Integrated Deep Neural Networks and Ensemble Models for Accurate Soil Moisture Prediction and Fertilizer Guidance

Soil moisture serves as an essential agricultural production variable since it influences crop growth and agricultural water use as well as fertilization practices. Through accurate soil moisture forecasts farmers can apply fertilizer effectively which results in better resources management alongside improved crop production. The study introduces a hybrid approach that uses ensemble learning (Random Forest, XGBoost, LightGBM and Deep Neural Networks (DNN) for precise forecasting of soil moisture and fertilizer recommendation. In order to ensure dependable data, fertilizer and soil moisture data are label-encoded and normalized. The ensemble models are averaged with weighted averaging and 80–20 train-test splits are used for training. The fertilizer recommendation system, based on a multi-layer DNN, incorporates soil parameters, crop, and forecasted moisture levels to develop optimal suggestions. Experimental results indicate an excellent R2 value, highly improving the accuracy of soil moisture prediction. The DNN-based fertilizer recommendation system also shows high accuracy, thereby qualifying as an automated and scalable precision agriculture solution. The machine-learning based architecture allows for predictive decision-making, full automation, and increased efficiency in fertilizer recommendation and soil moisture monitoring. With the application of ensemble learning and deep-learning, this work promotes data-driven precision agriculture, fostering sustainable agriculture. The model developed gave accuracy of 92 % for soil moisture prediction and accuracy of 91 % for fertilizer recommendation.