Phone: (+61 8) 6488 2238
Fax: (+61 8) 6488 1089
The Identification of Like-minded Communities on Online Social Networks
The efficient identification of communities with common interests is a key consideration in applying targeted advertising and viral marketing to online social networking sites. Existing approaches involve large-scale community detection on the entire social network before determining the interests of individuals within these communities. These approaches are both computationally intensive and may result in communities without a common interest. We propose two methods for detecting these like-minded communities using either topological or interaction links. Both methods are based on our selection algorithm for identifying users with common interests based on their following of celebrities that represent various interest categories. After identifying these users with common interests, we detect communities among them using either topological links (based on mutual friendship relationship) or interaction links (based on frequency and patterns of direct communication). Our evaluation on Twitter shows that both methods are able to detect communities comprising members that are well-connected and cohesive, and these communities become more connected and cohesive with the deepening or specialization of interest. Our proposed methods also result in communities that interact actively about their common interests (based on #hashtags and @mentions), with interaction-based method performing better than its topological-based counterpart.
Equally important in targeted advertising and viral marketing is the efficient detection of a community that is centered at an individual of interest (i.e. an influential individual). Most community detection algorithms are designed to detect all communities in the entire network graph. As such, it would be computationally intensive to first detect all communities followed by identifying communities where the individual of interest belongs to, especially for large-scale networks. We propose a community detection algorithm that directly detects the community centered at an individual of interest, without the need to first detect all communities. Our proposed algorithm utilizes an expanding ring search starting from the individual of interest as the seed user. Following which, we iteratively include users at increasing number of hops from the seed user, based on our definition of a community. This iterative step continues until no further users can be added, thus resulting in the detected community comprising the list of added users. We evaluate our algorithm on three real-life social networks and the YouTube online social network, and show that our algorithm is able to detect communities that strongly resemble the corresponding real-life communities.
With the rapid growth and proliferation of online social networking sites, many companies have embraced social media as new outlets for their targeted advertising and viral marketing efforts. For example, the Twitter social network comprises 500 million users who produce 2,200 tweets per second (as of 2012). This large user base and high user activity provide tremendous opportunities for these companies to effectively reach out to a large audience group (of potential consumers) on such online social networking sites. In turn, this audience group may further propagate information about the products/services provided by these companies. Thus, the identification of like-minded communities (comprising users with common interests in a product/service category) is essential to companies who want to utilize social media as an advertising and marketing platform.