Understanding co-evolution in large multi-relational social networks

Abstract

Understanding dynamics of evolution in large social networks is an important problem. In this paper, we characterize evolution in large multi-relational social networks. The proliferation of online media such as Twitter, Facebook, Orkut and MMORPGs 1 have created social networking data at an unprecedented scale. Sony's Everquest 2 is one such example. We used game multi-relational networks to reveal the dynamics of evolution in a multi-relational setting by macroscopic study of the game network. Macroscopic analysis involves fragmenting the network into smaller portions for studying the dynamics within these sub-networks, referred to as 'communities'. From an evolutionary perspective of multi-relational network analysis, we have made the following contributions. Specifically, we formulated and analyzed various metrics to capture evolutionary properties of networks. We find that co-evolution rates in trust based 'communities' are approximately 60% higher than the trade based 'communities'. We also find that the trust and trade connections within the 'communities' reduce as their size increases. Finally, we study the interrelation between the dynamics of trade and trust within 'communities' and find interesting results about the precursor relationship between the trade and the trust dynamics within the 'communities'.

Using coherencies to examine network evolution and co-evolution

“…Their focus is on community or group evolution rather than the evolutionary changes in the entire network and the relations among the changes of individual nodes. Similarly, Singhal et al ( 2014 ) examined the evolution and coevolution of communities in a multi-relational (trade and trust) network based on their connectivity. Toivonen et al ( 2009 ) compared two different models of network evolution, NEMs, in which the addition of new links is dependent on the (typically local) network structure and NAMs whose links are generated based only on nodal attributes.…”

Using Gaussian Measures for Efficient Constraint Based Clustering

“…Data mining is the process of finding hidden patterns and trends in databases and using that information to do a variety of tasks such as finding association rules, clustering heterogeneous groups of information and build predictive models [9]. It can also be considered as an important tool for data segmentation, selection, exploration and building models using the vast data stores to discover previously unknown patterns in various domains such as healthcare [12] [21], [23], [20] [19], social media analysis [18], [24], [2], [16], [11], [15], [17], finances and various other domains [22]. From the past few decades, data mining has been used extensively in various areas of decision making and decision analysis by financial institutions, for credit scoring and fraud detection; marketers, for direct marketing and cross-selling or up-selling; retailers, for market segmentation and store layout; and manufacturers, for quality control and maintenance scheduling.…”