An intercontinental machine learning analysis of factors explaining consumer awareness of food risk

Abstract

Food safety is a common concern at the household level, with important variations across different countries and cultures. Nevertheless, identifying the factors that best explain similarities and differences in consumer awareness pertaining to this topic is not straightforward. Starting from a questionnaire administered in seven countries from four continents (Argentina, Brazil, Colombia, Ghana, India, Peru, and the United Kingdom), we present an analysis of the answers related to food safety concerns, aimed at identifying possible explanatory factors. As classical statistical approaches can be limited when dealing with complex datasets, we propose an analysis with machine learning techniques, that can take into account both categorical and numerical values. With the questionnaire as a base, we task a machine learning algorithm, Random Forest, with predicting consumers’ answers to the target questions using information from all other answers. Once the algorithm is trained, it becomes possible to obtain a ranking of the questions considered the most important for the prediction, with the top-ranked questions likely representing explanatory factors. Top-ranked questions are then analyzed using a Random Forest regression algorithm, to test possible correlations. The results show that the most significant explanatory variables of safety concerns seem to be estimates of carbon footprints and calories associated with food products, and primarily with beef and chicken meat. These results tend to indicate that people who are most concerned about food safety are also those who are highly aware of environmental and nutritional impacts of food, hinting at differences in food education as a possible underlying explanation for the data.