Text mining-based categorization and user perspective analysis of environmental sustainability indicators for manufacturing and service systems

详细信息    查看全文

• 作者：Kijung Park^a ; ^{kzp5116@psu.edu" class="auth_mail" title="E-mail the corresponding author} ; Gü ; l E.Okudan Kremer^b ; ^{gkremer@iastate.edu" class="auth_mail" title="E-mail the corresponding author}
• 关键词：Environmental sustainability indicators ; Categorization ; Text mining ; Indicator utilization ; Indicator utility
• 刊名：Ecological Indicators
• 出版年：2017
• 出版时间：January 2017
• 年：2017
• 卷：72
• 期：Complete
• 页码：803-820
• 全文大小：2098 K

文摘

In response to increasing global consciousness about the environmental impact of companies, a wide variety of environmental sustainability indicators and frameworks have been developed. Despite the variety of available environmental sustainability indicators, the absence of a commonly accepted categorization framework often creates confusion and inhibits indicator deployment in practice. This paper addresses this issue with a bottom-up approach that categorizes environmental sustainability indicators using their text-based objective information, and investigates industrial perceptions on indicator use. As the foundation for this work, 55 environmental sustainability indicators were extracted from extant literature. Then, companies from manufacturing and service domain were surveyed to reveal perceptions on utilization status (i.e. used in practice and future implementation) and utility (i.e. usefulness and practicality) of each indicator. For indicator categorization, the text descriptions of the collected indicators were modeled using a text mining technique, the correlated topic model, to extract their latent topics as a basis to categorize the indicators. As a result, five categories and their relevant indicators were defined. Further, the utilization status and utility levels of the indicators within the derived categories were analyzed. Possible relationships between indicator utility levels and company characteristics were also identified through logistic regression. Utility levels of specific indicators were found to change subject to market location and industry sector. Findings from this study can complement top-down conceptual categorization and inform implementation of indicators.