DMDH 2025 Abstracts

Area 1 - DMDH

Full Papers

Paper Nr:	6
Title:	Process Mining and Machine Learning for Predicting Clinical Outcomes in Emergency Care: A Study on the MIMICEL Dataset
Authors:	Antonella Madau and Gianfranco Semeraro
Abstract:	The digitization of organizations and the increasing availability of data generated by Information Systems (IS) have led to the development of advanced techniques for business process improvement. Process Mining has emerged as a key discipline bridging the gap between Data Science and Business Process Management (BPM). In this study, we explore the application of classification techniques on the MIMIC-IV-ED dataset, which records patient-level event logs during their stay in the emergency department. The proposed approach starts with process mining to uncover underlying care pathways, followed by thorough data pre-processing and cleaning to construct a structured dataset suitable for classification tasks. In the final stage, we evaluate the performance of seven classification algorithms, encompassing both tree-based and boosting methods, to predict relevant clinical or operational outcomes. Our methodology highlights the synergy between process mining and machine learning, offering insights into patient flow and decision support in emergency care settings.

Paper Nr:	8
Title:	AMAKAN: Fully Interpretable Adaptive Multiscale Attention Through Kolmogorov-Arnold Networks
Authors:	Felice Franchini and Stefano Galantucci
Abstract:	This paper introduces AMAKAN, a novel method for tabular data classification combining the Adaptive Multi-scale Deep Neural Network with Kolmogorov–Arnold Network to ensure full interpretability without sacrificing predictive performance. The Adaptive Multiscale Deep Neural Network dynamically focuses on relevant features at different scales by using learned attention mechanisms. These multiscale features are then refined by Kolmogorov–Arnold Network layers, which replace typical dense layers with learnable univariate functions on network edges, providing transparency by allowing practitioners to visually see and inspect feature transformations directly. Experimental results on a variety of real-world datasets demonstrate that AMAKAN achieves performance equivalent to or better than state-of-the-art baselines while providing transparent and actionable explanations for its predictions. By the seamless combination of interpretable attention mechanisms with Kolmogorov–Arnold Network layers, the paper presents an explainable and efficient deep learning method for tabular data across a vast spectrum of application domains.

Short Papers

Paper Nr:	7
Title:	Explainable AI Approach for Cardiac Involvement Detection in Anderson-Fabry Disease
Authors:	Chiara Verdone, Matteo Gravina, Grazia Casavecchia, Rodolfo Belfiore and Benedetta Di Millo
Abstract:	Anderson-Fabry Disease (AFD) is a rare X-linked hereditary disorder caused by a deficiency of the enzyme alpha-galactosidase A, leading to the accumulation of globotriaosylceramide (Gb3) in multiple organs, including kidneys and the cardiovascular system. This study explores the role of deep learning techniques in the analysis of cardiac imaging data for the early detection and monitoring of AFD-related cardiac involvement. Using advanced image processing algorithms, we aim to improve diagnostic accuracy, assess myocardial fibrosis progression, and facilitate personalized patient management. Our findings highlight the potential of artificial intelligence in enhancing diagnostic workflows, reducing variability in interpretation, and aiding clinicians in making more informed decisions. Furthermore, the use of non-invasive imaging techniques and Native T1 sequences for mapping studies in cardiac magnetic resonance imaging (CMR) could reduce the need for contrast.