A Performance Assessment of Harvest Restoration in the Marathwada District over Weather Based Surface Rainfall Variance Employing Enhanced Optimizers

Abstract

For highly vulnerable regions such as Marathwada, correct estimation of crop compensation using periodic rainfall variability is essential for agricultural policy formulation and disaster prevention. The correlation between rainfall and yield has been studied extensively, but only a few studies have compared and used modern stochastic optimizers for predicting compensation. This research investigates the prediction performance of three recently-developed optimizers, AdaBelief, Muon, and Lion, in crop compensation with rainfall deviation and seasonal rainfall data as inputs. The model performance is assessed by the resulting coefficient of estimation (R2) and the mean of the absolute error (MAE).