Skip to content

README.md

Latest commit

 

History

History
85 lines (54 loc) · 3.55 KB

README.md

File metadata and controls

85 lines (54 loc) · 3.55 KB

RHESSysML

A comprehensive workflow to determine and visualize variable importance in RHESSys model output.

GitHub last commit GitHub issues GitHub pull requests GitHub repo size

About • How to Use • Choosing a Model • Wiki

About

RHESSysML provides a template and worked example for quickly exploring and identifying interesting variable relationships in RHESSys output data. This workflow is intended for users after the RHESSys model has been run and calibrated.

This repository contains the following directories:

  • /R: R functions used in the workflow.

  • /data: Folder to place RHESSys data for use in the workflow. Also contains data sets used in the completed example for Sagehen Creek.

  • /notebook_templates: Blank workflow notebooks for use with new datasets.

  • /notebooks: Notebooks used in the completed example for Sagehen Creek.

    • /supporting_docs: Notebooks supporting key choices made in the workflow.

  • shiny: Files and subdirectories associated with the Shiny application for interactive visualization of results.

  • renv: Files and subdirectories created by the renv package.

How to Use

  1. Fork and clone this repository.

  2. Place your RHESSys data in the data folder.

If this is your first time using this workflow, we suggest viewing the files within notebooks for steps 3-4 for more explanation and an example of a completed analysis.

  1. In notebook_templates, use "data_preparation.Rmd" to prepare data. We suggest aggregating by water year for the best results.

  2. In notebook_templates, run "rf_variable_importance.Rmd" or "gb_variable_importance.Rmd". For most use cases, "rf_variable_importance.Rmd" is preferred.

  3. In shiny, open "shiny_app.R" using RStudio and hit "Run App". The app can also be run via the command line using R -e “shiny::runApp(‘/shiny’)”.

Choosing a Model

Random Forest Gradient Boosting
Faster run time ✔️ ❌
Less tuning ✔️ ❌
Accurate predictive power ✔️ ✔️
Better maximum accuracy ❌️ ✔️

Wiki

Do you need some help? For help specific to this workflow, check the documentation and guidance within notebooks. For help with RHESSys, check the articles from the wiki.