Rough set (RS) theory is one of the most effective data mining tools. So far, researchers have proposed dozens or even hundreds of RS models to cope with various data challenges. However, scholars have not systematically compared and investigated the performance of these models in analyzing data. To address this issue, we select 12 most representative RS models derived from binary information tables and conduct comparative studies on them in terms of computational efficiency, approximation accuracy, classification accuracy, data dimensionality, and comprehensive performance. Through detailed experimental results, we reasonably evaluate the advantages of these models and offer reliable references for selecting suitable models to complete specific learning tasks.