### Citations

3242 | Mining Association Rules between Sets of Items in Large Databases
- Agrawal, Imielinski, et al.
- 1993
(Show Context)
Citation Context ...effectiveness of the algorithms. 1. INTRODUCTION Association rules have been well studied for discovering regularities between items in relational data, for promotional pricing and product placements =-=[4, 45]-=-. They have a traditional form X ⇒ Y , where X and Y are disjoint itemsets. There have been recent interests in studying associations between entities in social graphs. Such associations are useful in... |

300 | An apriori-based algorithm for mining frequent substructures from graph data
- Inokuchi, Washio, et al.
- 2000
(Show Context)
Citation Context ...pattern-query processing in big graphs. (d) Mining (diversified) GPARs is beyond rule mining from itemsets [46]. Graph pattern mining. There have been algorithms for pattern mining in graph databases =-=[22,24]-=- (see [25] for a survey). Large-scale mining techniques are also studied in a single graph [13], notably top-k algorithms [16, 27, 42, 44]. To reduce the cost, scalable subgraph isomorphism algorithms... |

179 |
A (sub)graph isomorphism algorithm for matching large graphs
- Cordella, Foggia, et al.
- 2004
(Show Context)
Citation Context ...AR. ✷ A naive way to compute Σ(x,G, η) is as follows. For each R(x, y) : Q(x, y) ⇒ q(x, y) in Σ, (a) enumerate all matches of Qq̄ and PR in G by using an algorithm for subgraph isomorphism, e.g., VF2 =-=[10]-=-; (b) compute supp(q,G) and supp(q̄, G) once in G; then based on the findings, (c) identify those R with conf(R,G) ≥ η, and return matches of x by these GPARs. This is cost-prohibitive (e.g., takes O(... |

114 |
Link prediction in complex networks: a survey
- Lü, Zhou
(Show Context)
Citation Context ... pattern Q4. ✷ The need for graph-pattern association rules (GPARs) is evident in social media marketing, community structure analysis, social recommendation, knowledge extraction and link prediction =-=[33]-=-. Such rules, however, depart from association rules for itemsets, and introduce several challenges. (1) Conventional support and confidence metrics no longer work for GPARs. (2) Mining algorithms for... |

100 | An axiomatic approach for result diversification
- Gollapudi, Sharma
- 2009
(Show Context)
Citation Context ... are too “homogeneous” [5]. Given a set Lk of k GPARs that pertain to the same predicate q(x, y), we define the objective function F (Lk) again by following the practice of social recommender systems =-=[19]-=-: (1− λ) ∑ Ri∈S conf(Ri) N + 2λ k − 1 ∑ Ri,Ri∈S,i<j diff(Ri, Rj). This, known as max-sum diversification, aims to strike a balance between interestingness (measured by revised Bayes Factor) and divers... |

92 |
A Complexity Theory of Efficient Parallel I Algorithms
- Kruskal, Rudolph, et al.
- 1990
(Show Context)
Citation Context ...multi-pattern matching. To the best of our knowledge, these are among the first algorithms on big graphs that guarantee a polynomial speedup over sequential algorithms with the increase of processors =-=[30]-=-. (c) We propose optimization strategies that are not studied by previous work. This said, prior optimization techniques can be incorporated into GPAR-based entity identification; e.g., the methods of... |

73 | Mining association rules in folksonomies
- Schmitz, Hotho, et al.
- 2006
(Show Context)
Citation Context ... Work. We categorize related work as follows. Association rules. Introduced in [4], association rules are defined on relations of transaction data. Prior work on association rules for social networks =-=[41]-=- and RDF knowledge bases resorts to mining conventional rules and Horn rules (as conjunctive binary predicates) [17] over tuples with extracted attributes from social graphs, instead of exploiting gra... |

