A deep learning approach for dealing with tabular data in crop classification

This paper presents the development of a deep learning approach for dealing with tabular data in crop classification. The input features are processed by slicing them and then transformed to capture the interaction between neighboring features. The combined features are transformed, and refined iteratively based on the learned weights using an adaptation of attention mechanism. To validate the effectiveness of the proposed deep learning approach, we conducted training on several deep neural networks like Multi-Layer Perceptron, ResNet and TabNet with the aim of accurately classifying crops. The experiments demonstrate that the proposed deep learning approach achieved an accuracy of 88%, precision of 87%, recall of 86%, and F-score of 86% in classifying crops compared to other deep learning approaches. Moreover, the results indicate that the proposed approach not only enhances the classification model but also potentially learns hidden patterns between the features.