Regression Model
Our motivation behind this model was because in Texas agriculture generates over $100 billion annually, and being able to understand climate-yield relationships can imform drought preparedness, irrigation planning, and crop selection decisions. Historical patterns show dramatic yield drops during years of drought, yet we lack quantitative models connecting specific climate conditions to production outcomes.
We will combine the USDA agricultural records with NOAA climate data spanning over 20 years covering 254 climates and 165 different crop types. We will start with a decision tree baseline and improve it using AdaBoost ensemble and other hyperparameter tuning methods, the model will help identify which climate factors most strongly influence crop outomces with potential applications in drought planning and resource allocation for Texas’s prime agriculture industry.
We will measure our success based on:
- having an R² > 0.70 on testing sets
- Model correctly predicts yield decreases during drough years
- Climate features appear in top 10 feature importance rankings
- RMSE below persistence model benchmark.
