Complex Event Processing (CEP) systems are used to combine low-level data from an input stream into high-level information. To account for workload peaks load shedding can be used to drop events. To determine when to drop events, the delay of the CEP system needs to be predicted by its workload. But if multiple operators of the CEP system share a resource, the workload of one operator does also influence the performance of the other operators. In this thesis, we examine the interference effect between multiple operators by building a prediction model. To solve this task we consider it a regression problem, where we use the arrival rate of an operator to predict the processing time of another operator on the same node. To also take into account the difference between the arrival rates of different event types, we introduce the balance score as the second input variable. Next, we design an experiment to generate diverse data. The data generated this way is then used to build prediction models by using two different methods: regression analysis and a neural network. After finding the best prediction model for each method, we compare the performance of these models. Here we show that which model is better mostly depends on the specific use case of the CEP system.