Twitter Association Rule Mining using Clustering and Graph Databases

In the actual scenario, the need to efficiently analyze this kind of data is increasing because of characteristics of such big data, especially their huge and sometimes unpredictable variety. Twitter alone, with 320 M active users every month and more than 500 M tweets per day, could represent an important source of information. For this research, we are focusing solely on social networks. The reason for this choice is that they are increasingly becoming a platform where people will comfortably update their status and share or retrieve information about the world in real time. Often news is spreading through them faster than in traditional channels because user capillarity worldwide makes it possible. In particular, we will focus on Twitter, because its micro-blogging nature makes it suitable for this kind of purpose. It questions the concept of a small private community of friends in favor of less private, less personal broadcast communications of common interest. Another reason why we chose Twitter is because semantic value of hashtags, their power in summarizing tweet content and the spreading model through the social network that allows us to highlight clusters of topics by focusing on these tags. One of the objectives of this thesis is to show how data mining can provide useful techniques to deal with these huge datasets for retrieving information to detect and analyze trending topics and the corresponding user’s interactions with them. We identified in Association Rules identification and evolution in time, a systematic approach to conduct the analysis.