A Consideration on Learning by Rule Generation from Tables with Missing Values

This paper considers rough set-based machine learning from tables. We coped with rule generation from tables with non-deterministic information, and proposed the NIS-Apriori algorithm. After executing this algorithm, as a side effect we obtain tables causing the best criterion values of the implications and tables causing the worst criterion values of the implications. We apply this property to tables with missing values, and propose the new framework rough set-based machine learning by rule generation. By showing examples, we describe the overview of this new framework.