[Japanese]

An Interactive Tool for Feature Analysis of Outliers in Multi-Dimensional Data

This paper presents an interactive visualization tool for the analy- sis of outliers in multi-dimensional data. Each data sample consists of multiple variables (features). For a given outlier data sample, our tool helps the user to identify which variable of the data sam- ple makes it an outlier. Our tool consists of a scatter plot view and variable relation graph view. The user first identifies an outlier data sample in a scatter plot. The user selects the outlier data sample, and the system updates the variable relation graph to visualize rela- tionship between the variables of the outlier data sample. The user then examines the variable relation graph and the scatter plot based on the visualization to identify outlying variables.

Thesis

Video

link