The problem 🌱 SoilSense solves

SoilSense addresses the critical issue of soil degradation, which negatively impacts crop yields, farmer income, and food security. Many farmers lack access to real-time soil information and rely on traditional or guesswork methods to manage soil health. This leads to:

• Poor crop selection and lower yields
• Overuse or misuse of fertilizers
• Long-term soil depletion and reduced sustainability

People can use SoilSense to:

• Quickly assess soil health: Get AI-powered insights on soil nutrients, pH, and moisture.
• Make informed decisions: Receive recommendations for optimal crops, fertilizers, and farming practices.
• Save time and resources: Reduce guesswork, avoid unnecessary fertilizer use, and improve efficiency.
• Promote sustainability: Implement soil conservation practices that protect soil health long-term.

Challenges I ran into

Challenges I Ran Into

1. Balancing Accuracy and Interpretability

   • Using a hybrid model (Random Forest + XGBoost) improved prediction accuracy but made it harder to explain results to non-technical users.
   • Solution: Applied SHAP explainability to highlight which soil features influenced predictions, making recommendations transparent and trustworthy.

2. Hyperparameter Tuning

   • Optimizing XGBoost and RF manually was time-consuming and inefficient.
   • Solution: Used Optuna to automatically search for the best hyperparameters, improving model performance and generalization.

3. Data Limitations

   • Soil datasets were incomplete, inconsistent, or in different formats.
   • Solution: Preprocessed the data, handled missing values, and created a synthetic dataset to ensure reliable model training.

Outcome: Overcoming these challenges improved the project’s accuracy, interpretability, and usability, making SoilSense both powerful and user-friendly.

Smart Contract Address(es)

This project uses AI to monitor soil health and provide actionable guidance to farmers; no blockchain smart contract is involved.