9th International Conference on ICT for Sustainable Development

Authors - Aishwarya V. Kadu, KTV Reddy
Abstract - Inadequate soil fertility and the increasing demand for research in data-driven agricultural tools significantly impact crop productivity. To address this challenge, utilizing high-throughput computational algorithms, particularly within machine learning, becomes indispensable for efficiently examining soil and achieving precise predictions of fertility status. This approach facilitates decision-making to optimize soil fertility management. However, challenges arise when selecting essential soil property criteria for accurate fertility forecasting and soil nutrient analysis. Additionally, the inherent limitations within individual ML algorithms and the subjectivity in implementing expert modeling procedures can result in variations in model performance, particularly in predicting lower fertility levels and soil fertility classification. This paper offers an all-encompassing survey of the present landscape of Applications of ML in agricultural soil nutrient analysis, fertility prediction, and soil mapping in Wardha District, Maharashtra. Our study incorporates essential elements, including location maps, district-specific information, climate data, soil maps, soil mapping units, and soil site suitability classifications. Notably, the most frequently examined soil nutrients in Wardha District, Maharashtra, encompass Potassium (K), Organic Carbon (O.C.), Manganese (Mn), Phosphorus (P), Copper (Cu), Zinc (Zn), Potential Hydrogen (pH). Predominant ML algorithms used in this context include Random Forest and Naïve Bayes, with Support Vector Machines, also significantly improving agricultural practices in the area.