Kablosuz Sosyal Sensör Ağlar

Online Social Networks (OSNs) have recently become the essential means of communication, networking and entertainment. One of the prominent applications of OSNs has originated from their frequent use during major events, known as social sensing, which is the utilization of the information shared in OSNs to estimate an observed yet unknown phenomenon. In this thesis, we analytically investigate social sensing capabilities of OSNs. To this end, we introduce Wireless Sensor Network (WSN) paradigm, Wireless Social Sensor Network (WSSN) and explore the WSSNs within the most widely used OSNs, i.e., Twitter and Facebook. First, we develop communication theoretical models for the mechanisms of information propagation in Twitter and then analytically model the social sensing with Twitter. The accuracy of the estimated signal is investigated with mean square error analysis. Later, using a simple observation model by considering the features of Twitter, i.e., tweet and retweet, we extend the performance analysis (in terms of mean square error) of social sensing by comparing with fundamental estimators in classical and Bayesian estimation theory for various factors such as user behavior, number of people participating to social sensing and geotag use percentage. Lastly, we model and investigate the main social sensing mechanism in Facebook, i.e., Facebook Comment Thread Network (FCTN). By developing an analytical model for user observations in CTN, we analyze the reliability of social sensing with Facebook CTN for varying user behaviors and relationships, event characteristics, Facebook features and network size. The results indicate that, in addition to network conditions, the reliability of social sensing, i.e., the accuracy of the estimated signal, is affected by the features of the OSN, user behavior patterns and source event characteristics.