A Review of Machine Learning and Data Mining Approaches for Business Applications in Social Networks

Introduction

Social Network Analysis (SNA) is a rapidly growing field which has received much attention in the recent years (Wasserman & Faust, 1994). The main goal of this scientific area is the study of relationships between individuals regarding their social position, the analysis of their roles, the discovery of social structures and many other issues related to social behavior. The relationships between individuals have been cast as links in huge networks and these have been traditionally constructed based on interviews and responses given by social actors. However this has always led to limited scalability of the analysis due to the lack of an infrastructure where interaction logs could be produced and saved. After this an automated approach could then be used for data collection from the interactions between individuals.

Recently, however, SNA has intensively evolved into an outstanding area not only in social sciences but also in computer science, due to the success of online social networking and media-sharing sites, and moreover, due to the availability of large repositories of social network data. With the rapid development of Internet and Web 2.0, SNA has gained even more importance which is mainly due to the combination between social media sites and social networks. The more these two combine and merge with each other, the more individuals have additional alternative ways to connect and build online relationships among them. Online social networks have boosted the capability to collect data as shown by the growing number of individuals interacting in large scale online social network platforms such as Facebook, LinkedIn, Flickr, Instant Messenger, etc. With the amount of data coming from these platforms, SNA is even more powerful because large scale networks of social entities can yield patterns that are normally not observed in small networks. With millions or even more actors in a network, it is now possible to discover patterns that can have a valuable business usage.

However, the advantage of large amounts of data that can help to perform a thorough analysis, comes together with a cost, which is the incapability of classical traditional machine learning and data mining methods to deal with this new scenario. This has given rise to challenges that are not only related to computational complexity, but also to the core methodological approaches of the learning and mining algorithms. These have to be rethought and redesign under the new perspective of the online social networking data. With the algorithmic solutions developed in the machine learning and data mining areas, it is strongly related a set of corresponding business applications. In online social networks, often every computational problem corresponds to an interesting business application and an effective algorithmic approach leads also to important developments from a business point of view. For this reason, it is essential for practitioners in the field of business related to social networks, to understand the computational challenges that lie behind the analysis of social networks.