Dataset Vertical Partitioning for Rough Set Based Classification

Dataset partitioning problem involves the vertical partitioning of the classification datasets into suitable subsets that preserve or enhance the classification quality of the original datasets. Typical classification model needs to be constructed for each subset and all generated models are then combined to form the classification model. This paper presents a dataset partitioning approach for rough set based classification. In this approach, the dataset is partitioned into two mutually exclusive subsets. Local reduct set is generated for each attribute subset which is then combined and used to generate the set of classification rules. A preliminary experimental result using the partitioning approach over some standard medical datasets showed that the approach preserves the classification accuracy.