Tree-based ML algorithms
Table of Contents
Description & principle
A clear, technical yet accessible explanation of the method, its core principle(s).
Major variants
optional
If the method has variants that seem important, either already widespread or promising and well documented.
Further online resources
References to useful online resources to get started, e.g. explanation blogs
Reference articles
Method
- One or a few key academic references that introduce or formalize the method.
Research applications
With RS data in Ecology / Biodiversity
- A
Without RS data (Ecology domain)
optional
- B
Packages
Python
R
Code Cells
optional
Assessment table
| Category | Criteria | Assessment |
|---|---|---|
| Outcome | Objective | Predictive task + interpretability, Scenario projection |
| Estimand | Maps & generalisations | |
| Data compatibility | Type | Spatial only (cross-sectional), Time-series (one sample), Panel data (many samples) |
| Required TS length | ||
| Handles few samples | 10 to 100 | |
| Handles huge datasets (n) | Yes | |
| Handles missing data | No: requires prelim. correction | |
| RS-data proven | Yes | |
| Assumptions | Functional form | Rule-based, Assumption-free |
| No unobserved confounders | Required | |
| No interference | Relaxes assumption, Not required | |
| Well-defined treatments | Relaxes assumption | |
| Common support (positivity) | Relaxes assumption | |
| Causal Markov Condition | Inapplicable | |
| Faithfulness | Inapplicable | |
| IDD | Relaxes assumption | |
| Model specific assumption | No specific | |
| Model properties | Requires explicit processes | Agnostic |
| Exposure type | Binary, Categorical, Continuous / Time-varying | |
| Number of variables | High-dimensional (p≫n) | |
| Handles lag effects | No | |
| Propagates uncertainty | Inapplicable | |
| Parametric nature | Non-parametric, Rule-based | |
| Packages | Language | R, Python |
| Usage | Technical but well documented |