Discovery of personalized knowledge based on rough set theory

In real world, a database usually is shared by many user, and all kind of knowledge can be take out of it. But it is not all the knowledge can meet a given user´s need. Some of them is redundant. The redundant knowledge not only is of utility to the user, but also intervenes the given user in their work. So finding just the knowledge which a given user really needs is important work in data mining. Unfortunately, it is a shortcoming in most currently existing knowledge mining algorithms that the algorithms usually generate redundant knowledge, which cannot satisfy user-defined requirements. In this paper, a concept of personalized knowledge, which can just satisfy the requirements, is given. To finding personalized knowledge, rough set theory is introduced. First, with rough set theory, and based on discernibility matrice, an other discernibility matrice, personalized discernibility matrice, is designed. By using this kind of matrice, a reduct, which includes most attributes that the user interested in, called personalized reduct, written as p_reduct, can be found. Secondly, by the way of designing threshold, redundant rules are delete. At last, we can describe a method for mining personalized knowledge, settling the existing problem to some extent.