Tutorial: ML for Hydrological Sciences

[aaarendt]

Lead: Savalan Neisary
Date: 20/08/2024
Start Time: 1045
Duration: 60
Description: Introduction to Applying XGBoost in Post-Processing Hydrological Models

Details

Learning Outcomes

• Understand the basics of machine learning and decision-tree algorithms.
• Learn how to apply and train an XGBoost model for hydrological modeling.
• Learn how to implement feature selection using the XGBoost algorithm.

People Developing the Tutorial (content creation, helpers, teachers)

Summary Description

The Decision-Tree workshop will explore simple XGBoost in hydrological modeling. The workshop will briefly introduce machine learning basics and decision tree algorithms and transition to hands-on activities in which participants will engage in the XGBoost model development pipeline, including data processing, hyperparameter tunning, feature selection, algorithm training, and model evaluation. The Python code and data will be available through GitHub. Participants can expect an improved understanding of the XGBoost algorithm and its applications within hydrological modeling and knowledge of data preprocessing and visualization.

Dependencies (things people should know in advance of the tutorial)

• Basic background in Python coding and using packages such as Numpy and Pandas.
• Knowledge of basic machine learning concepts and terminology.
• Experience with Jupyter notebooks

Technical Needs (GPUs? Large file storage? Unique libraries?)

• DMLC XGBoost library.
• Scikit-learn library.