The work described in this paper illustrates how social media is a valuable source of data which may be processed for informative knowledge discovery which may help in better decision making. We concentrate on Twitter as the source for the data to be processed. In particular, we extracted and captured tweets written in Turkish. We analyzed tweets online and real-time to determine most recent trending events, their location and time. The outcome may help predicting next hot events to be broadcasted in the news. It may also raise alert and warn people related to upcoming or ongoing disaster or an event which should be avoided, e.g., traffic jam, terror attacks, earthquake, flood, storm, fire, etc. To achieve this, a tweet may be labeled with more than one event. Named entity recognition combined with multinomial naive Bayes and stochastic gradient descent have been integrated in the process. The reported 95% success rate demonstrate the applicability and effectiveness of the proposed approach. © 2017 Association for Computing Machinery.