Decision Support System for Crop Management using Convolutional Neural Network

Effective crop management is critical for enhancing agricultural productivity and sustainability. Traditional farming methods often face challenges in accurately forecasting yields, optimizing resources, and determining the right harvest timing. To address these challenges, this study develops a decision support system for rice farming in South India by utilizing deep learning techniques, specifically convolutional neural networks (CNN s), combined with dimensionality reduction methods. The system uses a comprehensive dataset that includes factors such as rainfall, temperature, soil type, and rice variety to predict rice yield. Additionally, it optimizes the use of resources such as water, fertilizer, and labor on the basis of expected yields. The proposed CNN model, coupled with squared exponential kernel PCA (SEKPCA), achieved a prediction accuracy of 83.12% and an F1 score of 0.63, enabling farmers to make informed decisions and improve productivity. The system also evaluates resource requirements, offering insights into water usage, fertilizer, and labor needs. This approach enhances agricultural efficiency, reduces waste, and promotes sustainable farming practices. By integrating advanced deep learning technologies into traditional agriculture, the system contributes to more precise yield predictions and better resource management, paving the way for improved crop management and sustainability in rice farming in South India.