# AgriYield 2025: Predict Crop Yield from Soil, Weather & Remote Sensing Data By Intelligence Academy

### **Problem Statement**

Accurate crop yield prediction plays a vital role in ensuring food security, optimizing resource allocation, and informing evidence-based agricultural policies. In this competition, your task is to develop machine learning models that can predict maize yield (in kg/ha) using a rich set of features, including soil attributes, historical weather records, and satellite-derived vegetation indices.

This challenge aims to support precision agriculture by enabling early yield forecasting and smarter farm management. Your model’s success could directly contribute to improving agricultural sustainability and enhancing decision-making for farmers, agronomists, and policymakers worldwide.



### **Dataset Description**

The dataset consists of geotagged, field-level observations collected across diverse agricultural zones and cropping seasons. Each row represents an individual field and includes:

* **Soil Properties**: pH, organic matter, sand/silt/clay ratios, and other physicochemical indicators
* **Weather Data**: Daily and seasonal measurements such as rainfall, temperature (C), and relative humidity
* **Remote Sensing Indices**: NDVI, EVI, LAI, and other vegetation indices computed from satellite imagery
* **Crop Management Data** *(if available)*: Planting date, fertilizer application, irrigation flags
* **Target Variable**: Observed yield in kg/ha

#### Files provided:

* `train.csv` – Training data with input features and yield values
* `test.csv` – Test data with features only (for prediction)
* `sample_submission.csv` – Submission template

All continuous variables are normalized. Missing values are flagged. More metadata can be found in the accompanying data dictionary.



###  **Objective**

Your objective is to train a machine learning model that generalizes well and can **accurately predict the maize yield** for unseen fields in the test set.

This is a **regression problem**, and your final score will be determined by how closely your predictions match the true yields.



###  **Evaluation Metric**

Submissions will be evaluated using **Root Mean Squared Error (RMSE)**, a standard regression metric that penalizes large deviations.

$$
\text{RMSE} = \sqrt{ \frac{1}{n} \sum_{i=1}^{n} \left( y_i - \hat{y}_i \right)^2 }
$$

Where:

* yi is the ground truth yield
* $$
 \hat{y}
$$ is the predicted yield
* n is the number of test samples

A **lower RMSE** indicates **better predictive performance**.



### **Why Participate?**

* 💡 Apply machine learning to solve a **real-world challenge** in agriculture
* 🤝 Collaborate and learn from a **global data science community**
* 🧠 Improve your modeling, feature engineering, and generalization skills
* 🏅 Compete for recognition, resume-worthy results, and potential publications
* 🌱 Contribute toward **sustainable agriculture** and **climate-resilient food systems**


### **Timeline**

* 📥 Competition Launch: **July 27, 2025**
* 📤 Final Submission Deadline: **September 1, 2025**
* 🏁 Winners Announced: **September 15, 2025**

Note: The competition may include an interim leaderboard freeze period. Please monitor announcements.



###  **External Data & Notebooks**

You may use **external datasets** only if:

* They are **freely and publicly available**
* Their source is **fully documented** in your code
* The data is **uploaded to Kaggle Datasets** for transparency and reproducibility

Use of private or undisclosed external data is **strictly prohibited**.

Public code and notebooks are allowed **only after the Code Freeze Date** (to be announced).



###  Final Note

This is your chance to help revolutionize agriculture using data science. Let's build models that can drive **food security**, **sustainability**, and **global resilience**.

**Good luck** — may your **RMSE**  be low and your yields be high!