Why KNN Retrieval vs. Random Forest Classifier?

KNN (K-Nearest Neighbors) retrieval is used in this project to provide instant, crop-specific advisories by finding the most similar historical sample in the dataset. Unlike a Random Forest Classifier (or other predictive models), KNN does not require training or generalization—it simply matches the user's input to the closest real-world example. This approach is more scalable for large, diverse advisory datasets and avoids memory issues that can occur with thousands of unique advisory classes. It also ensures that every recommendation is based on real, previously seen data, making the system robust and transparent.

Dataset & Retrieval Info

• Total advisory samples: 10,000
• Unique advisory classes: 9,996
• Features used for retrieval: Crop, Season, Soil Type, Temperature, Humidity, Rainfall

Top 20 Advisory Class Distribution

How Retrieval Works

When a user requests an advisory, the system encodes their input and finds the most similar record in the dataset using KNN. The advisory from that record is returned, ensuring recommendations are always based on real, historical data.

Newsletter & Updates

Stay up to date with the latest improvements and research in Agri AI Weather Advisory.
Subscribe to our Newsletter