Smart Agriculture: IoT-Driven Soil Monitoring and Fertilizer Recommendation with CNN and Bidirectional GRU Models

By scheduling fertilizer application, precision agriculture maximizes economic efficiency and environmental friendliness. However, many fertilizer algorithms ignore crop type, soil fertility, and environmental variables. No comprehensive, real-time soil fertility mapping system makes context-sensitive advice difficult. The technique involves gathering sensor data, preprocessing it, selecting characteristics, and training the model. The sensors detect many metrics, including humidity, light intensity, temperature, and CO2 levels. Preprocessing methods vary the baseline and smooth data, improving data quality. The nonlinear neural network Self-Organizing Feature Map (SOFM) was utilized to analyze multivariate data. The model was trained using CNN-BiGRU.The CNN-BiGRU model outperformed BiGRU and CNN with 92.45% accuracy. This indicates the model can accurately propose fertilizer using real-time data.This proposed solves fertilizer recommendation system problems by merging cutting-edge neural network models with real-time contextual data. The suggested strategy improves precision agricultural accuracy, efficiency, and context awareness.