Complex Event Processing (CEP) is emerging as a new paradigm for continuous processing of streaming data in order to detect relevant information and provide support for timely reactions. The main role of a CEP engine is to detect the occurrence of event patterns on the incoming streaming data. However, the problem of discovering the event patterns, although strongly related to the data mining field, has not been studied from the perspective of CEP applications. This paper presents the first steps towards defining a framework that would allow seamless integration of CEP and data mining method. We present the smart cities scenarios as a good working-field for experimentation. A concrete use case is discussed and preliminary results are presented for real-live data that has been collected.