Multivariate Geochemical Tectonic Discrimination:
Practical Approaches, Limitations and Opportunities

Drop by my AEGC poster: #334 in Meeting Room 2 during or after lunch on Wednesday 4^th September, or contact me via one of the channels listed below.

Classical geochemical classification methods are human-centric, and commonly limited by the ability to visualise data. Modern approaches using machine classification are enabled by large public data repositories, and can be used to provide more objective perspectives and to investigate classification uncertainty.

This is a Binder-enabled repository to accompany a poster for the 2019 Australasian Exploration Geoscience Conference, which covers many of the key elements required to construct classifiers using geochemical data and open-source tools (scikit-learn and pyrolite). You can find the accompanying extended abstract here.

---

Contents

Start Here: Intro
Data: Compositional Data | Lambdas
Classifiers: Building Models | Feature Selection
Visualisation: Entropy | Manifolds
pyrolite: Overview

---

Maintainer:

Morgan Williams (morgan.williams at csiro.au)