Exploring Heterogeneous Sequential Data on River Networks with Relational Concept Analysis

Abstract

Nowadays, many heterogeneous relational data are stored in databases to be further explored for discovering meaningful patterns. Such databases exist in various domains and we focus here on river monitoring. In this paper, a limited number of river sites that make up a river network (seen as a directed graph) is given. Periodically, for each river site three types of data are collected. Our aim is to reveal user-friendly results for visualising the intrinsic structure of these data. To that end, we present an approach for exploring heterogeneous sequential data using Relational Concept Analysis. The main objective is to enhance the evaluation step by extracting heterogeneous closed partially-ordered patterns organised into a hierarchy. The experiments and qualitative interpretations show that our method outputs instructive results for the hydro-ecological domain.