Applied Machine Learning for Agriculture and Environmental Data

Aim

Workshop Objectives

Workshop Structure

📅 Day 1: Geospatial Engineering & Data Robustness

• Goal: Mastering the “Research-to-Data” pipeline using satellite and sensor inputs
• Handling Scientific Noise: Implementing automated pipelines for outlier detection and KNN-based imputation for missing environmental sensor logs
• Feature Engineering for Agri-Data: Programmatic calculation of Vegetation Indices (NDVI, EVI) and atmospheric corrections using Geopandas and Rasterio
• Dimensionality Reduction: Using PCA and RFE to identify the “minimal viable features” for publication-quality models
• Hands-on:
  • Building a pre-processing class in Python to clean a raw  CSV/GeoJSON dataset

📅 Day 2: Advanced Ensemble Modeling & Optimization

• Goal: Implementing high-performance algorithms that outperform traditional regression
• Beyond Linear Models: Implementing Gradient Boosted Decision Trees (XGBoost & LightGBM) for complex crop yield and soil carbon sequestration forecasting
• Spatial Validation Strategies: Addressing “Spatial Autocorrelation” using Group K-Fold Cross-Validation to ensure model generalizability across different geographical study areas
• Hyperparameter Tuning: Utilizing Optuna for Bayesian optimization to maximize $R^2$ and minimize RMSE
• Hands-on:
  • Training a multi-stage regressor to predict agricultural output based on multi-variate climate data

📅 Day 3: Explainable AI (XAI) & Research Deployment

• Goal: Interpreting the “Black Box” to meet peer-review standards for causality
• Global & Local Interpretability: Utilizing SHAP to quantify the impact of specific environmental variables on model outcomes
• Feature Dependency: Generating Partial Dependence Plots (PDP) to visualize non-linear relationships (e.g., tipping point where rainfall becomes detrimental to crop health)
• Interactive Prototyping: Deploying a live Gradio interface within Google Colab to allow peers to interact with your trained model in real-time
• Hands-on:
  • Generating a “Feature Importance Report” suitable for inclusion in a scientific manuscript

Who Should Enrol?

Important Dates

Workshop Outcomes

Fee Structure

What You’ll Gain