The increasing amount of IoT sensor event data in domains like smart cities or logistics demands efficient analysis techniques that give valuable and new insights from data. Process mining promises to discover valuable knowledge from data. However, the event data tracked from IoT devices is at a much lower level, on which a direct application of process mining yields few insights. This paper suggests an approach for process discovery on sensor location event data in single occupation settings by leveraging unsupervised learning in the form of clustering to raise the abstraction level of events. The main contribution is a feedback loop with a sensitivity analysis that measures how sensitive the discovered process model is to changes when processing and aggregating IoT sensor location event data in a (semi)automated manner. Our evaluation results show that the feedback loop automatically improves model quality considerably and can obtain high-level processes from sensor location event data.