67 | Conditional functional dependencies for capturing data inconsistencies
- Fan, Geerts, et al.
(Show Context)
Citation Context ... customers x of y, characterized by a social group of three members. (3) Association rules with graph patterns conveniently extend data dependencies such as conditional functional dependencies (CFDs) =-=[14]-=- in the context of social networks. ◦ If the addresses of x and x′ have the same country code “44” and same zip code, and if x′ shops at a Tesco store y with the same zip, then x may also shop at y. S... |

62 | Information diffusion and external influence in networks
- Myers, Zhu, et al.
- 2012
(Show Context)
Citation Context ...s negative example for R. However, as observed in [11, 17], the standard confidence is blind to the distinction between “negative” and “unknown”. This is particularly an overkill when G is incomplete =-=[11,34]-=-. Example 6: Consider patternQ2 in Fig. 1(b). LetQ2(x,G) contain three matches v1, v2, v3 of x1, x2, x3 in a social graph G, all living in Ecuador, where (1) v1 has an edge like to Shakira album, (2) ... |

43 | A fast bisimulation algorithm
- Dovier, Piazza, et al.
(Show Context)
Citation Context ...ine 6) and locally at each Si (localMine). It is costly to conduct pairwise automorphism tests on all GPARs in ∆E, since it is equivalent to graph isomorphism. To reduce the cost, we use bisimulation =-=[12]-=-. A graph pattern PR1 is bisimilar to PR2 if there exists a binary relation Ob on nodes of PR1 and PR2 such that (a) for all nodes u1 in PR1 , there exists a node u2 in PR2 with the same label such th... |

42 | What is frequent in a single graph
- Bringmann, Nijssen
- 2008
(Show Context)
Citation Context ... metrics for GPARs (Section 3). Conventional support for itemsets is no longer anti-monotonic for GPARs. We define support in terms of distinct “potential customers” by revising a measure proposed by =-=[7]-=-. We propose a confidence measure for GPARs by revising Bayes Factor [31] to incorporate the local closed world assumption [11,17]. This allows us to cope with (incomplete) social graphs, and to ident... |

42 |
Evaluating the interestingness of characteristic rules
- Kamber, Shinghal
- 1996
(Show Context)
Citation Context ...re matches x of Q(x, y), and is more discriminant if it is less likely to hold on more nodes from Qq̄. In addition, BF-based conf(R,G) is better justified than conventional confidence. As verified in =-=[26, 31]-=-, BF satisfies a set of principles for reasonable interestingness measures, including fixed under independence (conf(R,G) = 1 if Q and q are statistically independent), fixed under incompatibility (co... |

41 |
Association rule mining: models and algorithms
- Zhang, Zhang
- 2002
(Show Context)
Citation Context ...r inequality needed for functions. (c) Applying GPARs becomes an intractable problem of multipattern-query processing in big graphs. (d) Mining (diversified) GPARs is beyond rule mining from itemsets =-=[46]-=-. Graph pattern mining. There have been algorithms for pattern mining in graph databases [22,24] (see [25] for a survey). Large-scale mining techniques are also studied in a single graph [13], notably... |

37 |
Constrained frequent pattern mining: a pattern-growth view
- Pei, Han
- 2002
(Show Context)
Citation Context ...cision problem is NP-hard. ✷ 4.2 Discovery Algorithm One might want to follow a “discover and diversify” approach that (1) first finds all GPARs pertaining to q(x, y) by frequent graph pattern mining =-=[35]-=-, and then (2) selects top-k GPARs via result diversification [19]. However, this is costly: (a) an excessive number of GPARs are generated; and (b) for all GPARs R generated, it has to compute conf(R... |

36 | Extracting Redundancy-Aware Top-k Patterns
- Xin, Cheng, et al.
(Show Context)
Citation Context ...ning. There have been algorithms for pattern mining in graph databases [22,24] (see [25] for a survey). Large-scale mining techniques are also studied in a single graph [13], notably top-k algorithms =-=[16, 27, 42, 44]-=-. To reduce the cost, scalable subgraph isomorphism algorithms, e.g., [38], can be adopted to generate pattern candidates. Diversity of graph patterns is not studied there. However, (a) pattern mining... |

