Statistical and Machine Learning for Geospatial Big Data
International Biometric Society Conference, Atlanta, GA, 2024
Instructor
Abhirup Datta, PhD
abhidatta@jhu.edu Department of Biostatistics
Johns Hopkins University
Co-Instructor
Wentao Zhan, PhD Candidate
wzhan3@jhu.edu Department of Biostatistics
Johns Hopkins University
Outline
Traditional Geostatistical Analysis
- Exploratory data analysis
- Spatial linear mixed effect models
- Gaussian processes and kriging
- Methods for spatial big data
Introduction to Non-Linear Machine Learning Algorithms
- Random Forests
- Neural Networks
- Challenges of standard machine learning for spatially correlated data
Machine Learning Algorithms for Spatially Correlated Data
- How to use spatial correlation in machine learning algorithms?
- RF-GLS: Random Forests for spatially dependent data
- NN-GLS and geospaNN: (Graph) neural networks for geospatial data
- Demonstration of software:
RandomForestsGLS(R)geospaNN(Python)