Machine learning in food authentication: A sustainable solution to global food challenges

Authenticity of a food product is one of the major global food challenges. Ensuring consumers about origin, variety, composition, processing method, shelf life and purity of a food product is challenging for food producers and manufacturers. Traditional methods of food authentication are time consuming and labour intensive and can handle only limited datasets. While machine learning (ML) algorithms can serve as a sustainable and rapid approach through handling of larger datasets thereby overcoming the demerits of traditional methods of food authentication. However, no comprehensive review article covering all published literature of last 5 years is available till now. Therefore, this review was designed to discuss the implementation of various ML algorithms in authentication of various food commodities such as honey, beverages like beer, juices and wine, edible oils, seafoods, spices, cereals, meat, fish, egg, fruits, vegetables, milk and milk products etc. Neural networking (NN), decision trees, random forests, support vector machines (SVM), principal component analysis (PCA) and autoencoder with amalgamation with spectroscopic and chromatographic techniques provide a robust and rapid approach for building trusts between producers and consumers and fair-trade practices. ML has transformative potential in addressing food integrity challenges with profound successful implementation in food industry to enhance the food safety and quality. This information will help researchers and regulatory bodies to ensure consumers and traders for food quality and safety.