A Discretization Method Based on Statistical Indicator

Discretization of continuous data can improve classification ability of data mining algorithms.This paper proposes a discretization method based on statistical indicator.It measures the interdependence between class label and attributes with the aim to find optimal interval lists by means of correlation coefficent.It also introduces variable precision rough set model to effectively control information loss generated by discretization.The presented method is applied to sonar sensor data recognition and other fields.Experimental results show that this method significantly improves the classification ability on J48 decision tree