Enhancing Soil Fertility Prediction Through Advanced Modelling: Agrimind Intelligent Fertility Predictor

Soil fertility prediction (SFP) is a critical process that estimates nutrient availability in soil, directly impacting agricultural productivity and crop health. Achieving accurate SFP is essential for optimising agricultural productivity. However, traditional models face significant challenges due to the complex nature of soil composition, dynamic environmental processes and limited availability of high quality data. To address these issues, this study proposes the AgriMind Intelligent Fertility Predictor (AMIFP), an advanced predictive model that combines innovative preprocessing and deep learning techniques to enhance SFPs. This methodology introduces a novel preprocessing approach, SMOTEImputeScaler (SIC), which addresses missing data, normalises features and mitigates class imbalance by generating synthetic samples near class boundaries. Additionally, introduces the model Attentive LSTM Aware Dense Network (ALADN), an advanced modelling framework that ensures robust feature extraction and effective classification, ultimately enhancing the accuracy and reliability of SFPs. To optimise hyperparameters effectively, this work introduces Genetic probabilistic Search Hyperparameter Optimization (GPSHO), which improves model performance by exploring complex hyperparameter spaces while reducing computational overhead. Experimental results demonstrate that the AMIFP model significantly outperforms existing approaches, achieving 98.65% accuracy, 98.73% precision, 98.65% recall, 98.70% F1, and mean square error (MSE) of 0.007, indicating that the proposed AMIFP model offers a robust and reliable solution for SFP, aiding agricultural decision-making processes.