Data Mining on Social Interaction Networks

Martin Atzmueller

doi:10.46298/jdmdh.7

Abstract

Social media and social networks have already woven themselves into the very fabric of everyday life. This results in a dramatic increase of social data capturing various relations between the users and their associated artifacts, both in online networks and the real world using ubiquitous devices. In this work, we consider social interaction networks from a data mining perspective - also with a special focus on real-world face-to-face contact networks: We combine data mining and social network analysis techniques for examining the networks in order to improve our understanding of the data, the modeled behavior, and its underlying emergent processes. Furthermore, we adapt, extend and apply known predictive data mining algorithms on social interaction networks. Additionally, we present novel methods for descriptive data mining for uncovering and extracting relations and patterns for hypothesis generation and exploration, in order to provide characteristic information about the data and networks. The presented approaches and methods aim at extracting valuable knowledge for enhancing the understanding of the respective data, and for supporting the users of the respective systems. We consider data from several social systems, like the social bookmarking system BibSonomy, the social resource sharing system flickr, and ubiquitous social systems: Specifically, we focus on data from the social conference guidance system Conferator and the social group interaction system MyGroup. This work first gives a short introduction into social interaction networks, before we describe several analysis results in the context of online social networks and real-world face-to-face contact networks. Next, we present predictive data mining methods, i.e., for localization, recommendation and link prediction. After that, we present novel descriptive data mining methods for mining communities and patterns.

Highlights

The emergence of new social systems and organizational social applications has created a number of novel social and ubiquitous environments
We focus on elements of the modeling phase, i. e., the core data mining step in Sections 4-5: The applied data mining methods can be divided into descriptive and predictive methods [48]: While descriptive methods are used for summarizing the data, for identifying hidden information in the form of patterns, and for exploration, predictive methods are used for constructing models for inferring future properties given new data, e. g., for classification: We adapt and extend known predictive data mining algorithms on social interaction networks for supporting the users in typical tasks such as recommendation and localization in the context of the mentioned systems
We focus on social interaction networks [90,91,92,93], i. e., user-related social networks in social media capturing social relations inherent in social interactions, social activities and other social phenomena which act as proxies for social userrelatedness

Summary

Introduction

The emergence of new social systems and organizational social applications has created a number of novel social and ubiquitous environments By interacting with such systems, the user is leaving traces within the different databases and log files, e. We consider a link within such a network as evidence for user relatedness and interaction and call it social interaction network This connects and transcends private and business applications featuring a range of different types of networks, organizational contexts and corresponding interactions, e. G., networks that involve spatial proximity relations like co-location or face-to-face proximity With the growth and availability of the collected data, there is an increasing interest in the analysis of such social interaction networks

Objectives

Results

Conclusion