Finding Influential Users in Social Media Using Association Rule Learning

Erlandsson, Fredrik; Bródka, Piotr; Borg, Anton; Johnson, Henric

doi:10.3390/e18050164

Cited by 75 publications

(59 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Node selection and activation, known as a seeding process [23], are focused on selection of highly influential nodes [24] and can be based on a number of approaches including selection of candidates with the highest degree of other network centrality measures with assumed high potential for information spreading [23]. More complicated solutions with high computational costs are based on greedy approach and its extensions towards optimal seed set selection [14] or data mining techniques [25]. The majority of earlier studies assumed single stage seeding as well as selection of seeds at the beginning of the process.…”

Section: Introductionmentioning

confidence: 99%

“…Its initial degree D = 6 from the beginning of the process is not taken into account. Nodes 28,29,30,25,24 are activated in natural process; (B2) stage 3 of the process with nodes 11, 6, 21, 16, 22, 19, 18 and 3 is activated in natural process and node 27 is selected as a seed. Newly selected seed activates nodes 2 and 4 with propagation probability PP = 1 and as a result all nodes in network are activated within assumed three stages and with the use of three seeds.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Dynamic Rankings for Seed Selection in Complex Networks: Balancing Costs and Coverage

Jankowski

2017

Entropy

View full text Add to dashboard Cite

Information spreading processes within the complex networks are usually initiated by a selection of highly influential nodes in accordance with the used seeding strategy. The majority of earlier studies assumed the usage of selected seeds at the beginning of the process. Our previous research revealed the advantage of using a sequence of seeds instead of a single stage approach. The current study extends sequential seeding and further improves results with the use of dynamic rankings, which are created by recalculation of network measures used for additional seed selection during the process instead of static ranking computed only once at the beginning. For calculation of network centrality measures such as degree, only non-infected nodes are taken into account. Results showed increased coverage represented by a percentage of activated nodes dependent on intervals between recalculations as well as the trade-off between outcome and computational costs. For over 90% of simulation cases, dynamic rankings with a high frequency of recalculations delivered better coverage than approaches based on static rankings.

show abstract

Section: Introductionmentioning

confidence: 99%

mentioning

confidence: 99%

Dynamic Rankings for Seed Selection in Complex Networks: Balancing Costs and Coverage

Jankowski

2017

Entropy

View full text Add to dashboard Cite

show abstract

“…In addition, the US presents a very strong influence on its diverse collaboration partners so it is always the top country in all types of centrality measurements. By following the concept using association rule mining for identifying influential actors, the LHS countries are identified as influential countries [84]. This suggests that, compared to other centrality measurements, OutDegree Centrality is a direct proxy for measuring countries' influence in international collaboration.…”

Section: Global Interdependency Of Collaboration Randdmentioning

confidence: 99%

“…"Lift" value larger than one allows understanding the positive association between LHS and RHS. The obtained association rules provide insights on how items associate with each other and how to identify influential items [84].…”

Section: Mining Association Rule For Identifying Collaboration Interdmentioning

confidence: 99%

Global Interdependence of Collaborative R&D-Typology and Association of International Co-Patenting

2017

Sustainability

View full text Add to dashboard Cite

Economic globalization implies a growing interdependence of resources across countries. Technological R&D and cross-border collaboration are often identified as the primary driving forces in the process. This study aims to holistically analyze global landscape of international collaboration and identify influential countries and the interdependencies among countries. A total of 458,381 international patents granted by the United States Patent and Trademark Office (USPTO) from 1976 to 2013 are analyzed and the structure of international collaboration network is created. It is found that highly developed and small countries usually show a higher degree of internationalization. However, emerging countries such as China present high collaborative influences. The highly skewed collaboration distribution indicates significant inequality of internationalization, which is indeed a hurdle to a country's technological mobility. It can be observed that most pairs of interdependent countries are neighboring or even bordering countries because of their similar historical, linguistic and cultural heritages. Several contributions of this study are summarized as follows: (1) this study first proposes the II, IA, II-IA, IA-AA, and II-IA-AA system for classifying international patent; (2) all international patents (38-year time span) from USPTO are examined without sampling; (3) association rule mining is used to evaluate the interdependency of international collaboration; and (4) network structures illustrating 38 years international co-patenting are visually presented.

show abstract

“…Xu et al [23] analyzed the user posting behavior on a popular social media website. Erlandsson et al [5] proposed association learning to detect relationships between users. Benevenuto et al [2] presented a kind of analysis of user workloads in online social networks.…”

Section: Introductionmentioning

confidence: 99%

Estimating the similarity of social network users based on behaviors

Nguyen

Tran

Dam

et al. 2018

Vietnam J Comput Sci

View full text Add to dashboard Cite

Recently, with the express growth of social network, users have joined more and more of these networks and live their life virtually. Consequently, they create a huge data on these social networks: their profile, interest, and behavior such as post, comment, like, joining groups or communities, etc. This brings some new challenges to researchers: do users having the same profile/interest show the same behavior? And vice versa, do users having the same behavior have interest in the same things? One of the basic issues in these challenges is the problem of estimating the similarity among users on these social networks based on their profile, interest, and behavior. This paper presents a model for estimating the similarity between users based on their behavior on social networks. The considered behaviors are activities including posting entries, liking these entries, commenting and liking the comment in these entries. The model is then evaluated with a dataset-collected users from Twitter. The results show that the model estimates correctly the similarity among users in the majority of the cases.

show abstract

Finding Influential Users in Social Media Using Association Rule Learning

Cited by 75 publications

References 36 publications

Dynamic Rankings for Seed Selection in Complex Networks: Balancing Costs and Coverage

Dynamic Rankings for Seed Selection in Complex Networks: Balancing Costs and Coverage

Global Interdependence of Collaborative R&D-Typology and Association of International Co-Patenting

Estimating the similarity of social network users based on behaviors

Contact Info

Product

Resources

About