UCGIS Webinar 2024 Fall series kicks off 9/26 with Dr. Ziqi Li

3 views
Skip to first unread message

Song GAO

unread,
Sep 5, 2024, 5:37:05 PM9/5/24
to CPG...@listserv.buffalo.edu
GeoShapley: A Game Theory Approach to Measuring Spatial Effects in Machine Learning Models

Presenter: Dr. Ziqi Li (Florida State University)

Abstract: GeoShapley is a game theory approach to measuring spatial effects in machine learning models. The approach extends the Nobel Prize–winning Shapley value framework by conceptualizing location as a player in a model prediction game, which enables the quantification of the importance of location and the synergies between location and other features in a model. GeoShapley is model-agnostic and can be applied to models of various structures. The webinar will discuss the principles of GeoShapley, its effectiveness with both synthetic and empirical data, and a GeoShapley Python package. Lastly, the webinar will discuss some best practices and caveats when applying GeoShapley to real-world data.

Song GAO

unread,
Sep 25, 2024, 6:31:08 PM9/25/24
to CPG...@listserv.buffalo.edu
Dear colleagues and students,

A reminder for the following webinar tomorrow:-) 
Reply all
Reply to author
Forward
0 new messages