Interactive Mode Explorer Sonification Enhances Exploratory Cluster Analysis

Yang J, Hermann T. Interactive Mode Explorer Sonification Enhances Exploratory Cluster Analysis. J Audio Eng Soc. 2018;66(9):703-711. doi:10.17743/jaes.2018.0042

@article{Yang2018_19712,
  author = {Yang, Jiajun and Hermann, Thomas},
  title = {{Interactive Mode Explorer Sonification Enhances Exploratory Cluster Analysis}},
  journal = {Journal of the Audio Engineering Society},
  volume = {66},
  number = {9},
  pages = {703--711},
  year = {2018},
  month = sep,
  publisher = {Audio Engineering Society},
  doi = {10.17743/jaes.2018.0042},
  url = {https://doi.org/10.17743/jaes.2018.0042}
}

TY  - JOUR
TI  - Interactive Mode Explorer Sonification Enhances Exploratory Cluster Analysis
AU  - Yang, Jiajun
AU  - Hermann, Thomas
T2  - Journal of the Audio Engineering Society
J2  - J. Audio Eng. Soc.
VL  - 66
IS  - 9
SP  - 703
EP  - 711
PY  - 2018
DA  - 2018/09/06
DO  - 10.17743/jaes.2018.0042
UR  - https://doi.org/10.17743/jaes.2018.0042
PB  - Audio Engineering Society
LA  - en
AB  - Exploratory Data Analysis (EDA) refers to the process of detecting patterns of data when explicit knowledge of such patterns within the data is missing. Because EDA predominantly employs data visualization, it remains challenging to visualize high-dimensional data. To minimize the challenge, some information can be shifted into the auditory channel using humans’ highly developed listening skills. This paper introduces Mode Explorer, a new sonification model that enables continuous interactive exploration of datasets with regards to their clustering. The method was shown to be effective in supporting users in the more accurate assessment of cluster mass and number of clusters. While the Mode Explorer sonification aimed to support cluster analysis, the ongoing research has the goal of establishing a more general toolbox of sonification models, tailored to uncover different structural aspects of high-dimensional data. The principle of extending the data display to the auditory domain is applied by augmenting interactions with 2D scatter plots of high-dimensional data with information about the probability density function.
ER  -