The use of information granules to detect anomalies in spatial behavior of animals

Abstract

Nowadays, in the era of automation of ecological measurements, more and more often we are dealing with large data sets in which various unexpected anomalies may occur. Their detection is often crucial for a proper assessment of ecological trends and processes. Therefore, methods allowing for identification of anomalous data are especially important for a deep understanding of ecological phenomena and their relationships in practical domains. In this study, we present an innovative application of information granules to the in-depth study of spatial behavior of the European bison (Bison bonasus), based on GPS data. As evidenced by a series of numerical experiments, this granular computing-based approach allows to detect both anomalies and regularities in the atypical behavior of the European bison, a species important for local ecosystems. The transformation of the original data space into a new semantic multidimensional space, defining the degree of membership in the anomaly class represented by new descriptors allows for more unambiguous analysis of non-standard animal behavior.

Keywords

• Fuzzy numbers
• Information granules
• Outlier detection
• Data transformation
• European bison
• Telemetry data