Crop Recommendation System based on Soil and Environmental factors using Graph Convolution Network and Graph Neural Network: A systematic Literature Review

The rapid advancement of agricultural technology has led to an increasing reliance on data-driven approaches for optimizing crop yields and resource management. In this research paper, we present a comparative study between two graph-based models, namely the Graph Convolutional Network (GCN) and the Graph Neural Network (GNN), for the task of crop recommendation based on various environmental factors. Our study focuses on leveraging a dataset encompassing critical parameters such as nitrogen level, potassium level, phosphorus level, temperature, local humidity, pH of soil, and rainfall, with the target being the selection of a suitable crop for a given season. To address the complexity and interdependencies of the provided dataset, we harness the power of graph-based models that are adept at capturing intricate relationships among features. Both the GCN and GNN are well-suited for such tasks due to their ability to process structured data represented as graphs. We adopt a supervised learning approach where the input features are organized as nodes in a graph, and edges represent potential associations between these features. The objective is to predict the most appropriate crop label for a given set of environmental conditions. Our experimentation involves pre-processing the dataset to construct an appropriate graph representation. We evaluate the performance of both models using metrics such as accuracy, precision, recall, and F1-score to ascertain their effectiveness in recommending crops. Additionally, we investigate the model's ability to generalize by employing techniques like k-fold cross-validation to mitigate overfitting concerns

In conclusion, this research contributes to the ongoing exploration of graph-based models in agricultural applications. By showcasing the comparative performance of GCN and GNN in the context of crop recommendation, we offer valuable insights into the potential of these models for aiding precision agriculture practices. Our findings underline the importance of choosing an appropriate graph-based model based on the nature of the dataset and its inherent relationships, leading to more informed decisions in crop management and resource allocation.