Multiple alternative clusterings and dimensionality reduction

Clustering is the process of grouping objects based on some notion of similarity. It is commonly applied for exploratory data analysis, segmentation, preprocessing and data summarization.Traditional clustering algorithms only find one clustering solution. However, data can be grouped and interpreted in many different ways. Moreover, different clustering solutions are interesting for different purposes. Instead of committing to one clustering solution, here we introduce four methods that can provide several possible alternative clustering solutions to the user for exploratory data analysis.

Autorentext

Donglin Niu received Bachelor degree in Nanjing University, China and PhD degree in Northeastern University, Boston, in electrical engineering. His interest includes machine learning, web scale data mining, distributed computing and digital advertising. He currently holds a research engineer position at Yahoo Inc.