<p>The deterioration of food freshness, particularly meat, due to bacterial growth is a major concern for food safety. This study aimed to detect the growth of the harmful bacteria using machine learning algorithms and analyze the influence of other bacteria on harmful bacteria growth through explainable artificial intelligence method analysis. Using genetic sequencing to study bacterial diversity, bacterial composition in pork meat samples was analyzed. Bacteria with relative abundances below the sensitivity thresholds of 0.1%, 0.25%, and 0.5%, which indicate their presence percentages in the samples, were excluded from the dataset for evaluation. This approach enabled a focus on more dominant bacterial populations and was utilized to assess the growth of harmful bacteria alongside traditional culture-based methods. Statistical tests revealed significant relationships between bacterial species, notably a negative correlation between one type of bacteria often found in meat spoilage and another potentially harmful type (correlation coefficient: <InlineEquation ID="IEq1"> <EquationSource Format="TEX">$-0.385, p < 0.05$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mo>-</mo> <mn>0.385</mn> <mo>,</mo> <mi>p</mi> <mo><</mo> <mn>0.05</mn> </mrow> </math></EquationSource> </InlineEquation>). Machine learning algorithms, particularly a prediction model based on probabilities, showed marked improvement in performance when the day variable of testing was included, achieving an accuracy rate of 0.89 for the most stringent dataset (0.5%). In addition, an explainable artificial intelligence analysis indicated that a high presence of bacteria associated with spoilage was linked to a reduction in the growth of harmful bacteria on specific laboratory growth media. Integrating genetic sequencing method, machine learning algorithms, and explainable artificial intelligence analysis offers valuable insights into bacterial growth dynamics, representing a breakthrough in predictive food safety. This approach contributes significantly to the development of effective food safety and quality assurance strategies within the food industry by enhancing quality control, extending product shelf life, and reducing financial losses associated with spoilage and recalls.</p>

Machine learning-based prediction of Clostridium growth in pork meat using explainable artificial intelligence

期刊论文
发表日期：

2025年1月21日

查看全文

Volkan Ince,
Mohamed Bader-El-Den,
Jack Alderton,
Farzad Arabikhan,
Omer Faruk Sari,
Annette Sansom

摘要

The deterioration of food freshness, particularly meat, due to bacterial growth is a major concern for food safety. This study aimed to detect the growth of the harmful bacteria using machine learning algorithms and analyze the influence of other bacteria on harmful bacteria growth through explainable artificial intelligence method analysis. Using genetic sequencing to study bacterial diversity, bacterial composition in pork meat samples was analyzed. Bacteria with relative abundances below the sensitivity thresholds of 0.1%, 0.25%, and 0.5%, which indicate their presence percentages in the samples, were excluded from the dataset for evaluation. This approach enabled a focus on more dominant bacterial populations and was utilized to assess the growth of harmful bacteria alongside traditional culture-based methods. Statistical tests revealed significant relationships between bacterial species, notably a negative correlation between one type of bacteria often found in meat spoilage and another potentially harmful type (correlation coefficient: $-0.385, p < 0.05$ $- 0.385, p < 0.05$ ). Machine learning algorithms, particularly a prediction model based on probabilities, showed marked improvement in performance when the day variable of testing was included, achieving an accuracy rate of 0.89 for the most stringent dataset (0.5%). In addition, an explainable artificial intelligence analysis indicated that a high presence of bacteria associated with spoilage was linked to a reduction in the growth of harmful bacteria on specific laboratory growth media. Integrating genetic sequencing method, machine learning algorithms, and explainable artificial intelligence analysis offers valuable insights into bacterial growth dynamics, representing a breakthrough in predictive food safety. This approach contributes significantly to the development of effective food safety and quality assurance strategies within the food industry by enhancing quality control, extending product shelf life, and reducing financial losses associated with spoilage and recalls.