Classification of Starch Adulteration in Milk Using Spectroscopic Data and Machine Learning

Abstract

In this comprehensive research, an in-depth evaluation of several machine learning algorithms, including Multilayer Perceptron, IBk, KStar, M5Rules, and RandomForest, is conducted to ascertain their effectiveness in detecting adulteration in milk products using spectroscopic data. The algorithms were rigorously deployed and assessed through a series of controlled experiments involving both raw and adulterated milk samples. Notably, IBk and KStar algorithms emerged with a perfect accuracy rate of 100% in identifying adulteration, highlighting their superior capability in this domain. Additionally, the Decision Table algorithm also performed exceptionally well, achieving a remarkable correlation coefficient of 0.9871. These promising results emphasize the undeniable potential of machine learning algorithms as reliable and precise tools for detecting adulteration in milk. Such technological interventions play a critical role in elevating the safety and quality standards of milk and milk-based products in the market. Moreover, the deployment of these advanced machine-learning techniques provides an invaluable layer of consumer protection, plays a significant role in combating widespread fraudulent practices in the milk industry, and ensures compliance with stringent food safety standards. These methodologies could be indispensable for both industry players and regulatory bodies, significantly contributing to the safeguarding of public health.