Social Networking: Mining, Visualization, and Security
Format: PDF / Kindle (mobi) / ePub
With the proliferation of social media and on-line communities in networked world a large gamut of data has been collected and stored in databases. The rate at which such data is stored is growing at a phenomenal rate and pushing the classical methods of data analysis to their limits. This book presents an integrated framework of recent empirical and theoretical research on social network analysis based on a wide range of techniques from various disciplines like data mining, social sciences, mathematics, statistics, physics, network science, machine learning with visualization techniques and security. The book illustrates the potential of multi-disciplinary techniques in various real life problems and intends to motivate researchers in social network analysis to design more effective tools by integrating swarm intelligence and data mining.
optimally choose the initial active nodes so as to maximize the range of information diffusion. For instance, the problem of viral marketing wherein recommendation by the top k influential nodes (k is given) resulting in the largest cascade is studied in . They show that their method signicantly out-performs heuristic methods such as degree centrality and closeness centrality. It is believed that the most influential node has the highest value of network centrality such as betweenness
is essential in a complete analysis [71, 72]. Wrapper based approach by Brodley and Friedl  identifies and eliminates mislabelled instances to improve quality of training data. In first step, they try to learn candidate’s instances using “m learning algorithms” to tag correctly or incorrectly labelled instances. Then they build a classifier to remove mislabelled instances. Filtering can be based on one or more of the m base level classifiers’ tags.  highlights instance selection mechanism
Facebook’s privacy model can be adapted. In social networks, people subscribe to their likes such as news, blog etc.; join groups such as group of certain music band, fan club of a certain person, etc. These activities highlight one’s involvement and interests. Zheleva et al.  used such important subscription based information and using the concept of homogeneity within a group, they inferred hidden attribute values of actors. They proposed several methods, and the method using Support
(Select), collecting those members into possible subgroups (Collect) and choosing the cohesive subgroups over time (Choose). Social network analysis, clustering and partitioning, and similarity measurement are then used to implement each of the steps. The Social Cohesion Analysis of Networks (SCAN)  method was developed for automatically identifying subgroups of people in social networks that are cohesive over time. The SCAN method is to be applied based on the premise that a social graph can
sources of livelihood to the villagers in all the selected villages, the latter is particularly the major source in all the villages (Table 1: Panel C). Our summary data regarding the villages also show the presence of different categories of Hindu castes and other communities in each village. Again, Hindu high and middle caste households own most of the land in all the villages except Raspur where the low castes, tribals and Muslims own major share. We can now discuss to explain how these