Linear Correlation Analysis of Numeric Attributes for Government Data

To analyze the linear correlations of numeric attributes of government data, this paper proposes a method based on the clustering algorithm. A clustering method is adopted to prune outliers and the linear correlation analysis is performed for each cluster, instead for the whole dataset. In this way, the method can obtain multiple correlations between the same two attributes. The paper presents the experiment on the government social security data. Experimental results show that the proposed method is much better than the traditional regression analysis and association rule analysis.

• Type: Conference
• Year: 2007
• Where: IMSCCS
• Authors: Ying Chen, Guochang Gu, Tian-yang Lv, Shaobin Huang, Jun Ni
• Keywords: Clustering Method, IMSCCS 2007, Linear Correlation, Linear Correlation Analysis, Theoretical Computer Science