Testing weighting approaches for forecasting in a Group Wisdom Support System environment

详细信息    查看全文

• 作者：Heiko A. von der Gracht^a ; ¹ ; ^{heikovdg@gmx.de" class="auth_mail" title="E-mail the corresponding author} ; Ulrich Hommel^b ; ² ; ^{ulrich.hommel@ebs.edu" class="auth_mail" title="E-mail the corresponding author} ; Tobias Prokesch^b ; ^{tobias.prokesch@ebs.edu" class="auth_mail" title="E-mail the corresponding author} ; Holger Wohlenberg^c ; ³ ; ^{holger.wohlenberg@deutsche-boerse.com" class="auth_mail" title="E-mail the corresponding author}
• 关键词：Forecasting ; Group wisdom ; Decision support system ; Decision making ; Collective intelligence
• 刊名：Journal of Business Research
• 出版年：2016
• 出版时间：October 2016
• 年：2016
• 卷：69
• 期：10
• 页码：4081-4094
• 全文大小：860 K

文摘

Decision makers usually seek the best possible information to support their decisions. Yet the more experts a decision maker consults, the more divergent opinions he or she might collect. In particular, the approach of attaining an adequate level of information is of crucial importance for many stakeholders such as financial and political institutions as well as sales departments. Inspired by fact that simple heuristics oftentimes outperform complex optimization models, we test and compare several simple forecast-combining methods, including multiple equally weighted approaches, an “imitate-the-successful” heuristic as well as several other weighting approaches (based on self-assessment, knowledge, and hit rate). Forecasts are collected and processed from the crowd using a novel Group Wisdom Support System (GWSS), which provides an entire forecast distribution and information on the consensus evolution over time. We find that the equally weighted triangular forecasts, a simple 1/N heuristic, delivers the most accurate results.