Deep Generative Models in Process Mining

Several deep learning approaches have been proposed to address tasks in process mining, but little attention has been given to how events and traces are represented. In this thesis, we explore the use of deep generative models ---particularly autoencoders and their variants--- to learn process-aware representations that capture the contextual and structural dependencies in event logs. We proposed three novel architectures for generating embeddings, imputing missing values, and detecting anomalies, and validate them through extensive experiments on real-world datasets. The results show that these models outperform existing methods and offer robust, generalizable solutions for improving process mining tasks. Additionally, we present a benchamarking frameworks to ensure fair comparisons and support reproducible research.

Palabras clave: Process mining, Deep Learning, Deep Generative Models