Axes Re-Ordering in Parallel Coordinate for Pattern Optimization

Abstract

Visualization of multidimensional dataset is a challenging task due to non-uniformity of the data. It requires new ways to display data for better analysis and interpretation. Parallel coordinate is one of the popular techniques for visualization of multi dimensional dataset. Parallel coordinate technique emphasis various types of patterns present in the dataset. Here, pattern is shown by a poly-line. Slope of poly-line indicates the difference between data values. Variation in slope creates the different types of pattern. Based on slope, pattern can be classified and this kind of classification helps to explore distinct pattern available in dataset. Ordering of the axis affects pattern available in any dataset. Specific arrangement of axis may provide maximum patterns and another arrangement may provide minimum patterns. Ordering of axis in different order to find maximum or minimum pattern requires exponential time. Here, we propose a novel clustering technique using heuristic based branch & bound based axis reordering mechanism to solve this problem in polynomial time.