Target Clustering Based on Integrated Discrepancy-Degree for Mixed-Type Attributes
Jia Zeng, Lingyu Yang, Shen Gongzhang · 2010
Target clustering forms the targets with similar attributes into several independent clusters, which is essential for the situation assessment. In this paper, a novel clustering method for the targets with mixed-type attributes is proposed. First of all, the attribute types are given according to the fundamental types of data, two new types, the "Vector-type attribute" and the "Interval-type attribute", are defined. Based on the characteristics of attribute types, the discrepancy-degree for each type is built, which extends the existing calculation method. An integrated discrepancy-degree for mixed-type attributes is established, where the quantitative or qualitative requirement of attribute is considered. The integrated discrepancy-degree solves the problem that some attributes are unsuitable for weighted sum. There is only one comprehensive comparison between individual targets and only one threshold needed to set. Several targets with mixed-attributes are selected into simulations and the results demonstrate the effectiveness of this approach.