🌱 Crop Intelligence Architecture
Ecological Suitability Optimization: Multi-Class Agronomic Classifier
22 Crop Varieties
96.4% Accuracy
XGBoost Ensemble
Top-5 Priority Logic
flowchart TD
subgraph Data_Pipe ["1. Agronomic Ingestion (Crop_Engine.py)"]
A1["Soil Health Card Data
(N, P, K, pH)"] & A2["Environmental Logs
(Temp, Hum, Rain)"] --> B1["Input Normalization
+ Unit Conversion"]
end
subgraph Feature_Engineering ["2. Agronomic Feature Expansion"]
B1 --> C1["Nutrient Ratios
(n_ratio, p_ratio, k_ratio)"]
B1 --> C2["Environmental Proxies
(Aridity & Heat Indices)"]
B1 --> C3["Categorical Regimes
(pH & Rainfall Bins)"]
C1 & C2 & C3 --> C4["Expanded Feature Vector
(40+ Dimensions)"]
end
subgraph Model_Architecture ["3. Classification Engine"]
C4 --> D1["XGBoost Ensemble Model
(Decision Tree Forest)"]
D1 --> D2["Multi-Class Softmax
(22 Probability Scores)"]
D2 --> D3["Confidence Ranking
(Top-K Selection)"]
end
subgraph System_Deployment ["4. AgriSense AI Core (serve.py)"]
D3 -.-> H1["Recommendation API
/api/crop"]
H1 --> H2["Optimized Crop List
+ Confidence Scores
+ Agronomic Tips"]
end
classDef source fill:#f0fdf4,stroke:#16a34a,stroke-width:2px;
classDef feat fill:#fffbeb,stroke:#d97706,stroke-width:2px;
classDef model fill:#eff6ff,stroke:#2563eb,stroke-width:2px;
classDef artifact fill:#f9fafb,stroke:#4b5563,stroke-width:2px;
class A1,A2,B1 source;
class C1,C2,C3,C4 feat;
class D1,D2,D3 model;
class H1,H2 artifact;
🧪 Agronomic Context Scaling
The model expands 7 raw inputs into **40+ engineered features**, identifying critical biological thresholds like the Aridity Index and Nutrient Dominance.
🏆 Multi-Class Precision
Achieves a **96.4% accuracy** across 22 crop varieties, successfully resolving overlaps between visually similar cereal and pulse ecological niches.
⚡ Risk-Averse Recommendations
Instead of a single prediction, the engine provides a **Top-5 prioritized list**, allowing farmers to weigh AI confidence against seed availability and market demand.
📡 Soil Health Integration
Designed to consume data directly from Soil Health Card (SHC) schemas, bridging the gap between raw laboratory measurements and field-level planning.