Case–Based Ranking for Environmental Risk Assessment

Abstract

Abstract. The assessment of environmental risk may be conceived as a process of ranking among alternative scenarios. First principle approach tends to define in advance the general criteria of risk assessment which not necessarily discriminate a meaningful priority relation afterwards. Moreover ex-ante methods relies on the tacit assumption of data homogeneity for describing risk scenarios. In this work we propose a method for risk assessment where the process of ranking is driven by cases. Steps of pairwise case analysis for rank elicitation are interleaved with steps of computational learning to estimate the ranks for unseen cases. The case-based approach is twofold: on one hand supporting the users to elicit their knowledge by examples, on the other hand enabling an approximation of target rank looking at case descriptions. A case study of environmental risk assessment in a northern region of Italy illustrates how the case-based ranking method works in practice. A second case study in a eastern region of Italy shows how the reference method can be extended to support a collaborative and distributed assessment of environmental risk.