Epidemiology and Biostatistics Dissertation Defense: Melissa Meeker
Monday, May 8, 2023
2:00 PM-3:00 PM
"A Spatial Extension of the Random Forest Algorithm"
Melissa Meeker, PhD student at Dornsife
The random forest machine learning algorithm is often used in spatial
datasets with strong predictive performance. However, the random forest
algorithm does not acknowledge the correlation structure in spatial datasets.
In this dissertation, we propose three spatial extensions of the random forest
which include (1) implementing a geographically stratified sampling approach,
(2) incorporating a neighbor-based predictor set, and (3) modifying the split
criterion used to produce the decision trees. We examine the effect of these
modifications in three types of data: simulated data, Philadelphia Police
Department investigations data, and Christiana Hospital NICU data.
Melissa Meeker is a PhD candidate in biostatistics at Drexel University
with a B.S. in mathematics and computer science from Ursinus College. Melissa’s
research interests include methods for spatial data, machine learning, and
public health disparities.
Contact Information
Nancy Colon-Anderson
nanderson@drexel.edu