一种面向用户需求的属性约简算法

摘要

In existing attribute reduction algorithms, attributes in an attribute set are assumed to be equally important. However, the assumption is unreasonable and impractical for some practical applications, since the importance of attributes is usually related to demands or preferences which are different from one to another. The results of many other user-oriented algorithms fail to provide the exact match to needs. In this paper, considering the attribute order of users’demands and simplifying attribute reduction into set cover reduction, an user-oriented attribute reduction algorithm is proposed to get a minimum attribute reduction to satisfy users’demands or preferences. As a result, the proposed algorithm is proved to be feasible and effective.%在已有的属性约简算法中，一般假定属性集中的属性同等重要。然而，对于实际问题，这种假定既不合理也不实际，因为属性的重要性往往与用户的需求相关。许多已经提出的面向用户需求的学习算法给出的结果不能保证与用户的需求完全匹配。将描述用户需求的属性序纳入考虑，并将属性约简问题转化为集合覆盖的约简问题求解，提出一种面向用户需求的属性约简算法，旨在获得满足用户需求或偏好的最小属性约简。理论分析、实验和实例显示，算法可行且有效。