Food safety remains a critical public health conceated food causing significant morbidity and mortality worldwide. Accurate and transparent risk assessment is essential to safeguard consumer health and support evidence-based regulatory decisions. This study introduces a comprehensive and novel AI framework that leverages machine learning, deep learning, and transformer-based models to classify food safety risks using the Rapid Alert System for Food and Feed (RASFF) dataset, enhanced through advanced data enrichment techniques. The enriched dataset addressed challenges such as short explanations and class imbalances, resulting in significant performance improvements across all model categories. Transformer-based models, including BERT with an accuracy of 0.978 and RoBERTa with an accuracy of 0.979, outperformed traditional machine learning methods such as logistic regression, which achieved an accuracy of 0.954, and SVM, which reached 0.959, as well as deep learning models like LSTM, which obtained 0.971, and BiLSTM, which achieved 0.97 3. Explainable Artificial Intelligence (XAI) techniques were applied to uncover critical insights into the models’ decision-making processes. Influential features such as salmonella, aflatoxins, and listeria were identified, enhancing model transparency and interpretability. The framework’s combination of data augmentation and XAI improves predictive accuracy while ensuring interpretability, making it suitable for real-time risk prioritization in regulatory surveillance. This study demonstrates the potential of AI-powered approaches for food safety risk assessment, bridging predictive accuracy with transparency to support more reliable and actionable food safety management systems. Graphic

Spoilage in processed meat products, such as poultry and pork sausages, presents significant challenges for food safety, quality control, and waste reduction. This study presents a machine learning-based framework to classify spoilage intensity levels using sensory, physicochemical, and microbiological features. To overcome limitations caused by small datasets, we applied synthetic data augmentation using a tabular variational autoencoder (TVAE) to generate high-fidelity samples that enhance model generalization. Additionally, traditional oversampling techniques such as SMOTE and ADASYN were employed for comparative purposes and to further address class imbalance issues. Seven machine learning classifiers were evaluated logistic regression, support vector machine, K -nearest neighbors, random forest, gradient boosting, voting classifier, and multilayer perceptron. The best classification performance was achieved when models were trained on GAN-based synthetic data and tested on real samples. For poultry sausage spoilage prediction, the gradient boosting classifier reached the highest accuracy of 97%. For pork sausages, random forest achieved the highest accuracy of 95%. These results confirm the effectiveness of data augmentation in improving predictive robustness. To ensure model transparency, we integrated explainable AI techniques SHAP and LIME into the pipeline. These analyses revealed that sampling time, CO $$_2$$ 2 concentration, pH, and microbial species such as Lactobacillus curvatus and Leuconostoc carnosum were among the most influential features in spoilage prediction. The combination of synthetic data generation and interpretable machine learning enables a reliable, scalable, and explainable approach to spoilage classification. This methodology has strong potential for enhancing quality control systems in the meat industry while reducing waste and improving safety along the food supply chain. Graphical

Food spoilage prediction is a critical challenge in food safety and quality management, particularly for meat products exhibiting complex microbiological and biochemical dynamics. This study presents an explainable machine learning framework for predicting sausage spoilage intensity using volatile organic compound (VOC) profiles and physicochemical parameters, enhanced through Generative Adversarial Network (GAN)-based data augmentation. The proposed framework integrates interpretable machine learning models, random forest, gradient boosting, logistic regression, multi-layer perceptron, and a voting classifier with the TVAESynthesizer generative model to address data scarcity and imbalance in experimental food datasets. SHapley Additive exPlanations (SHAP) were employed to quantify the contribution of individual VOCs and physicochemical variables to spoilage classification, thereby enhancing model transparency and biological interpretability. Results revealed that GAN-augmented datasets substantially improved predictive performance compared to models trained on original data. For poultry sausages, the gradient boosting and random forest models achieved an accuracy of 0.92, while for pork sausages, both models reached an accuracy of 0.89. In addition, fold-wise regeneration of synthetic data during cross-validation yielded highly stable model performance, with Random Forest and Gradient Boosting achieving accuracies and F1-scores above 0.90 for poultry sausages, and consistently robust peak accuracies around 0.89 for pork sausages, confirming the reliability of the GAN-augmented training strategy. SHAP analysis revealed that Sampling Time and pH are the dominant predictors of spoilage for both poultry and pork sausages, with alcohol-related volatile compounds such as 1-propanol, 2-butanone, and 2-butanol driving predictions in poultry, and ethyl acetate, methanethiol, dimethyl sulfide, and hexanal playing a major role in pork spoilage classification. Overall, integrating generative modeling with explainable AI significantly improves both predictive accuracy and interpretability. The proposed framework offers a sustainable, data efficient, and interpretable solution for real time, non-destructive monitoring of meat freshness and quality. Graphic

Foodborne pathogens remain a major public health concern, and the early identification of unsafe conditions is essential for preventive control. Routine inspections generate rich textual and structured data that can support real-time assessment of pathogen-related risk. The objective of this study is to develop an explainable transformer-based framework for predicting food safety inspection grades using multimodal inspection data. We combine structured metadata with unstructured deficiency narratives and evaluate classical machine learning models, deep learning architectures, and transformer models. RoBERTa achieved the highest performance (F1 = 0.96), followed by BiLSTM (F1 = 0.95) and LightGBM (F1 = 0.92). SHapley Additive exPlanations (SHAP) analysis revealed linguistically meaningful indicators of pathogen-related hazards such as temperature abuse, pests, and unsanitary practices. The findings demonstrate that transformer-based models, combined with explainable AI (XAI), can support pathogen-oriented monitoring and real-time risk assessment. This study highlights the potential of multimodal AI approaches to enhance inspection efficiency and strengthen public health surveillance.

The microbial spoilage of meat products poses a significant challenge to food safety and shelf-life, primarily due to the proliferation of spoilage-related bacteria under varying environmental conditions. Among beneficial microorganisms, Lactococcus lactis species are notable for their ability to produce nisin, a natural antimicrobial that enhances the microbial stability of meat. This study presents an interpretable machine learning framework to predict the relative abundance of Lactococcus in meat using 16S rRNA sequencing data, enhanced by synthetic data generation and explainable AI. A dataset comprising microbial profiles from rump steak samples across seasonal and temporal variables was expanded using a Generative Adversarial Network (GAN), addressing data scarcity. Multiple regression models, including Artificial Neural Networks, Random Forest, Gradient Boosting, and k-Nearest Neighbors, were trained and evaluated. The ANN model achieved the highest performance (\((R^2 = 0.89)\)) when trained on synthetic data and tested on real samples. Explainable AI analysis revealed Lactobacillus, Carnobacterium, and seasonal variables as the most influential predictors. Notably, seasonal shifts significantly impacted Lactococcus abundance, with higher levels observed in autumn. This study demonstrates the effectiveness of GAN-augmented data and interpretable machine learning in modeling microbial behavior, providing a scalable, cost-efficient tool for predictive microbiology. The approach not only enhances our understanding of microbial co-occurrence patterns but also supports the development of intelligent, data-driven strategies for natural food preservation.