Evermann, Rehse, Fettke 2017 — Deep Learning for Process Behaviour at Runtime

The first paper applying deep learning (specifically RNNs) to PPM. Explicitly positioned as an exploratory contribution — demonstrating feasibility, not final performance.

Contribution

Frames next-event prediction as sequence modelling, borrowing the NLP analogy: events are words, traces are sentences.
Uses RNNs (with embedding layer) instead of explicit process-model representations (HMMs, annotated transition systems).
Notes three structural differences between process logs and NL: smaller vocab, potentially very long traces, rule-constrained sequences — all favourable to neural approaches.

Significance

Serves as the baseline that Tax et al. 2017 (sources/2017-tax-lstm-process-prediction) explicitly compares against and improves upon with LSTM and the broader multi-task setting. Together these two 2017 papers anchor the neural-PPM era.

Connections

Concepts: concepts/predictive-process-monitoring · concepts/next-activity-prediction · concepts/lstm-ppm Authors: entities/peter-fettke · entities/joerg-evermann · entities/jana-rebecca-rehse Related sources: sources/2017-tax-lstm-process-prediction