作者: Valentina Prado , Ben A. Wender , Thomas P. Seager
DOI: 10.1007/S11367-017-1281-3
关键词: Machine learning 、 Stakeholder 、 Decision support system 、 Normalization (statistics) 、 Probability distribution 、 Artificial intelligence 、 Comparative life cycle assessment 、 Data mining 、 Mathematics 、 Alternative methods
摘要: Identification of environmentally preferable alternatives in a comparative life cycle assessment (LCA) can be challenging the presence multiple incommensurate indicators. To make problem more manageable, some LCA practitioners apply external normalization to find those indicators that contribute most their respective environmental impact categories. However, cases, these results entirely driven by reference, rather than performance alternatives. This study evaluates influence methods on interpretation facilitate use decision-driven applications and inform latent systematic biases. An alternative method based significance mutual differences is proposed instead. paper performs evaluation describes an called overlap area approach for purpose identifying relevant issues LCA. The utilizes probability distributions characterized assess significant differences. effects three LCIA methods, through application four studies. For each application, we call attention category highlighted approach. External suffers from bias emphasizes same categories regardless application. Consequently, studies employ guide selection may result recommendations dominated reference insensitive data uncertainty. Conversely, via calls with between does not show across because it depend references sensitive changes Thus, decisions will draw tradeoffs alternatives, highlight role stakeholder weights, generate assessments are responsive solution LCAs this different algorithm capable evaluating integrating uncertainty results.