SAIFeR-embedded system design and implementation of soil analysis and intelligent fertiliser recommendation system

In Pakistan, the agriculture sector contributes 23% to the Gross Domestic Product (GDP) while ensuring food security and rural development. The persistent challenges of crop cultivation and productivity can be mitigated with the application of suitable fertilisers. However, due to inefficient fertiliser management, the crop yield is below the regional benchmarks. In some areas, the soil health has declined due to incorrect fertiliser usage causing nutrient imbalance and negative environmental impact. To address these challenges, an intelligent fertiliser recommendation system (SAIFeR) has been proposed that enables farmers to determine fertiliser requirements for the optimum growth of specific crops. An embedded electronic system measures the soil parameters which are then subjected to machine learning algorithms for a comprehensive analysis and resultant fertiliser recommendations. One of our proposed models, Expanded-Input Multi-Layer Perceptron with LeakyReLU, performed well for predicting across the 3 widely used fertilisers (Urea, DAP and SOP) and provides accurate prediction with an average RMSE of 12–14% against the ground truth. This data-driven application of fertilisers would empower local farmers to adopt sustainable farming and minimise adverse environmental impacts due to overdose of fertilisers. This research provides a foundation for further advancements in precision agriculture, and smarter and environmentally responsible farming practices.