Rapid Detection of Microbial Load in Chicken Meat via Gas Sensor Array and Machine Learning

Today, both classical culture methods and modern molecular and rapid analysis methods are used to determine microbial growth in food products. Conventional methods involve long turnaround times, high labour and costs, and reliance on expert interpretation; thus, there is a need for low-cost, digitally integrable, and mobile-adaptable alternatives. In this study, a gas sensor matrix integrated with machine learning algorithms was utilised to rapidly predict microbial growth in chicken meat, enabling estimation of microbial load under controlled storage conditions. Based on results from models built with different heater profiles and algorithms, the Gradient Boosting Regressor emerged as the most effective approach for regression, while the Support Vector Machine achieved the highest classification performance. In terms of heater profiles, the temperature modulation of HP-324 was observed to be the most suitable for this type of study. The proposed models, with their high predictive accuracy, are intended not to replace conventional microbiological methods, but to serve as a rapid pre-screening layer that complements and supports them in food quality control workflows. The results demonstrate that sensor-based predictive models can provide a non-contact, rapid, non-destructive, and scalable solution for assessing microbial growth, thereby contributing significantly to food safety monitoring and shelf life prediction.