35 | Mining graph evolution rules
- Berlingerio, Bonchi, et al.
- 2009
(Show Context)
Citation Context ...bases resorts to mining conventional rules and Horn rules (as conjunctive binary predicates) [17] over tuples with extracted attributes from social graphs, instead of exploiting graph patterns. While =-=[6]-=- studies time-dependent rules via graph patterns, it focuses on evolving graphs and hence adopts different semantics for support and confidence. GPARs extend association rules from relations to graphs... |

32 |
la Fuente. An empirical study of real-world SPARQL queries
- Arias, Fernández, et al.
(Show Context)
Citation Context ...vx), i.e., the subgraph of G induced by Nd(vx), is in some fragment; and (b) the fragments have roughly even size. These are possible since 98% of real-life patterns have radius 1, 1.8% have radius 2 =-=[18]-=-, and the average node degree is 14.3 in social graphs [8]; thus Gd(vx) is typically small compared with fragment size. Fragment Fi is stored at worker Si, for i ∈ [1, n− 1]. (2) DMine discovers GPARs... |

28 | Aiding the detection of fake accounts in large scale social online services
- Cao, Sirivianos, et al.
(Show Context)
Citation Context ...ermines street. 1502 (4) The applications of association rules are not limited to marketing activities. They also help us detect scams. As an example, the rule below is used to identify fake accounts =-=[9]-=-. ◦ If (a) account x′ is confirmed fake, (b) both x and x′ like blogs P1, . . . , Pk, (c) x posts blog y1, (d) x ′ posts y2, and (e) if y1 and y2 contain the same particular content (keyword), then x ... |

27 | M.: A survey of frequent subgraph mining algorithms. The Knowledge Engineering Review
- Jiang, Coenen, et al.
- 2013
(Show Context)
Citation Context ... processing in big graphs. (d) Mining (diversified) GPARs is beyond rule mining from itemsets [46]. Graph pattern mining. There have been algorithms for pattern mining in graph databases [22,24] (see =-=[25]-=- for a survey). Large-scale mining techniques are also studied in a single graph [13], notably top-k algorithms [16, 27, 42, 44]. To reduce the cost, scalable subgraph isomorphism algorithms, e.g., [3... |

24 | Parallel evaluation of conjunctive queries
- Koutris, Suciu
- 2011
(Show Context)
Citation Context ...better still, the larger n is, the smaller T (|G|, |Σ|, n) is. Remark. Algorithm DMine (Section 4.2) takes t(|A|/n, k) time and is parallel scalable if the problem size |A| is measured as |G|+|Q|+|Σ| =-=[29]-=-. Indeed, if one wants all candidate GPARs R with supp(R,G) ≥ σ, then |Σ| is the size of the output, and |Σ| is not large (due to small d and large σ). 5.2 Optimization Strategies Algorithm Matchc jus... |

24 | Knowledge discovery with genetic programming for providing feedback to courseware author”, User Modeling and User-Adapted
- Romero, Ventura, et al.
- 2004
(Show Context)
Citation Context ...owing [4]. However, a consequent can be readily extended to multiple predicates and even to a graph pattern. (2) Conventional association rules [4] and a range of predication and classification rules =-=[39]-=- are a special case of GPARs, since their antecedents can be modeled as a graph pattern in which nodes denote items. Conditional functional dependencies [14] can also be represented by GPARs (see Q3 o... |

23 |
Mining social networks for targeted advertising
- Yang, Dia, et al.
- 2006
(Show Context)
Citation Context ...effectiveness of the algorithms. 1. INTRODUCTION Association rules have been well studied for discovering regularities between items in relational data, for promotional pricing and product placements =-=[4, 45]-=-. They have a traditional form X ⇒ Y , where X and Y are disjoint itemsets. There have been recent interests in studying associations between entities in social graphs. Such associations are useful in... |

