Symbolic Deterministic Finite Automata-based Automated Fertilizer Prescription

Abstract

Sustainable agriculture requires the use of an adequate amount of fertilizers. In this research work, initially, an attempt was made to classify soil fertility using machine learning-based classifiers. To overcome the drawbacks of machine learning-based classifiers, this research uses Symbolic Deterministic Finite Automata (SDFA) for soil fertility classification. The proposed method classifies soil fertility as LOW, MEDIUM (MED), or HIGH using the levels of four soil parameters, including pH, Electrical Conductivity (EC), Organic Carbon (OC), and Nitrogen (N). The proposed approach was assessed using Sentinel-2 remotely sensed data and laboratory-measured soil-health data. The experiments' outcomes show that the proposed approach effectively classifies soil fertility. The accuracy achieved using Sentinel-2 data was 100%, while the accuracy gained using laboratory-measured data with four and twelve soil parameters were 100% and 98.37%, respectively. The results of soil fertility classification were used to recommend fertilizers. © 2023 IEEE.