A systematic review of explainable artificial intelligence for spectroscopic agricultural quality assessment

The introduction of complex machine learning models has greatly improved the accuracy and practical use of spectroscopic analyses in agriculture. However, users often struggle to understand how these models operate internally or how specific features contribute to the predictions. This lack of clarity can hinder innovation in agricultural spectroscopy, especially in selecting appropriate spectral wavelengths for domain specific applications or designing portable and low-cost devices. Therefore, the integration of Explainable Artificial Intelligence (XAI) techniques is essential to address these challenges in the agricultural sector. This review systematically examines recent advancements in XAI techniques and highlights their substantial effects on enhancing spectroscopic models for assessing the quality of agricultural and food products. This study also highlights current challenges and explores prospects, emphasizing how these innovative techniques can support more advanced and widely adopted applications within the agricultural industry.