Towards Ontology-based Explainable Classification of Rare Events

Towards Ontology-based Explainable Classification of Rare Events

Working paper. In The open archive HAL, 2019.

Abstract:

Rare events (e.g. major floods, violent conflicts) are events that have potentially widespread and/or disastrous impact on society. The overall goal is to build a framework capable to classify, predict and explain such rare events. To do so, we envisage the usage of a mixture of sub-symbolic Machine Learning (ML) and Ontology-based Statistical Relatio-nal Learning (OSRL) techniques to generate rare events classifiers and predictors, which additionally may be mapped into natural language to ease human interpretability of the decision process.