Complex Event Processing (CEP) is becoming more and more popular in service-oriented practices, especially to monitor the behaviour of continuous tasks within manual business processes, such as in logistics. The inference mechanisms of CEP engines are completely guided by rules, which are specified manually by domain experts. We argue that this user-based rule specification is a limiting factor that complicates the integration of CEP within the realm of Business Process Management (BPM) in a seamless way. Therefore, we present autoCEP as a two-phase data mining-based approach that automatically learns CEP rules from historical traces. In the first phase, complex temporal patterns are learned using early classification on time series techniques, then these patterns are algorithmically transformed into CEP rules in the second phase. Satisfactory results from evaluations on real data demonstrate the effectiveness of our framework.