user list of writing comment in the books field and estimating the comment of books field then, choose 30782 nodes in front, grasp trusting relationship, commodity scoring relation and the comment each other of these 30782 nodes then and estimate relation.Be about to these 30782 nodes in twos each other relation all excavate out.
(6) according to commodity scoring similarity and comment evaluation similarity trust network is expanded, formed a complicated directed networks figure who comprises multiple heuristic information.The user is counted as the node in the digraph, and commodity scoring relation or comment evaluation concern the limit of regarding as in the digraph, and commodity scoring similarity or comment are estimated the size of similarity as the weight on the directed edge.
(7) several different parameters according to social network analysis sort to the node in the whole directed networks, comprise using common greedy algorithm, hill-climbing algorithm, high in-degree heuristic information algorithm.M both candidate nodes before selecting from every kind of algorithm ranking results, the value of M are to want the final core customer's that determines quantity, can choose 3 * M both candidate nodes altogether.The node that satisfies one of following condition will be chosen for both candidate nodes:
Preceding M node according to high in-degree heuristic information algorithm ordering;
Preceding M node according to common greedy algorithm ordering;
Preceding M node according to the hill-climbing algorithm ordering;
If interstitial content no show 3 * M that above-mentioned three kinds of situations are chosen (both candidate nodes of algorithms of different ordering has repetition, and just ordering is different) continues to continue to add from M+1 node of three kinds of algorithms so, make the number of both candidate nodes arrive 3 * M.
(8) in candidate's 3 * M node, choose M node core customer the most, total
Plant combination, we enumerate search entirely to various combinations, thereby draw the optimum solution of excavating M core node in 3 * M node, just overall solution space more excellent separating.Owing to utilize preliminary heuristic information to carry out the screening of both candidate nodes, selected 3 * M candidate's node, compressed solution space significantly.In 3 * M both candidate nodes, adopt and enumerate search entirely, can consider this 3 * M the interactive information between the point fully, thereby can avoid being absorbed in the local optimum of hill-climbing algorithm.
A node itself contains four kinds of information, usefulness Node (c d) represents for a, b,
Wherein, a is that a node can activate the quantity of node on every side separately; B is a node to effective influence power summation of node on every side.B=sum (min (P, w[i] [j])), wherein sum is summation, and min gets smaller value, and i is that we discuss current node, j=1...30782, w[i] [j] be the influence power of i (current node) to j, P is an activation threshold.
C is the quantity of the in-degree summation of a node
D is the cooperation information between the node.Be that an independent node can't be with node activation on every side, still two node gangs then can activate two nodes.
Common greedy algorithm has only been utilized a of all nodes, b information; Hill-climbing algorithm has only been considered the b information of all nodes and preliminary d information, and the c information of all nodes has only been used in the in-degree ordering, the algorithm of enumerating entirely in the both candidate nodes of the present invention's structure has utilized a of some " elite's nodes ", b, c, the d full detail, thereby can obtain best result.
Concrete grammar is as follows:
(1) suppose that N user arranged in the community network, existing M sample commodity will be given to client on probation, and the quantity of candidate user is 3 * M.Utilize the above-mentioned several method of mentioning to choose both candidate nodes collection C then.
(2) utilize common greedy algorithm, choose M both candidate nodes, add up the quantity of each node energy " separately " activation node, promptly add up " the effectively weight summation " of each node, promptly suppose w[i] [j] be the influence power of node i to j, " effectively weight summation "=sum of i node (min (P, w[i] [j]) so), wherein its ordering rule of j=1..N is: activate earlier node quantity more separately, by the row that successively decreases (M is individual before promptly choosing); Under identical situation, according to effective weight summation row's (promptly choose before M) that successively decreases.
(3) utilize the greedy algorithm of climbing the mountain, choose M both candidate nodes, its core algorithm is a kind of simplification of search, promptly has an evaluation function f (), is used for estimating all candidate solutions, and the candidate of choosing optimum is separated and expanded.Its step:
(a) initialization solution space is changeed (b)
(b) obtain the f value of current each point, change (c)
(c) choose maximum f value and with its adding formation q; If formation q is full, then changes (d), otherwise change (b)
(d) finish hill climbing process, calculating can activate number.
The definition of f () then is " the effectively weight summation " under the mutual situation, has promptly deducted the associated weights between the node that is activated.
