Development of a model for estimation of soil parameters using deep learning

Earlier, crop cultivation was undertaken based on farmers' hands-on expertise. However, farmers cannot choose suitable crops based on soil and environmental factors. Therefore, this research proposed a Novel EDLAM: Efficient Deep Learning with Attention Mechanism Model to predict the appropriate yield based on the soil moisture and texture. In that model, to pre-process the data, this research proposed a Progressed Synthetic Minority Over-sampling _Support Vector Machine (PSMOTE_SVM) model to reduce the noise. Then, this research presented a Boosted Sequential Forward Feature Selection (BSFFS) to extract the best features. Finally, this research proposed Bidirectional Long Short Term Memory with a Global Attention Mechanism to predict the crop. As a result, our study provides 95.1% accuracy, 95.74% precision, 94.91% recall and 95.06% F1 score.