BEST: Bilaterally Expanding Subtrace Tree for Event Sequence Prediction

In Predictive Process Monitoring, handling uncertainty regarding future case execution is the core building block for reliable predictive or prescriptive methods. In the last decade, deep learning methods are increasingly the preferred approach when it comes to Next Activity Prediction and/or Remaining Trace Prediction. However, it remains an open question whether deep learning models finally surpass traditional data mining techniques for these tasks. In our paper, we contribute to answering this question by proposing a sequence prediction framework based on bilaterally expanding hierarchical subtraces that serves as an alternative approach for currently established deep learning techniques. We mine sequential patterns from activity traces and arrange them into a hierarchical subtrace tree by their structural relationship and inter-pattern distances. The tree structure can directly be leveraged for forecasting the most probable future activities given the trace history. We achieve competitive forecasting results for Remaining Trace Prediction, even surpassing state-of-the-art deep learning approaches on the majority of the analyzed real-world benchmark process event logs while only relying on the available control-flow information.