Structural Positional Encoding for knowledge integration in transformer-based medical process monitoring

TL;DR

The paper addresses predictive process monitoring in medical settings, focusing on forecasting the next activity in running process traces while accounting for domain knowledge. It introduces a Transformer-based PPM framework augmented with Structural Positional Encoding, which injects ontology-derived structure via Laplacian eigenmaps into activity embeddings. Across a stroke-management dataset, SPE improves predictive accuracy at $k$-values and stabilizes performance, with gains over standard positional encoding and baseline configurations. The findings suggest that integrating domain knowledge through graph-based embeddings enhances decision-support capabilities in complex medical workflows, with potential for generalization to other clinical processes.

Abstract

Predictive process monitoring is a process mining task aimed at forecasting information about a running process trace, such as the most correct next activity to be executed. In medical domains, predictive process monitoring can provide valuable decision support in atypical and nontrivial situations. Decision support and quality assessment in medicine cannot ignore domain knowledge, in order to be grounded on all the available information (which is not limited to data) and to be really acceptable by end users. In this paper, we propose a predictive process monitoring approach relying on the use of a {\em transformer}, a deep learning architecture based on the attention mechanism. A major contribution of our work lies in the incorporation of ontological domain-specific knowledge, carried out through a graph positional encoding technique. The paper presents and discusses the encouraging experimental result we are collecting in the domain of stroke management.

Figures (4)

• Figure 1: The model architecture. The left side of the figure represents the Transformer-decoder model, the right side shows the Structural positional encoding module.
• Figure 2: Excerpt of the domain ontology
• Figure 3: Sequence length histogram.
• Figure 4: Model performance with different embedding sizes on testset.