Analytic algorithms developed and refined at Carnegie Mellon School of Computer Science Auton Lab are helping to identify food safety problems before they impact human health. The lab specializes in the research of new approaches to statistical data mining.
Lab co-director Artur Dubrawski, PhD, a senior systems scientist, says that the analytic algorithms were developed to enable food safety and public health analysts to more systematically screen their data for evidence of interest “so that no important clue is ever missed.” The system partially automates the process of analyzing large amount of diverse data “so that good outcomes can scale with the data,” he says.
Using machine learning and key data points, the analytic algorithms can, for example, connect two cases of foodborne illnesses that occur in different parts of the country and then trace them back to the source of the contamination.
“In our analyses we primarily rely on data routinely collected by the government agencies in the process of their food safety and public health oversight duties,” Dr. Dubrawski says. The USDA’s Food Safety and Inspection Service, for example, collects multiple types of data, ranging from veterinary health assessments of cattle received at slaughterhouses to results of microbial testing of samples of food taken at various stages of production, logs of regulatory inspections of food processing facilities by food safety inspectors, and consumer complaints about food. In addition to data collection by the FDA, the CDC also collects data about human cases of potential foodborne diseases to inform epidemiologic tracking and other analyses, he says.
Dr. Dubrawski notes that the algorithms help detect patterns in that type of data. “We improve visibility of food safety concerns by facilitating identification of informative patterns in data, such as calling off a batch of ground meat suspected of being contaminated with Salmonella before it ships to distributors.”
The algorithms are used to discern patterns of pathogens at various stages of production, helping “to prevent bad things from happening in operations as well as at more strategic, policy-making levels,” he explains.
In addition to tracking pathogens such as Salmonella, E. coli, and Listeria, similar efforts are underway to track chemical and physical contaminants and allergens, he says.
Dr. Dubrawski has more than 25 years of experience researching machine intelligence and its applications. His work has resulted in many uses of analytic solutions and software for various government and industrial applications.