1 Human Systems Engineering, Arizona State University Polytechnic , Mesa, Arizona.
2 School of Computing, Informatics, and Decision Systems Engineering, Arizona State University , Tempe, Arizona.
Big Data. 2017 Mar;5(1):53-66. doi: 10.1089/big.2016.0044. Epub 2017 Mar 10.
Historically, domains such as business intelligence would require a single analyst to engage with data, develop a model, answer operational questions, and predict future behaviors. However, as the problems and domains become more complex, organizations are employing teams of analysts to explore and model data to generate knowledge. Furthermore, given the rapid increase in data collection, organizations are struggling to develop practices for intelligence analysis in the era of big data. Currently, a variety of machine learning and data mining techniques are available to model data and to generate insights and predictions, and developments in the field of visual analytics have focused on how to effectively link data mining algorithms with interactive visuals to enable analysts to explore, understand, and interact with data and data models. Although studies have explored the role of single analysts in the visual analytics pipeline, little work has explored the role of teamwork and visual analytics in the analysis of big data. In this article, we present an experiment integrating statistical models, visual analytics techniques, and user experiments to study the role of teamwork in predictive analytics. We frame our experiment around the analysis of social media data for box office prediction problems and compare the prediction performance of teams, groups, and individuals. Our results indicate that a team's performance is mediated by the team's characteristics such as openness of individual members to others' positions and the type of planning that goes into the team's analysis. These findings have important implications for how organizations should create teams in order to make effective use of information from their analytic models.
从历史上看，商业智能等领域需要一位分析师来处理数据、开发模型、回答运营问题并预测未来行为。然而，随着问题和领域变得更加复杂，组织开始雇佣分析师团队来探索和建模数据，以生成知识。此外，鉴于数据收集的快速增长，组织正在努力开发大数据时代的情报分析实践。目前，有多种机器学习和数据挖掘技术可用于对数据进行建模，并生成见解和预测，而可视化分析领域的发展则侧重于如何有效地将数据挖掘算法与交互式视觉效果联系起来，以使分析师能够探索、理解和交互数据和数据模型。尽管已经有研究探讨了单个分析师在可视化分析管道中的作用，但很少有研究探讨团队合作和可视化分析在大数据分析中的作用。在本文中，我们提出了一个集成统计模型、可视化分析技术和用户实验的实验，以研究团队合作在预测分析中的作用。我们将实验围绕社交媒体数据的分析展开，以解决票房预测问题，并比较团队、群体和个人的预测性能。我们的结果表明，团队的表现受到团队成员对他人立场的开放性以及团队分析中计划类型等团队特征的影响。这些发现对于组织应该如何创建团队以有效利用其分析模型中的信息具有重要意义。
