Using deep neural networks for evaluation of soil quality based on VIS–NIR spectroscopy

The CNN model outperforms other regression algorithms in terms of accuracy and low predicting errors using non-preprocessed data, and the use of diverse soil properties for modeling spectral data can provide sufficient information on the advantages and disadvantages of VIS–NIR data in predictions.