Star Coordinate Dimension Arrangement using Euclidean Distance and Pearson Correlation

Abstract

Star Coordinate (SC) is a circular visualization technique that maps k-dimensional data. Its interactive features allow user to manipulate projections to search for hidden information. Without prior knowledge of relationship between dimensions users will be blindly searching for clusters. This paper proposes dimension rearrangement using Euclidean Distance and Pearson Correlations to reveal the clusters in SC. The methodology consists of four phases; Calculate the distance between individual attributes against a dependent attribute using Euclidean distance; Pearson correlation is used to obtain the correlation data attributes; Sort the correlation values in ascending order; finally, attributes table are reordered with the positive values to the right and negative values to the left according to the correlation value. The resulting tables are applied to produce the SC. This method is successful in producing clusters that makes it easier for the users to further manipulate the SC for their data analysis.