Towards Reproducibility in Predictive Process Mining: SPICE -- A Deep Learning Library

TL;DR

This paper introduces SPICE, a Python framework that standardizes and reimplements deep learning methods for Predictive Process Mining, enhancing reproducibility, comparability, and transparency across different datasets and models.

Contribution

The paper presents SPICE, a comprehensive, configurable Python library that reimplements existing PPM methods in PyTorch for improved reproducibility and benchmarking.

Findings

SPICE achieves comparable results to original methods on benchmark datasets.

The framework enables fair comparison across different PPM approaches.

Reproducibility and transparency are significantly improved with SPICE.

Abstract

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.