众包是一种通过网络平台汇聚群体智慧求解问题的一种新兴商业模式。与质量相关的一系列决策直接影响到众包任务成败及众包商业价值的发挥，因此如何有效提高众包运营质量成为当前众包领域的迫切需求。针对已有研究中考虑工人的异质性、工人与任务的依赖性不足，项目提炼出动态环境下众包工人选择及任务分配等关键问题。（1）结合线下线上信息对工人质量进行动态更新，提出基于贝叶斯自适应学习的工人选择模型。（2）基于部分可观测的马尔科夫决策模型，构建工人动态筛选的优化模型，用于主动识别及剔除低质量的工人。（3）考虑工人去向、工人质量及任务完成程度等动态反馈信息，建立众包任务的在线分配决策模型。对所提出的模型设计求解算法，并通过实际数据及模拟数据验证模型的有效性。研究结果丰富了基于数据驱动的决策模型和方法。项目预期在众包决策领域取得实质进展，为解决众包的质量控制与决策问题提供借鉴。

Crowdsourcing is an emerging business mode that gathers swarm intelligence to solve various kinds of problems through network platforms. A series of quality related decisions directly affect the success or failure of the crowdsourced task and the realization of the commercial value for crowdsourcing. Therefore, how to improve the quality effectively in crowdsourced operation process has become a hot topic being discussed nowadays in the academic field. The existing literature did not fully address the heterogeneity of workers, the dependency between workers and tasks. Based on these shortcomings, this project exploits some of the key issues such as worker selection and task assignment. The main contributions are summarized in the following. (1) The worker quality is dynamically updated based on the available information of both online and offline. A Bayesian adaptive learning based worker selection model is then proposed. (2) Based on the partially observable Markov decision model, an optimization model for dynamic screening workers is constructed to actively identify and eliminate low-quality workers. (3) Considering the dynamic feedback information on workers' whereabouts, worker quality and task completion degree, the dynamic assignment model for crowdsourcing task is established. For each of the proposed decision models, corresponding algorithms are designed, and these models and algorithms are validated by both the realistic data sets and simulated data sets. The results of this study enrich the research of data-driven decision-making models and methods. It is expected to make substantial progress in the field of crowdsourcing decision making, and provide new insights for solving the quality control and decision problems in crowdsourcing.