AgriPredict

Empowering Farmers with AI: Predicting Decisions That Feed a Nation.

Comment

🌾 About the Project: AgriPredict

Inspiration

While exploring some of the most pressing issues faced by India, I was struck by a tragic but often overlooked statistic: over 10,000 farmers die by suicide each year, according to NCRB. Farming — the backbone of our food system — has become one of the most dangerous occupations, not only in India but also in countries like the US, UK, and Canada.

These suicides are often rooted in financial stress, driven by volatile crop prices, unpredictable weather, and disasters. Farmers are unable to plan ahead due to a lack of reliable forecasting tools. That’s when the idea for AgriPredict was born — a platform to forecast crop prices and help farmers make informed, confident decisions.

How I Built It

  • The project is built using Flask for the backend and HTML/CSS for the frontend.
  • I used artificially generated datasets simulating seasonal patterns and sudden price fluctuations (such as those caused by disasters).
  • To capture the seasonal trends and abrupt changes in prices, I used SARIMA (Seasonal AutoRegressive Integrated Moving Average) — a time-series forecasting model well-suited for agriculture data.
  • To enhance accuracy, I combined SARIMA with Gradient Boosting (XGBoost), creating a hybrid model that learns residual patterns effectively.
  • Since a single model isn’t feasible for a large country like India, I divided the nation into 100 centers, each with its own localized model for regional price forecasting.

Challenges Faced

  • I discovered this hackathon at the last moment, so I focused on building and integrating the core functionality: price prediction.
  • I wasn’t able to integrate weather forecasting and disaster alerts, which I plan to add in the next version.
  • Creating regional models and tuning them efficiently within the limited time was also a technical challenge.

What I Learned

  • The importance of combining domain knowledge (agriculture) with machine learning for real-world impact.
  • Hands-on experience in time series modeling, hybrid ML systems, and Flask deployment.
  • Simulating realistic datasets and structuring solutions for nation-wide scalability.

What’s Next

  • Integrate weather and disaster forecasting APIs for more robust predictions.
  • Improve UI/UX to make it more user-friendly for farmers and decision-makers.
  • Deploy the application and test it with real-world data for feedback and improvement.

🔗 Check out the GitHub repository for the source code.

Share this project:

Updates