Online Program Home
My Program

Abstract Details

Activity Number: 529 - Regression Trees and Random Forests
Type: Contributed
Date/Time: Wednesday, August 1, 2018 : 10:30 AM to 12:20 PM
Sponsor: Section on Statistical Learning and Data Science
Abstract #330905 Presentation
Title: Spectral Clustering via Unsupervised Random Forests
Author(s): William Biscarri* and Robert J. Brunner and Ruoqing Zhu
Companies: University of Illinois at Urbana-Champaign and University of Illinois at Urbana-Champaign and University of Illinois Urbana-Champaign
Keywords: Spectral Clustering; Random Forest; Affinity Matrix; Network Link Probability Estimation

Due to their ability to handle non-convex data, spectral clustering methods have proven to be effective in a wide use of clustering problems. A key step of spectral clustering algorithms is the construction of an affinity matrix, which contains a measure of similarity between each of the observed data points. Once constructed, the affinity matrix serves as the primary component for determining the how the data are clustered. Thus, effective methods for constructing affinity matrices are crucial to the success of spectral clustering. Motivated by the kernel representation of Random Forests, we present an unsupervised Random Forest algorithm for constructing affinity matrices. We demonstrate that this algorithm leads to sensible data clusters, and is also able to effectively handle high dimensional data. Further possible uses of the algorithm are discussed, with particular attention paid to network link probability estimation.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program