(4) utilize the in-degree of node in the community network to sort, select M both candidate nodes.
(5) if node repeats, not enough 3 * M node, several nodes before choosing according to (2) individual rule of common greedy algorithm so make candidate node sum reach 3 * M.
(6) in 3 * M node, enumerate search entirely, draw 3 * M the optimum solution in the node combination then, as more excellent separating in the whole network.
The foregoing description is used for the present invention that explains, rather than limits the invention, and in the protection domain of spirit of the present invention and claim, any modification and change to the present invention makes all fall into protection scope of the present invention.
Claims (2)
1. compression space high-efficiency search method based on complex network is characterized in that: may further comprise the steps:
(1) grasps trusting relationship from each user at the personal homepage of commodity online comment website Epinions.com by web crawlers, the user is counted as the node among the figure, trusting relationship is counted as a directed edge, points to the user who is trusted from the user who trusts others;
(2) after two users mark to two or more identical goodses simultaneously, calculate these two users' commodity scoring similarity, when calculating two users' commodity scoring similarity, each user is regarded in the scoring of the commodity of common evaluation as an one-dimensional vector, utilize the distance between the vector to calculate two users similarity each other then, computing formula is as follows:
Wherein, Sim
p(p represents commodity for A, the B) scoring of the commodity between expression user A and user B similarity, and n represents the quantity of the common commodity of estimating of two users,
Represent the maximum possible distance of two vectors, be best result when a user gives a mark to all common commodity of estimating, and another user is when giving a mark to minimum branch to all common commodity of estimating, the distance between two vectors reaches maximum so;
(3) often read comment that user C writes and the words of marking as a user D, can regard user D as an one-dimensional vector to the scoring of user C so, setting user C is a best result to the scoring of the own comment of being write, and we still utilize the distance between the vector to calculate two users comment scoring similarity each other; Computing formula is as follows:
Wherein, Sim
r(r represents comment for C, the D) scoring of the comment between expression user C and user D similarity, the quantity of the article that k representative of consumer C writes in some concrete fields,
Represent two maximum possible distances between the vector, when the marking of the comment of user D being write as user C all was minimum branch, the distance between two vectors was maximum;
(4) selected a certain specific field, grasp out the user list of writing comment in this field and estimating this field comment then, choose a front N node, grasp each other trusting relationship of this N node, commodity scoring relation and comment then and estimate relation, be about to this N node in twos each other relation all excavate out;
(5) according to commodity scoring similarity and comment evaluation similarity trust network is expanded, form a complicated directed networks figure who comprises multiple heuristic information, the user is counted as the node in the digraph, commodity scoring relation or comment evaluation concern the limit of regarding as in the digraph, and commodity scoring similarity or comment are estimated the size of similarity as the weight on the directed edge;
(6) several different parameters according to social network analysis sort to the node in the whole directed networks, comprise using common greedy algorithm, hill-climbing algorithm, high in-degree heuristic information algorithm; M both candidate nodes before selecting from every kind of algorithm ranking results, the value of M are to want the final core customer's that determines quantity, can choose 3 * M both candidate nodes altogether; The node that satisfies one of following condition will be chosen for both candidate nodes:
Preceding M node according to high in-degree heuristic information algorithm ordering;
Preceding M node according to common greedy algorithm ordering;
Preceding M node according to the hill-climbing algorithm ordering;
If interstitial content no show 3 * M that above-mentioned three kinds of situations are chosen continues to continue to add from M+1 node of three kinds of algorithms so, make the number of both candidate nodes arrive 3 * M;
(7) in candidate's 3 * M node, choose M node core customer the most, total
Plant combination, draw the optimum solution of in 3 * M node, excavating M core node, just overall solution space more excellent separating.
