Abstract: We present an incremental rough set inductive learning approach to two classes of decision systems. Firstly, based on rough set and user-require, we also propose an algorithm of feature reduction to reduce the decision table, such that a reduced decision table is generated. Secondly, we modifies definitions of decision matrix and decision function and a method based on modified decision matrix and decision function is introduced to produce the decision rules from the reduced decision table. Finally, An example is also illustrated in the paper, and proves that the approach is very effective.