Predicting Irrigation Need

Achieved 95.9% accuracy in the Kaggle Predicting Irrigation Need challenge as part of the Digilians AI & Machine Learning program. The task required applying deep learning on tabular data, which introduced additional complexity and demanded careful feature engineering. To address this, a robust pipeline was developed to transform raw agricultural data into highquality inputs, including water stress indicators, temperature–humidity interaction features, rainfall-to-irrigation ratios, rolling trend features, and evapotranspiration-based indices. This approach significantly improved model performance, enabling the model to surpass the 95% accuracy threshold. The project demonstrates strong capabilities in deep learning, feature engineering, and solving real-world problems in smart agriculture.