18 | AMIE: association rule mining under incomplete evidence in ontological knowledge bases
- Galárraga, Teflioudi, et al.
- 2013
(Show Context)
Citation Context ...terms of distinct “potential customers” by revising a measure proposed by [7]. We propose a confidence measure for GPARs by revising Bayes Factor [31] to incorporate the local closed world assumption =-=[11,17]-=-. This allows us to cope with (incomplete) social graphs, and to identify interesting GPARs with correlated antecedent and consequent. (3) We study a new mining problem, referred to as the diversified... |

15 | An incremental bisimulation algorithm
- Saha
- 2007
(Show Context)
Citation Context ...ks automorphism between R1 and R2 only if so. It takes O(|∆E| 2) time to check pairwise bisimilarity Ob for all GPARs in ∆E [12]. Moreover, Ob can be incrementally maintained when new GPARs are added =-=[40]-=-. These allow us to use efficient (incremental) bisimulation tests instead of automorphism tests. Trivial GPARs. DMine detects trivial GPARs R(x, y): Q(x, y) ⇒ q(x, y) at Sc as follows: (1) if supp(q,... |

13 | Association rule interestingness: Measure and statistical validation
- Lallich, Teytaud, et al.
- 2007
(Show Context)
Citation Context ...longer anti-monotonic for GPARs. We define support in terms of distinct “potential customers” by revising a measure proposed by [7]. We propose a confidence measure for GPARs by revising Bayes Factor =-=[31]-=- to incorporate the local closed world assumption [11,17]. This allows us to cope with (incomplete) social graphs, and to identify interesting GPARs with correlated antecedent and consequent. (3) We s... |

10 | Scalable multi-query optimization for SPARQL
- Le, Kementsietsidis, et al.
- 2012
(Show Context)
Citation Context ...discovering diversified graph patterns. Graph pattern matching. Several parallel algorithms have been developed for subgraph isomorphism, e.g., [28, 37, 38], and for multi-pattern optimization, e.g., =-=[23, 32]-=-. Our algorithms for EIP differ from the prior work in the following. (a) Instead of enumerating isomorphic matches, EIP identifies a potential customer once one match is found, and moreover, computes... |

10 |
Jabe-Ja: A distributed algorithm for balanced graph partitioning
- Rahimian, Payberah, et al.
- 2013
(Show Context)
Citation Context ...tchc (Section 5.1), (b) disVF2, a parallel implementation of VF2 for EIP, and (c) Matchs, Match by using the method of [38] instead of VF2. Fragmentation and distribution. We revised the algorithm of =-=[36]-=- to evenly partition graph G into n fragments (see Section 4.2). We find that the gap between maximum and 1510s100s200s300s400s500s600s700s800s900s1000 4 8 12 16 20 Ti m e ( sec on d) DMine DMineno (a... |

7 | Efficient discovery of frequent correlated subgraph pairs
- Ke, Cheng, et al.
(Show Context)
Citation Context ...ning. There have been algorithms for pattern mining in graph databases [22,24] (see [25] for a survey). Large-scale mining techniques are also studied in a single graph [13], notably top-k algorithms =-=[16, 27, 42, 44]-=-. To reduce the cost, scalable subgraph isomorphism algorithms, e.g., [38], can be adopted to generate pattern candidates. Diversity of graph patterns is not studied there. However, (a) pattern mining... |

5 | Distributed graph simulation: Impossibility and possibility
- Fan
(Show Context)
Citation Context ...oblem A is parallel scalable if there exists a parallel scalable algorithm for it. Unfortunately, parallel scalability is not warranted for all problems, e.g., it is beyond reach for graph simulation =-=[15]-=-. The good news is as follows. Theorem 6: EIP is parallel scalable. ✷ As a proof, we outline a parallel algorithm for EIP, denoted by Matchc. Given Σ, G = (V,E, L), η and a positive integer n, it comp... |

