An Algorithm For Discretization Of Continuous Attributes Based On NBC Clustering In Rough Set Theory

Discretization of continuous attributes is an important preprocessing in machine leaning and data mining, researchers propose many algorithms about the discretization problem, for example, equal-width-intervals algorithm and equal-frequency-intervals algorithm In early time and discretization algorithm based on GA and so on in present. Discretization algorithms can be classified into supervised and unsupervised algorithms according to whether considering class information in discretization process.In this paper, firstly, evaluation criteria of discretization, several classical discretization algorithms and RS is introduced; Secondly, the algorithm thought and basic concept about NBC (Neighborhood-Based Clustering) is introduced, a new algorithm ADBNBC (Algorithm for discretization of Continuous Attributes Based on NBC Clustering) is proposed and apply it into Rough Set theory; At last, comparison experiments have been done about different discretization algorithms in the end of the paper, the result of the experiments shows the algorithm is effective.