光华讲坛——社会名流与企业家论坛第 5633 期

主题：Gender classification of product reviewers in china: a data driven approach

主讲人：俄勒冈州立大学 朱斌 副教授

主持人：经济信息工程学院 郑海超 教授

时间：2019年11月29日（星期五）上午10：00

地点：西南财经大学柳林校区格致楼J311B

主办单位：经济信息工程学院  金融智能与金融工程四川省重点实验室  科研处

主讲人简介：

Dr. Bin Zhu is an Associate Professor and the program director of Business Analytics in the College of Business. Prior to OSU she was an assistant professor at Boston University. She earned her Ph.D. in Management Information Systems from University of Arizona. Her current research interests include business intelligence, information analysis, social network, human-computer interaction, information visualization, computer-mediated communication, and knowledge management systems. She has been a lead author for papers that have appeared in Information Systems Research, Decision Support Systems, Journal of the American Society for Information Science and Technology, IEEE Transaction on Image Processing, and D-Lib Magazine. Her research also received an IBM faculty award.  Her teaching interests are business intelligence; database analysis and design; telecommunication; web technology; business programming; data structure and algorithms; e-commerce; information security/assurance; management information systems.

朱斌博士是俄勒冈州立大学商学院的副教授和商业分析专业项目主任。在此之前，她曾是波士顿大学的助理教授。她在亚利桑那大学获得了管理信息系统博士学位。目前的研究方向包括商业智能、信息分析、社交网络、人机交互、信息可视化、计算机沟通和知识管理系统。她作为主要作者在 Information Systems Research、Decision Support Systems、Journal of the American Society for Information Science and Technology、IEEE Transaction on Image Processing、and D-Lib Magazine等期刊上发表过论文。研究成果获得过IBM教员奖。教学兴趣包括商业智能、数据库分析与设计、电信、网络技术、程序设计、数据结构和算法、电子商务、信息安全/保障、管理信息系统。

内容提要：

While it is crucial for organizations to automatically identify the gender of participants in product discussion forums, they may have difficulties adopting existing gender classification methods because the performance of a classification method is highly contextual, given that the discriminative power of gender features used by a classification method varies with context. This paper proposes and validates a framework to develop a classification method that uses a more “data-driven” approach to accommodate the contextual changes. We demonstrated that in addition to optimizing a gender classification method, its performance can also be improved by optimizing the way in which it is applied to the archived data of online product discussion forums. Our study also indicates that for any given online discussion forum data and a given classification method, the classification accuracy varies with the size of input data. And there is an optimal input data size to achieve highest accuracy. This is different from the commonly accepted assumption that larger data size always leads to better classification performance.

对于企业而言，自动识别产品论坛中参与者的性别至关重要。但是，目前企业在应用性别分类方法过程中仍存在困难，一个很大的因素是分类模型高度情景化，即不同情境中识别性别的特征有很大差异。主要提出并验证了一个框架，来开发“数据驱动”的、适应情境变化的分类方法。除了优化性别分类方法之外，还可以通过调整其应用于在线产品论坛数据的方式来提高模型的性能。对于任何给定的在线论坛数据和给定的分类方法，分类准确性会随着输入数据的大小而变化。存在一个最佳的输入数据样本大小来后的最高的准确性。这与我们认为的“更大的数据大小总是导致更好的分类性能”这一常识不一致。