Towards Reproducibility in Predictive Process Mining: SPICE -- A Deep Learning Library
This work addresses reproducibility issues for researchers and practitioners in predictive process mining, though it is incremental as it builds on existing methods.
The paper tackles the lack of reproducibility and comparability in Predictive Process Mining (PPM) by proposing SPICE, a Python framework that reimplements three existing deep-learning methods, resulting in a common base for robust benchmarking across 11 datasets.
In recent years, Predictive Process Mining (PPM) techniques based on artificial neural networks have evolved as a method for monitoring the future behavior of unfolding business processes and predicting Key Performance Indicators (KPIs). However, many PPM approaches often lack reproducibility, transparency in decision making, usability for incorporating novel datasets and benchmarking, making comparisons among different implementations very difficult. In this paper, we propose SPICE, a Python framework that reimplements three popular, existing baseline deep-learning-based methods for PPM in PyTorch, while designing a common base framework with rigorous configurability to enable reproducible and robust comparison of past and future modelling approaches. We compare SPICE to original reported metrics and with fair metrics on 11 datasets.