知识密集型任务的众包参与者特征建模及其绩效预测研究

Knowledge-intensive tasks are the most important task type in crowdsourcing systems. For these tasks, crowdsourcing participants are the critical elements that determine the efficiency and outcomes of tasks completion. In order to analyze participants properly and select right participants, identifing participants’ features comprehensively is a basic work, modeling their features reasonable is critical, and predicting their performance precisely is premise. Consequently, in this research, we focus on knowledge-intensive tasks, define participant’s performance, and identify the participants’ features that impact their performance by reviewing related literatures and investigating widely. On this basis, the relationships of tasks, participants’ behaviors, and their features are invetigated, and a mapping model is constructed. Furthermore, by using bayesian network theory, fuzzy cognitive maps, and data mining approaches, we model participants’ features and build their feature charts based on their participation behaviors. Finally, the relation model between participants’ features and their performance are constructed, the mechanism of participants’ features impact on their performance is discovered, and participants’ performance on a task are predicted further. The research results provide a uniform framework for analyzing participants, suggest the approaches for describing and measuring participants’ features and provide a new idea for improving the current studies on participants analysis and selection. Also, the research has practical values to increase the quality of knowledge-intensive tasks crowdsourcing.

知识密集型任务是最重要的众包对象之一。对于该类任务，众包参与者是决定任务完成效率和效果的关键。为准确分析和选择合适的众包参与者，参与者特征的系统识别是基础，特征的合理建模是关键，绩效的准确预测是前提。为此，本项目以知识密集型任务为研究对象，定义参与者绩效，通过文献分析和问卷调查等方式系统识别影响参与者绩效的个体特征因素；在此基础上，探究任务、参与者行为及其特征之间的关系，构建三者之间的映射模型，进而基于参与者行为，利用贝叶斯网络、模糊认知图、数据挖掘技术等对参与者特征建模，构建参与者特征“图谱”；基于模糊神经网络构建参与者特征与其绩效之间的关系模型，揭示参与者特征对其绩效的影响机理，进而预测参与者绩效。本项目研究成果能为众包参与者分析提供相对统一的框架，为合理刻画和度量参与者特征提供方法和途径，为进一步完善参与者分析和选择的现有研究成果提供支持，对保障知识密集型任务众包质量具有实用价值。

本项目围绕知识密集型众包中的参与者，主要研究了三部分内容：（1）众包参与者的特征提取及选择；（2）众包参与者的行为决策过程建模及其影响因素分析；（3）面向参与者的众包平台的服务特性识别及设计。研究取得的成果进一步丰富和完善了知识密集型众包中参与者个体特征分析和参与行为的建模与促进的理论研究成果，也为知识密集型众包平台更好的服务和管理参与者提供理论和方法支持。.项目期间，项目负责人和合作者共发表期刊论文16篇。其中以第一作者发表SCI/SSCI期刊论文共5篇，国家自然科学基金委管理科学部认定的A类重要期刊论文1篇，会议论文2篇；以通讯作者发表SCI/SSCI期刊论文2篇；作为其他作者合作发表SCI/SSCI期刊论文共6篇。发表的期刊包括Computers & Industrial Engineering、Behaviour & Information Technology、Electronic Commerce Research、International Journal of Technology Management、 Applied Ergonomics、管理科学等。另外有两篇论文已经完成，一篇投稿到SCI期刊Expert systems with application在二审中，另外一篇拟投稿到SCI/SSCI期刊Information Technology and management。.在项目经费的部分支持下，项目负责人分别于2019-2020年前往瑞典皇家理工学院和2021-2020年前往武汉大学信息与管理学院交流学习，讨论和解决项目开展中的相关问题，逐步明确后续的研究方向与重点。同时，项目负责人和项目主要成员均多次参加了国际/国内学术会议，加深了与国内外专家学者之间的交流和学习。.最后，在项目实施过程中，项目负责人协助培养了4名硕士研究生并于2019年毕业，目前作为导师指导3名硕士研究生，进一步共同推进项目研究工作的开展。