-Predicting-Irrigation-Need-

This project focuses on building a machine learning system to predict crop irrigation requirements based on environmental and soil conditions. The model classifies irrigation needs into three categories: Low, Medium, and High, helping optimize water usage and support efficient agricultural decision-making.

A complete data preprocessing pipeline was applied, including handling categorical features using Ordinal Encoding, feature scaling with StandardScaler, and addressing class imbalance using SMOTE to improve model fairness and performance.

Extensive feature engineering was performed to enhance predictive power by capturing meaningful relationships in the data, such as:

Temperature–Humidity interaction

Rainfall-to-Soil Moisture ratio

Evaporation Index based on temperature, wind speed, and sunlight

Soil Thirst indicator combining temperature and irrigation history

The model was implemented using a neural network built with TensorFlow and Keras, optimized with techniques like:

Batch Normalization

Dropout for regularization

Adaptive learning rate scheduling

Early stopping for better generalization

The system achieved strong performance with approximately 97.5% validation accuracy, indicating its effectiveness in predicting irrigation needs from structured data.

Key Features:

Strong feature engineering for improved predictions

Handling imbalanced datasets using SMOTE

Efficient neural network model for classification

High accuracy with reliable generalization

Technologies Used:

Python

TensorFlow / Keras

Scikit-learn

Imbalanced-learn (SMOTE)

Pandas & NumPy