Machine Learning · Data Science · Agricultural Analytics · 2026
Rice Yield Prediction Based on GDHY Data
An analysis and prediction project for rice yield using the global GDHY dataset. Climate grid data in NetCDF format is processed into a yield anomaly prediction model. The focus is on spatial data exploration and time series model evaluation.
- 3 metricsModel evaluationRMSE, MAE, and Pearson correlation
- Temporal validationData split strategyTimeSeriesSplit preserves temporal order
Case study contents
Objective
Analyze global rice yield patterns and build a model that can predict yield anomalies from historical climate data.
Results and limitations
The model was evaluated using RMSE, MAE, and Pearson correlation mapped spatially, along with identification of regions with the highest prediction difficulty. Limitation, the project is still at an exploratory stage and the documentation does not report explicit final metric values, and model performance likely varies across regions due to differing climate characteristics.
Technical details
Open implementation details
Role and contribution
Worked on the entire project alone. Processed NetCDF data, performed data exploration and spatial analysis, built a regression model, and evaluated model performance by region.
Methodology
Extracted and merged yearly NetCDF files into a single data structure with time and location dimensions. Cleaned sentinel values and capped unrealistic yield values. Performed spatial analysis to observe global production patterns, then trained a regression model using TimeSeriesSplit validation suited to time series data.
Technologies
- Python
- xarray
- pandas
- NumPy
- scikit-learn
- matplotlib
- seaborn
- scipy