The STEM is an ensemble of locally defined, randomly located, spatially overalpping decision trees and generates a final prediction by aggregating over all overlapping pixels with either the mean, mode, or weighted mean. These scripts offer a Python implementation of the STEM algorithm developed by Fink and others (2010)
Fink, Daniel, Wesley M. Hochachka, Benjamin Zuckerberg, David W. Winkler, Ben Shaby, M. Arthur Munson, Giles Hooker, Mirek Riedewald, Daniel Sheldon, and Steve Kelling. 2010. “Spatiotemporal Exploratory Models for Broad-Scale Survey Data.” Ecological Applications 20 (8): 2131–2147.