Abstract: Classical analysis tools are not able to satisfy this era of big data, which is full of multifarious, complicated and dynamic changed data. How to obtain valuable information from large-scale data quickly and effectively has became a challenging problem. Some scholars combined the rough set attribute reduction theory with other theories to process high-dimensional, dynamic and massive data effectively. The attribute reduction algorithms based on parallel computing, incremental learning and granular computing are classified and summarized. Then their characteristics, present problems and the key future research directions are analyzed.

Key words: big data, rough set, attribute reduction, parallel computing, incremental learning, granular computing

摘要： 在大数据时代，数据不仅类型多样、结构复杂还具有动态变化的特点，传统的分析工具已经不能满足大数据分析的需求。如何快速有效地从大规模数据中获取有价值的信息成了一个具有挑战性的问题。一些学者将粗糙集属性约简理论与其他理论相结合，从而可以有效地处理高维动态的海量数据。重点对基于并行计算、增量学习、粒计算的属性约简算法进行分类总结，分析了它们各自的特点，剖析了当前研究中存在的问题，展望了未来研究的重点关注方向。

关键词: 大数据, 粗糙集, 属性约简, 并行计算, 增量学习, 粒计算