International Journal of Social Network Mining (3 papers in press)
Research Contributions Published on Betweenness Centrality Algorithm: Modelling to Analysis in the Context of Social Networking
by Sagar S. De, Satchidananda Dehuri, Sung-Bae Cho
Abstract: Social network analysis has become an inevitable tool for the prosperity of modern civilization. The process of accumulating relational information from structured/unstructured sources, modeling networks, and extracting actionable information requires expertizing in several knowledge domains. This paper presents an approach for the analysis of documents in the context of social networking. The approach is illustrated by using a case study related to research contributions published on Betweenness Centrality algorithm. Distinct networks in terms of article, article-author, and author are modeled and analyzed to understand the insights. Consequently, it is possible to identify crucial articles, active authors, groups along with their expertise, research directions, the correlation among documents, and many more. Thus the paper conferred techniques for document collection, preprocessing, network modeling, and network analysis methods for the directed, undirected, weighted, unweighted, connected, disconnected, and bipartite networks.
Keywords: Betweenness Centrality; Bibliographic Data Collection; Document Labelling; Network Modeling; Network Analysis Methods; Preferential Attachment; Two-mode Analysis; Cluster Analysis.
Action Rules for Sentiment Analysis using Twitter
by Angelina Tzacheva, Jaishree Ranganathan, Arunkumar Bagavathi
Abstract: Actionable patterns are interesting and usable knowledge mined from large datasets. Action rules are rules that describe the possible transition of objects from one state to another with respect to a decision attribute. In this work, we extract actionable recommendations in the form of action rules, that can be applied to social networking data in a scalable manner to achieve the desired user goals. We propose extraction of actionable patterns based on Sentiment Analysis of social network data. In Sentiment Analysis there are two approaches to determine the polarity: Corpus based, and Lexicon based. We use corpus-based Sentiment Analysis approach, where sentiment values are generated based on sentence structure rather than words. Results show actionable patterns in tweet data and provide suggestions on how to change sentiment of the tweet to a more positive one. The experiment is performed with Twitter data in a distributed environment using Hadoop MapReduce for scalability with large data.
Keywords: Sentiment Analysis;Natural Language Processing;Action Rules;Meta-Actions;MapReduce.
Spatial patterns of the French rail strikes from social networks using Weighted k-Nearest Neighbour
by OUARET Rachid, BIRREGAH Babiga, JAAFOR Omar
Abstract: The information analysis provided by millions of social network users is one of the most important sources of information yielding interesting insights of spatial patterns about Socio-Political events. rnDuring the recent French National Railway Strikes (from April to June), Twitter was used as platform where people expressed their opinions, with millions of “SNCF” tweets posted over the Strike period.rnIn this paper, we have discussed a methodology which allows the utilization and interpretation of Twitter data to determine spatial patterns over French territory. The identification of a geographic strike landscape is achieved through spatial interpolation using Weighted k-Nearest Neighbour. This study shows the benefits of geo-statistical learning for extracting sentiment polarities of social events across France.
Keywords: Weighted k-Nearest-Neighbor; spatial interpolation; Twitter; social networks; railway network; social polarities; sentiments analysis