Exploration of Heterogeneous Data Using Robust Similarity

TL;DR

This paper introduces a generic similarity-based exploration method for heterogeneous data that enhances visualization and analysis by revealing subtle structures, outliers, and modes across diverse data types.

Contribution

It proposes a novel similarity measure applicable to various data types and a visual encoding framework for multi-level exploration of heterogeneous datasets.

Findings

Effective detection of multiple modes and outliers.

Applicable to diverse heterogeneous datasets including ensembles.

Enhances data exploration and visualization capabilities.

Abstract

Heterogeneous data pose serious challenges to data analysis tasks, including exploration and visualization. Current techniques often utilize dimensionality reductions, aggregation, or conversion to numerical values to analyze heterogeneous data. However, the effectiveness of such techniques to find subtle structures such as the presence of multiple modes or detection of outliers is hindered by the challenge to find the proper subspaces or prior knowledge to reveal the structures. In this paper, we propose a generic similarity-based exploration technique that is applicable to a wide variety of datatypes and their combinations, including heterogeneous ensembles. The proposed concept of similarity has a close connection to statistical analysis and can be deployed for summarization, revealing fine structures such as the presence of multiple modes, and detection of anomalies or outliers. We then propose a visual encoding framework that enables the exploration of a heterogeneous dataset in different levels of detail and provides insightful information about both global and local structures. We demonstrate the utility of the proposed technique using various real datasets, including ensemble data.