2. the compression space high-efficiency search method based on complex network according to claim 1 is characterized in that: in 3 * M both candidate nodes search is enumerated in various combinations employings entirely.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200810121364XA CN101388024B (en) | 2008-10-09 | 2008-10-09 | Compression space high-efficiency search method based on complex network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200810121364XA CN101388024B (en) | 2008-10-09 | 2008-10-09 | Compression space high-efficiency search method based on complex network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101388024A CN101388024A (en) | 2009-03-18 |
| CN101388024B true CN101388024B (en) | 2010-11-10 |
Family
ID=40477448
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200810121364XA Expired - Fee Related CN101388024B (en) | 2008-10-09 | 2008-10-09 | Compression space high-efficiency search method based on complex network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101388024B (en) |
Families Citing this family (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9990429B2 (en) | 2010-05-14 | 2018-06-05 | Microsoft Technology Licensing, Llc | Automated social networking graph mining and visualization |
| CN101937455B (en) * | 2010-08-27 | 2012-02-08 | 北京鸿蒙网科技有限公司 | Method for establishing multi-dimensional classification cluster based on infinite hierarchy and heredity information |
| CN102880608A (en) * | 2011-07-13 | 2013-01-16 | 阿里巴巴集团控股有限公司 | Ranking and searching method and ranking and searching device based on interpersonal distance |
| CN102270239A (en) * | 2011-08-15 | 2011-12-07 | 哈尔滨工业大学 | Evolution analysis method for associated networks in forum |
| CN102651030B (en) * | 2012-04-09 | 2013-10-30 | 华中科技大学 | Social network association searching method based on graphics processing unit (GPU) multiple sequence alignment algorithm |
| CN104158840B (en) * | 2014-07-09 | 2017-07-07 | 东北大学 | A kind of method of Distributed Calculation node of graph similarity |
| CN110209923B (en) * | 2018-06-12 | 2023-07-25 | 中国人民大学 | Topic influence user pushing method and device |
| CN109657105B (en) * | 2018-12-25 | 2021-10-22 | 杭州灿八科技有限公司 | Method for acquiring target user |
| CN111694900B (en) * | 2019-02-28 | 2023-06-13 | 阿里巴巴集团控股有限公司 | Network graph processing method and device |
| CN110110529B (en) * | 2019-05-20 | 2020-12-11 | 北京理工大学 | Software network key node mining method based on complex network |
| CN110719106B (en) * | 2019-09-27 | 2021-08-31 | 华中科技大学 | Social network graph compression method and system based on node classification and sorting |
| CN111666420B (en) * | 2020-05-29 | 2021-02-26 | 华东师范大学 | Method for intensively extracting experts based on subject knowledge graph |
-
2008
- 2008-10-09 CN CN200810121364XA patent/CN101388024B/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN101388024A (en) | 2009-03-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101388024B (en) | Compression space high-efficiency search method based on complex network | |
| Yang et al. | We know what@ you# tag: does the dual role affect hashtag adoption? | |
| US11070643B2 (en) | Discovering signature of electronic social networks | |
| Cui et al. | Who should share what? item-level social influence prediction for users and posts ranking | |
| Zhu et al. | Online purchase decisions for tourism e-commerce | |
| CN101694652B (en) | Network resource personalized recommended method based on ultrafast neural network | |
| CN110213164B (en) | Method and device for identifying network key propagator based on topology information fusion | |
| Ahmed et al. | A supervised learning approach to link prediction in Twitter | |
| CN104899229A (en) | Swarm intelligence based behavior clustering system | |
| CN111651678B (en) | Personalized recommendation method based on knowledge graph | |
| CN102456064B (en) | Method for realizing community discovery in social networking | |
| CN108038746A (en) | Method is recommended based on the bigraph (bipartite graph) of key user and time context | |
| Yin et al. | Exploring social activeness and dynamic interest in community-based recommender system | |
| CN109635206A (en) | Merge the personalized recommendation method and system of implicit feedback and user's social status | |
| Li et al. | Collaborative filtering based on user attributes and user ratings for restaurant recommendation | |
| Ahmed et al. | Enhancing link prediction in Twitter using semantic user attributes | |
| Juszczuk et al. | Real-world data difficulty estimation with the use of entropy | |
| CN103353865A (en) | Barter electronic trading commodity recommendation method based on position | |
| Alobaid et al. | Typology-based semantic labeling of numeric tabular data | |
| Zhou et al. | The crude oil international trade competition networks: Evolution trends and estimating potential competition links | |
| Bączkiewicz | Temporal SWARA-SPOTIS for multi-criteria assessment of European countries regarding sustainable RES exploitation | |
| Nath et al. | Ad impression forecasting for sponsored search | |
| Ferreira | Capturing asymmetry in real exchange rate with quantile autoregression | |
| CN113516553A (en) | Credit risk early warning method and device | |
| KR101549188B1 (en) | Apparatus and method for measuring brand image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20101110 Termination date: 20191009 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |