Abstract:

Abstract Data mining is the finest technique for extracting knowledge from patient routes. Medical occurrences are intricately organized in event logs and frequently documented using several medical codes. Before applying process mining analysis, labeling these occurrences properly is challenging. This study presents a new method for managing complex events in medical records. Improved deep auto-encoding (IDAE) generates precise labels by grouping similar events in latent space. Also, an explanation is given by decoding the instances that correspond to the generated labels. When tested on simulated events, the method successfully uncovered hidden clusters in sparse binary data and provided precise justification for created labels. Real medical data are used in a case study. The outcomes support the theory's effectiveness in knowledge extraction from complicated event logs depicting patient pathways.

Keywords: