l16 广西师范大学学报:自然科学版 第27卷 [8]LEI Yu,LIU Huan.Efficient feature selection via analysis of relevance and redundancy[J-].Journal of Machine Learn— ing Research,2004,5:1205—1224. Correlation Calculation for Mixed Attribute LUO Rong—hal,JIANG Sheng-yi (School of Informatics・Guangdong University of Foreign Studies,Guangzhou 510006,China) Abstract:The current way of calculating the mixed attribute correlation iS to discretize the continuous at— tribute and then process the dataset based on frequency.But discretization will probably result in losing .parts of information,which brings some uncertainties in the attribute evaluation.A new approach of cal— culating mixed attribute correlation without discretization is propesed.All data of a continuous attribute are grouped by the values of a discrete attribute.On the basis of the relation between the variances of pre and post grouping,correlation between these tWO attributes can be measured.The experiments,conduct— ed with tWO real datasets and two UCI datasets。show that this approach can measure the mixed attribute correlation effectively. Key words:mixed attribute;correlation;variance (责任编辑 马殷华)