An incremental reduct algorithm based on generalized decision for incomplete decision tables

Attribute reduction is an important issue of data mining. In this paper an incremental reduct algorithm is proposed for incomplete decision tables. A reduct definition is firstly presented. And then based on the concept of generalized decision the different cases caused by adding a new object to an incomplete decision table are deeply analyzed and some important conclusions are proved by theorems. Finally an algorithm is proposed for incrementally computing the reducts of an incomplete decision table. An example shows that the proposed algorithm is very efficient because in many cases it can avoid recomputing the new reducts.