Application of Machine Learning Algorithms for Monitoring of Spoilage of Cow’s Milk Using the Cheap Gas Sensor

Abstract

Cow’s milk is very sensitive to spoilage processes that could negatively affect its quality. Several methods have been developed for the analysis of milk spoilage, and in recent times solutions using e-nose are more and more common. This paper analyzes the possibility of using the Bosch BME688 (BioMedical Engineering) sensor, which can detect different gases by analyzing their different electronic signatures. Samples of fresh, pasteurized and UHT (Ultra High Temperature) sterilized milk, which were at room temperature (22°C) during the 4-day experiment, were tested. The obtained data for every day were used to train nine models based on different classification algorithms (K Nearest Neighbor, Logistic Regression, Support Vector Machine, Naive Bayes, Decision Tree, Random Forest, Gradient Boosted Trees, Probabilistic Neural Network) with the aim of distinguishing the type of milk with known time after opening, i.e. milking. The focus of the research was on the analysis of classification algorithms that can be used for the classification of milk with regard to the standing time at room temperature. The manufacturers of the sensor itself enabled the use of neural networks exclusively in the BME AI-Studio program, which is used for configuration and training of the model, and their decision is correct in most cases.

Topics & Concepts