Abstract: Soil fertility plays a critical role in agricultural productivity and food security. With increasing global population and diminishing arable land, optimizing crop yield through intelligent soil management has become imperative. This work presents an enhanced deep learning-based model for soil fertility assessment and crop recommendation using a multi-parameter agricultural dataset. The proposed system integrates key soil features such as N, P, K, pH, EC, micronutrients, and organic carbon, followed by feature scaling and optimized training to achieve improved classification performance. Experimental results demonstrate an accuracy of 88.06% with a macro F1-score of 0.72, indicating strong predictive capability for major soil fertility classes. The confusion matrix confirms high precision and recall for Classes 0 and 1, while Class 2 shows lower performance due to limited representation. Overall, the model provides an efficient and data-driven solution for agricultural decision-making, helping farmers and agronomists select suitable crops based on soil characteristics.

Keywords: Deep Learning, Fertilizer Recommendation, Crop Recommendation, soil data; soil analysis

Downloads: | DOI: 10.17148/IARJSET.2025.1211053