Evaluating the Effectiveness of Soil Amendment Practices on Groundwater Recharge and Irrigation Sustainability in Semi-Arid Farmlands

Abstract

Irrigation requirement prediction plays a crucial role in increasing the efficiency of farming and sustainable utilization of groundwater resources in semi-arid lands. Prediction of irrigation requirements is difficult because of various factors involved in the process like interaction between soil type, climate, and crop type. In this research, Hybrid Feature Selection-Ransom Forest (HFS-RF) scheme is proposed to improve the irrigation requirement prediction using the Irrigation Water Requirement Prediction Dataset. Firstly, preprocessing is done in the form of dealing with missing values, duplicate removal, label encoding, and Min-Max normalization. Further, HFS scheme, which is based on the use of correlation analysis and RF feature importance, is utilized to select the optimal features having an impact on irrigation demand. The selected features are used to train thevRFclassifier, which classifies the irrigation requirements into Low, Medium, and High. The overall accuracy obtained from the proposed scheme is 97.00% with 98.00% recall, 97.00% precision, and 97.00% F1- score. Thus, the proposed scheme outperforms other schemes. The experimental results clearly reveal the effectiveness of the HFS-RF scheme in enhancing the prediction accuracy and decreasing the computational complexity