A Method of Weather Cases Generation Based on Similarity Rough Set

Case selection from weather database is a key step in disaster weather forecasts based on CBR. The selection of representative weather cases without noise and reduces time and space complexity is its essential target. This paper proposes the SRS algorithm based on similarity-based rough set theory. By reducing undirected graph, it can select a reasonable number of the typical cases from a large data set for future case-based reasoning tasks. It also can handle noise and inconsistent data. Experimental result has confirmed the algorithm feasibility and the validity.