Recurrent process mining with live event data

In organizations, process mining activities are typically performed in a recurrent fashion, e.g. once a week, an event log is extracted from the information systems and a process mining tool is used to analyze the process’ characteristics. Typically, process mining tools import the data from a file-based source in a pre-processing step, followed by an actual process discovery step over the pre-processed data in order to present results to the analyst. As the amount of event data grows over time, these tools take more and more time to do pre-processing and all this time, the business analyst has to wait for the tool to finish. In this paper, we consider the problem of recurrent process discovery in live environments, i.e. in environments where event data can be extracted from information systems near real time. We present a method that pre-processes each event when it is being generated, so that the business analyst has the pre-processed data at his/her disposal when starting the analysis. To this end, we define a notion of intermediate structure between the underlying data and the layer where the actual mining is performed. This intermediate structure is kept in a persistent storage and is kept live under updates. Using a state of the art process mining technique, we show the feasibility of our approach. Our work is implemented in the process mining tool ProM using a relational database system as our persistent storage. Experiments are presented on real-life event data to compare the performance of the proposed approach with the state of the art.