Rice Grain Protein Content Predictions Using Machine Learning and Multi-Year Field Data

Multi-year field data was used to evaluate year-independent prediction of rice grain protein content with machine learning. The dataset, authored by Hyun-Jin Jung and last updated on 2026-06-01, includes plot-level and replicate-mean data structured for agronomic experimental design. Models were validated using leave-one-year-out cross-validation to simulate predictions for unseen growing seasons.

Use Cases

• Predicting rice grain protein content based on multi-year agronomic field data.
• Evaluating model generalization for unseen growing seasons using leave-one-year-out cross-validation.
• Analyzing the impact of replicate aggregation on predictive robustness for plot-level variability.
• Assessing systematic bias across protein levels and nitrogen application rates via residual diagnostics.
• Predicting deviations from year-variety baselines using anomaly-based analysis for relative variations.

Strengths

• Dataset employs a stringent year-independent validation framework using leave-one-year-out cross-validation.
• Analysis includes both plot-level and replicate-mean datasets to assess variability reduction.
• Residual diagnostics indicated no clear systematic bias across protein levels or nitrogen application rates.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Dataset size is 16.5 KB, indicating a very limited scope.

Provenance

Source

figshare

Collection Method

Multi-year field data collection aligned with agronomic experimental design.

Freshness

Last updated 2026-06-01 04:15:16; freshness should be verified.