2-Irrigation Need Prediction

Irrigation Need Prediction

This project aims to predict irrigation requirements (Low, Medium, High) using environmental and agricultural data to support efficient water management in smart farming.

? Approach

Applied feature engineering to create meaningful variables such as water efficiency and environmental ratios.

Used a preprocessing pipeline including:

Standardization for numerical features

Ordinal, binary, and one-hot encoding for categorical features

Built a deep neural network (DNN) using TensorFlow/Keras with dropout and L2 regularization.

Addressed class imbalance using computed class weights.

Applied Early Stopping and ReduceLROnPlateau for better training stability.

? Results

Validation Accuracy: 98%

Strong performance across all classes, including the minority class.

? Output

Generated predictions for the test dataset and saved them in a submission file ready for Kaggle.