Explainable AI Framework for Multivariate Hydrochemical Time Series

The understanding of water quality and its underlying processes is important for the protection aquatic environments. With rare opportunity access to a domain expert, an explainable AI (XAI) framework proposed that applicable multivariate time series. XAI provides explanations are interpretable by experts. In three steps, it combines data-driven choice distance measure with supervised decision trees guided projection-based clustering. series consists measurements, including nitrate, electrical conductivity, twelve other environmental parameters. relationships between parameters investigated identifying similar days within cluster dissimilar clusters. framework, called DDS-XAI, does not depend on prior knowledge about data structure, tendentially contrastive. in can be visualized topographic map representing high-dimensional structures. Two state art XAIs eUD3.5 iterative mistake minimization (IMM) were unable provide meaningful relevant from data. DDS-XAI swiftly applied new Open-source code R all steps provided structured application-oriented.