5 | Mining top-k non-redundant association rules
- Fournier-Viger, Tseng
(Show Context)
Citation Context ...ning. There have been algorithms for pattern mining in graph databases [22,24] (see [25] for a survey). Large-scale mining techniques are also studied in a single graph [13], notably top-k algorithms =-=[16, 27, 42, 44]-=-. To reduce the cost, scalable subgraph isomorphism algorithms, e.g., [38], can be adopted to generate pattern candidates. Diversity of graph patterns is not studied there. However, (a) pattern mining... |

4 | GRAMI: frequent subgraph and pattern mining in a single large graph
- Elseidy, Abdelhamid, et al.
(Show Context)
Citation Context ...itemsets [46]. Graph pattern mining. There have been algorithms for pattern mining in graph databases [22,24] (see [25] for a survey). Large-scale mining techniques are also studied in a single graph =-=[13]-=-, notably top-k algorithms [16, 27, 42, 44]. To reduce the cost, scalable subgraph isomorphism algorithms, e.g., [38], can be adopted to generate pattern candidates. Diversity of graph patterns is not... |

4 |
Gong et al. Evolution of social-attribute networks: measurements, modeling, and implications using google
- Z
- 2012
(Show Context)
Citation Context ...mental setting. We used two real-life graphs: (a) Pokec [3], a social network with 1.63 million nodes of 269 different types, and 30.6 million edges of 11 types, such as follow, like; and (b) Google+ =-=[20]-=-, a social graph with 4 million entities of 5 types and 53.5 million links of 5 types. We also designed a generator for synthetic graphs G = (V,E, L), controlled by the numbers of nodes |V | (up to 50... |

4 | Query optimization of distributed pattern matching
- Huang, Venkatraman, et al.
- 2014
(Show Context)
Citation Context ...discovering diversified graph patterns. Graph pattern matching. Several parallel algorithms have been developed for subgraph isomorphism, e.g., [28, 37, 38], and for multi-pattern optimization, e.g., =-=[23, 32]-=-. Our algorithms for EIP differ from the prior work in the following. (a) Instead of enumerating isomorphic matches, EIP identifies a potential customer once one match is found, and moreover, computes... |

4 | Parallel processing of multiple graph queries using MapReduce. In: The fifth international conference on advances in databases, knowledge, and data applications (DBKDA
- Kim, Lee, et al.
(Show Context)
Citation Context ..., apart from [16,44]. We are not aware of prior work on discovering diversified graph patterns. Graph pattern matching. Several parallel algorithms have been developed for subgraph isomorphism, e.g., =-=[28, 37, 38]-=-, and for multi-pattern optimization, e.g., [23, 32]. Our algorithms for EIP differ from the prior work in the following. (a) Instead of enumerating isomorphic matches, EIP identifies a potential cust... |

3 | Battling predictability and overconcentration in recommender systems
- Amer-Yahia, Lakshmanan, et al.
(Show Context)
Citation Context ...The Diversified Mining Problem We are interested in GPARs for a particular event q(x, y). However, this often generates an excessive number of rules, which often pertain to the same or similar people =-=[5, 44]-=-. This motivates us to study a diversified mining problem, to discover GPARs that are both interesting and diverse. Objective function. To formalize the problem, we first define a function diff(, ) to... |

3 |
et al. Knowledge vault: A web-scale approach to probabilistic knowledge fusion
- Dong
- 2014
(Show Context)
Citation Context ...terms of distinct “potential customers” by revising a measure proposed by [7]. We propose a confidence measure for GPARs by revising Bayes Factor [31] to incorporate the local closed world assumption =-=[11,17]-=-. This allows us to cope with (incomplete) social graphs, and to identify interesting GPARs with correlated antecedent and consequent. (3) We study a new mining problem, referred to as the diversified... |

3 |
Three-objective subgraph mining using multiobjective evolutionary programming
- Shelokar, Quirin, et al.
- 2014
(Show Context)
Citation Context |

2 |
global online consumer survey. http://www.nielsen.com/content/dam/corporate/us/en/ newswire/uploads/2009/07/pr global-study 07709.pdf
- Nielsen
(Show Context)
Citation Context ...studying associations between entities in social graphs. Such associations are useful in social media marketing; indeed, “90% of customers trust peer recommendations versus 14% who trust advertising” =-=[2]-=-, and “60% of users said Twitter plays an important role in their shopping” [43]. Nonetheless, association rules for social graphs are more involved than rules for itemsets. Example 1: (1) Association... |

2 |
An NSA big graph experiment
- Burkhardt, Waring
- 2013
(Show Context)
Citation Context ...fragment; and (b) the fragments have roughly even size. These are possible since 98% of real-life patterns have radius 1, 1.8% have radius 2 [18], and the average node degree is 14.3 in social graphs =-=[8]-=-; thus Gd(vx) is typically small compared with fragment size. Fragment Fi is stored at worker Si, for i ∈ [1, n− 1]. (2) DMine discovers GPARs in parallel by following bulk synchronous processing, in ... |

2 |
Collecting and analyzing data from e-government facebook pages
- Grujic, Bogdanovic-Dinic, et al.
- 2014
(Show Context)
Citation Context ...l graphs. This is costly: graph pattern matching by subgraph isomorphism is intractable. Worse still, real-life social graphs are often big, e.g., Facebook has 13.1 billion nodes and 1 trillion links =-=[21]-=-. Contributions. This paper proposes GPARs, and provide effective algorithms for discovering and applying GPARs. (1) We introduce graph-pattern association rules (GPARs) for social media marketing (Se... |

2 |
et al. Substucture discovery in the subdue system
- Holder, Cook, et al.
- 1994
(Show Context)
Citation Context ...pattern-query processing in big graphs. (d) Mining (diversified) GPARs is beyond rule mining from itemsets [46]. Graph pattern mining. There have been algorithms for pattern mining in graph databases =-=[22,24]-=- (see [25] for a survey). Large-scale mining techniques are also studied in a single graph [13], notably top-k algorithms [16, 27, 42, 44]. To reduce the cost, scalable subgraph isomorphism algorithms... |

2 |
PGX.ISO: Parallel and efficient in-memory engine for subgraph isomorphism. GRADES
- Raman, Rest, et al.
- 2014
(Show Context)
Citation Context ..., apart from [16,44]. We are not aware of prior work on discovering diversified graph patterns. Graph pattern matching. Several parallel algorithms have been developed for subgraph isomorphism, e.g., =-=[28, 37, 38]-=-, and for multi-pattern optimization, e.g., [23, 32]. Our algorithms for EIP differ from the prior work in the following. (a) Instead of enumerating isomorphic matches, EIP identifies a potential cust... |

2 | Exploiting vertex relationships in speeding up subgraph isomorphism over large graphs
- Ren, Wang
- 2015
(Show Context)
Citation Context ...5] for a survey). Large-scale mining techniques are also studied in a single graph [13], notably top-k algorithms [16, 27, 42, 44]. To reduce the cost, scalable subgraph isomorphism algorithms, e.g., =-=[38]-=-, can be adopted to generate pattern candidates. Diversity of graph patterns is not studied there. However, (a) pattern mining over graph databases [24,27] cannot be used to mine GPARs, as their anti-... |

2 |
Twitter users say they use the site to influence their shopping decisions
- Smith
- 2013
(Show Context)
Citation Context ...useful in social media marketing; indeed, “90% of customers trust peer recommendations versus 14% who trust advertising” [2], and “60% of users said Twitter plays an important role in their shopping” =-=[43]-=-. Nonetheless, association rules for social graphs are more involved than rules for itemsets. Example 1: (1) Association rules for social graphs are defined on graphs rather on itemsets. Below is an e... |