An Improved Fuzzy Discretization Way for Decision Tables with Continuous Attributes

In this paper, a discretization method for continuous decision tables is put forward. The method includes two steps called crisp discretization and fuzzy discretization. The first step is to get some crisp "cuts" based on information entropy and discernibility matrix for the decision table. The second step is to carry out fuzzy operation based on the former crisp discretization. In the second step, in spite of constructing crisp discrete intervals or fuzzy discrete intervals, the crisp "cut intervals" are even more important than the single "cuts".