Abstract: In recent years, in order to ensure the family financial difficulties students successfully completed their studies, central and local governments have developed a variety of policies, including the subsidy policy for poor students in university. But will also face a problem, is how to judge whether students are poor students. There are many factors that affect the assessment of funding, leading to the existence of funding assessment is not fair. In order to help the major colleges and universities to do a better job of poor student funding work, firstly, this paper designs the questionnaire of the college students’ poverty support assessment standard, to collect data from college freshmen to seniors. Then, taking advantage of the improved rough set theory which based on the pessimistic multi-granularity information reduction algorithm, optimistic multi-granularity reduction algorithm and the two-layer absolute granularity reduction algorithm, it has excavated the key factors that affect the evaluation criteria of poor students, and verified the correctness of the results by experiments. The research results can make the poverty assessment standard more just, so that poor students can better finish their studies under the assistance of the state, and lay a good foundation for future life.

Key words: multi granularity rough sets, college students poverty aid, absolution reduction, particle size reduction, amount of information

摘要： 近年来，为保证家庭经济困难学生顺利完成学业，中央和地方各级政府制定了各种各样的政策，其中包括了大学贫困生资助这项政策。但同时也面临一个问题，就是如何判断学生是否贫困生。由于影响资助评定的因素有很多，导致资助评定存在不公平性。为了帮助各大高校更好地做好贫困生资助工作，设计了大学生贫困资助评定标准调查问卷，向某校大一至大四学生分发调查问卷收集数据，利用粗糙集理论改进的基于悲观多粒度约简算法、乐观多粒度约简算法以及双层绝对粒度约简算法，挖掘出影响贫困生评定标准的关键因素，并通过实验验证了结果的正确性。研究成果使得贫困生评定标准更加公正，让贫困生能在国家的资助下更好地完成学业，为以后的人生打好基石。

关键词: 多粒度粗糙集, 大学生贫困资助, 绝对约简, 粒度约简, 信息量