Soil Moisture Quantity Prediction using Optimized Deep Learning Supported model for Sustainable cultivation of Groundnut plant

Abstract

In the hydrological cycle, soil moisture is the predominant component for regulating the evapotranspiration, vegetation production and runoff. The process of soil moisture is directly essential for farmers involved in sustainable agriculture and their related social and economic activities. This soil moisture is the potential soil indicators that aids in determining the conditions of early water deficit in the earlier stage, decision making about crop planning, policy making by government, food security conditions and uncertainty in crop yield. In specific, the soil moisture prediction of groundnut plant is essential as it is the principal crop cultivated by significant number of farmers in the semi-arid regions of Southern India. In this paper, Generative Adversarial Networks-based Deep Learning Model with Sailfish Optimization Algorithm (GAN-DLMSOA) for predicting the quantity of soil moisture. This GAN-DLMSOA is proposed with the potential characteristics of GAN that aids in significant prediction through the comprehensive exploration of exhaustive parameters associated with the problem from the dataset to render accurate prediction process. In specific, Sailfish Optimization Algorithm (SOA) is included to perform remarkable feature optimization that contributes towards accurate soil moisture prediction. SOA is also used to train the GANs with fast convergence degree for accomplishing the task of prediction and facilitate precise results. The data perturbation-based stability analysis of the proposed GAN-DLMSOA is also conducted with different weather conditions to perceive the stability of the incorporated model. The proposed GAN-DLMSOA confirmed with a mean RMSE value of 0.04065 and a  maximized accuracy level of 99.66% compared to all the baseline soil prediction schemes.