Artificial intelligence and statistical methodologies applied to food safety, risk assessment and traceability of supply chains

The industry supply chain is nowadays facing the need for optimized and customized models useful to retrieve and pull together data on food safety, risk assessment and traceability.
Not integrated, sparse data from multiple sources and multiple study fields often leads to the incomplete and inaccurate evaluation of factors impacting the food safety, but at the same time this huge quantity of data provide the bases for statistical methods implemented in the food metrology. Within this field, repeatability and reproducibility are the two measures of precision which allowed for obtaining reliable outputs from common standard models. In this scenario, pathogens are one of the main factors undermining food safety and traceability (toxic compounds, allergens and bacterial pathogens). A link between pathogen related metadata and food shelf-life is fundamental to understand and monitoring changes in physicochemical properties at specific spoilage time points.

A systematic collection of multi-integrative data from food metrology and supply chain metadata will be the basic requirement for subsequent refined analyses.
The ReCaS DataCenter facility will host all the useful data, linked metadata and will provide the suitable environment to physically run the AI algorithms and biostatistics pipelines.
A systematic literature search will provide a first mock data set of harmful pathogens, toxic compounds and allergens identified in manufacturing processes for both traditional and novel foods that will be used for the AI training.
Standardization and harmonization strategies on gathered datasets, inclusive of toxic compounds, bacterial pathogens and allergens, will be performed to grant stable inputs with reduced inter-sample variations and methodological bias.
A machine learning algorithm will work on batches of data relative to traditional and novel foods to ensure an efficient supply chain management at different levels and it will be specifically run on product shelf-life large-scale heterogeneous datasets.
An ad hoc SQL database will specifically collect and integrate metrology, food safety and bacterial taxa metadata (inclusive of those relative food safety in terms of potential pathogenicity, toxic compounds and allergens). The different set of metadata matrices will form the structure of the interconnected tables through primary key elements.

The data economy approach (IoT, Big Data, Edge Computing, etc.) used on food supply chains, both for traditional and novel foods, will support food safety, risk assessment and traceability.
Evaluation of the reliability measurement and confidence results (suitable for data FAIR principles) obtained from the application of AI algorithms.
Creation of an SQL database on food traceability data obtained by implementing an IoT platform using a specific AI algorithm in real time, stored on the ReCaS DataCenter (at University of Bari).
Creation of a new platform on the ReCaS DataCenter, aimed at implementing and sharing at the national level the here produced data together with those carried out by WP1 and WP2.