AgriNet: A Soil Fertility Dataset for Deep Learning-Based Fertilizer Recommendation in Rice Cultivation

The AgriNet dataset contains laboratory-tested soil fertility records collected from agricultural fields in Bareilly district, Uttar Pradesh, India. Soil samples were analyzed at the Soil Testing Laboratory, Krishi Vigyan Kendra (KVK), Indian Veterinary Research Institute (IVRI), Bareilly, using standard soil analysis procedures. The dataset was developed to support research in precision agriculture, soil fertility assessment, fertilizer recommendation, and machine learning-based decision support systems. The dataset consists of 258 soil samples, each containing measurements of soil pH, Electrical Conductivity (EC), Organic Carbon (OC), available Nitrogen (N), available Phosphorus (P), available Potassium (K), Sulphur (S), Zinc (Zn), Boron (B), Iron (Fe), Manganese (Mn), Copper (Cu), along with cultivated land area (Acres and Bigha). These laboratory-derived soil properties were extracted from official Soil Health Cards and manually digitized following data validation and quality assurance procedures. The target fertilizer recommendation variables (Target_Urea, Target_DAP, and Target_MOP) were generated using the Soil Test Crop Response (STCR) methodology for alluvial soils under high-yielding rice cultivation with a target yield of 5.5 tonnes per hectare. The recommendations were computed programmatically using standard STCR equations to ensure consistency and reproducibility. The dataset has been cleaned by removing duplicate records, validating numerical values, handling missing observations, and excluding statistical outliers using the Z-score (±3 standard deviations) criterion. The final dataset is provided in CSV format together with documentation describing each variable and its measurement units. This dataset can be used for developing and benchmarking machine learning, deep learning, multi-output regression, precision nutrient management, soil fertility prediction, and fertilizer recommendation models. It also serves as a valuable resource for researchers, agricultural scientists, educators, and developers working in digital agriculture and decision-support systems.