A Scalable Correlation-Driven Framework for Compact Visualization of Multidimensional Data using Adaptive Projection Technique

---

Abstract

In the era of rapidly developing digital technology, the application of data visualization has increased exponentially. But in data visualization, the graphical charts or tools are only limited to represent the relationship among two or three parameters of the dataset. However, most real-world datasets contain many more attributes. Projecting two parameters on to two-dimensional display prevents the discovery of multidimensional associations like correlations. In this paper, a scalable correlation-driven framework is proposed for Compact Visualization of Multidimensional Data (ComVisMD). The proposed method, ComVisMD, is based on an adaptive projection technique that maintains essential relationships among visualization elements. It is simulated using the ‘graphics’ package in the Python programming language. It is used to display five dimensions for mapping five attributes of dataset into 2D compacted formation. It easily helps to illustrate the hidden information of the dataset. Moreover, users can interactively map the display to search for data insights. The performance of ComVisMD is compared with other common graphical methods. Additionally, comparative analysis is also done with some existing methods. Finally, it is validated with a cricket players’ dataset based on visual analytics parameters.

---

Keywords

Multidimensional Data Visualization; Data Projection; Data Aggregation; Correlation Analysis; Visual Analytics for Complex Data; Intelligent Decision Support Systems.