Machine Learning-Driven Raman Spectroscopy Techniques for Rapid Detection of Chemical Compounds and Contaminants in Foods

Raman spectroscopy is an emerging nondestructive and rapid detection technology and has been used for food analysis. However, the time-consuming and low-throughput Raman spectral analysis is the major hurdle for its wide implementations. Innovations and advancements in artificial intelligence–based data analytics such as machine learning with reliable test accuracy promise the improving potential of using this technique for numerous domains of food science and engineering applications. This chapter provides an overview of the application of different machine learning algorithms(random forest, 2D-CNNs, etc.)combined with Raman spectral techniques in the rapid analysis of chemical compounds and contaminants in foods. The underlying principle of Raman spectroscopy as well as the advantages and limitations of machine learning algorithms are addressed. Data analysis such as standardized protocol and methodology of data file size, partition method of training and test dataset, and cloud database and online web server for the application, particularly for online/real-time detection, are discussed. This knowledge will fundamentally help develop rapid detection methods using the machine learning–driven Raman spectroscopy technique.