Online Program Home
My Program

Abstract Details

Activity Number: 177 - Section on Statistical Learning and Data Science CPapers 2
Type: Contributed
Date/Time: Monday, July 30, 2018 : 10:30 AM to 12:20 PM
Sponsor: Section on Statistical Learning and Data Science
Abstract #330802 Presentation
Title: WPSVM for Spatial Point Processes Directed by Gaussian Random Fields
Author(s): Subha Datta*
Companies: New Jersey Institute of Technology
Keywords: WPSVM; PSVM; SDR; SVM
Abstract:

Sufficient Dimension Reduction (SDR) has become an essential tool in dealing with high-dimensional data in the past few years. In this paper, we attempt to tackle the issue of high-dimensionality for inhomogeneous spatial point processes. The spatial point processes have been treated as binary response. We have used an SDR method, called weighted principal support vector machine (WPSVM). Most SDR methods don't perform well when the response is binary. It is, hence, the desire of the researcher to develop an efficient method to reduce dimension of the data without loss of information with binary response. PSVM combines techniques of SDR and SVM and can extract sufficient predictors for both linear and non-linear models. However, PSVM suffers from estimating at most one direction of the central subspace for binary response. Shin et. al. has shown that the WPSVM can estimate more than one direction with binary response. We have conducted simulation studies to evaluate the finite sample performance of the SDR methods and WPSVM. WPSVM seems to work reasonably well for inhomogeneous spatial point processes when compared with other methods and hence is our suggested approach.


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

Back to the full JSM 2018 program