A novel incremental attribute reduction approach for incomplete decision systems

With the increasing diversification and complexity of information, it is vital to mine effective knowledge from information systems. In order to extract information rapidly, we investigate attribute reduction within the framework of dynamic incomplete decision systems. Firstly, we introduce positive knowledge granularity concept which is a novel measurement on information granularity in information systems, and further give the calculation method of core attributes based on positive knowledge granularity. Then, two incremental attribute reduction algorithms are presented for incomplete decision systems with multiple objects added and deleted on the basis of positive knowledge granularity. Furthermore, we adopt some numerical examples to illustrate the effectiveness and rationality of the proposed algorithms. In addition, time complexity of the two algorithms are conducted to demonstrate their advantages. Finally, we extract five datasets from UCI database and successfully run the algorithms to obtain corresponding reduction results.