Abstract: Many data sets often vary dynamically in practical applications. Static attribute reduction algorithms consume a lot of time and space to solve these dynamical data sets. Since the data in set-valued decision information system are usually changed, a heuristic dynamic updating algorithm of attribute reduction is designed by introducing the concepts of conditional information quantity and significance of attribute. When new condition attributes are added to the set-valued decision information system, the proposed algorithm makes use of the attribute reduction results of the old system, updates the attribute reduction results quickly with the variation of attribute set, and deletes some redundant attributes in the new result of attribute reduction in reverse order, which keeps the knowledge concise and improves the computational efficiency. Finally, the effectiveness and feasibility of the proposed algorithm are validated by the examples.

Key words: rough sets, attribute reduction, significance of attribute, heuristic algorithm

摘要： 在现实应用中许多数据往往是动态变化的，静态的属性约简算法处理此类数据需消耗大量的计算时间和存储空间。针对集值决策信息系统中数据的动态变化情况，通过引入条件信息量和属性重要性概念，提出了一种启发式的动态属性约简算法，当新的属性集增加到决策信息系统时，算法能够利用原系统的属性约简结果，快速更新属性集增加后的属性约简，并对更新后的属性约简中可能存在的冗余属性进行反向剔除，保持了知识获取的简洁，提高了算法的计算效率。最后，通过实例验证进一步分析了算法的有效性和可行性。

关键词: 粗糙集, 属性约简, 属性重要度, 启发式算法