Parallel Coordinates with Data Labels

Parallel coordinates have been widely used to analyze high-dimensional data. Numerous methods have been designed to provide overview patterns in parallel coordinate plots. However, detailed information is also important in data analysis. When several lines overlap or are close to one another, distinguishing detailed information of polyline crossings is difficult. In this paper, we present a novel approach to address the problem of polyline crossing ambiguity by using data labels. We place different labels along various polylines to give cues for differentiation of lines. We bend the lines and optimize the arrangement of curved lines to provide space for clear visible labels. An energy system that models attractive and repulsive forces of lines is used to guide the search for optimized line arrangement. The experiments on several real datasets demonstrate the effectiveness of our approach.