Prediction of Crop Yield using Deep Learning and Dimensionality Reduction Approaches

Abstract

India has agriculture as its primary occupation. IBEF (Indian Brand Enquiry Foundation) estimates that 58% of Indians who live in rural areas depend on agriculture. Agriculture plays an important role in the life span of human being not only for the survival but for the better economic growth of the country too. Field-level crop yield prediction (CYP) is essential for quantitative and economic analysis, enabling the formulation of agricultural commodity plans, import-export strategies, and increasing farmer incomes. Crop breeding traditionally demands significant time and resources, making CYP a critical tool for forecasting higher crop production. This study introduces an effective approach combining deep learning (DL) and dimensionality reduction (DR) techniques for CYP, focusing on regional crops in India. The methodology comprises three phases: preprocessing, dimensionality reduction, and classification. Initially, agricultural data from South India is collected and preprocessed through data cleaning and normalization. Subsequently, dimensionality reduction is performed using squared exponential kernel-based principal component analysis (SEKPCA). Finally, a weight-tuned deep convolution neural network (WTDCNN) is employed to predict high crop yields profitably. Simulation results demonstrate that the proposed method achieves superior performance compared to existing techniques, with an impressive accuracy of 98.96%. The novelty of this approach lies in its integration of DL, DR, and WTDCNN, offering precise crop yield predictions tailored to Indian regional crops.