The complex event processing (CEP) paradigm defines a set of systems that have the goal of finding important patterns in a streams of low-level events. They combine them in order to form informationwith a higher level of abstraction.These systems are susceptible to fluctuation in the workload of the event streams. Load peaks can lead to an increased latency which is undesirable in realtime scenarios. On limited hardware nodes this effect can be mitigated by using load shedding i.e. dropping a portion of the incoming events. The latency is also influenced by other operators sharing the same physical node. This thesis proposes a workflow which reduces processing time and therefore the latency by using this interference effect. For this purpose a neural network is trained. It is used for constructing a lookup table containing target system states leading to a specific reduction in processing time. These states are then translated into shedding configurations. While the implemented solution has its limitations, the result strongly indicates that the handling of the interference effect can be used to reduce processing times.