Correlating Restaurant Health Code Violations and Online Customer Reviews via Machine Learning Methods

According to the Centers for Disease Control and Prevention, 48 million Americans become sick, 128,000 are hospitalized, and 3000 die of food- borne diseases annually. These severe outcomes often stem from poor restaurant health code adherence and inefficient governmental health code inspection and enforcement systems. This study adds context and analysis to existing large datasets, affording more actionable information from health inspections to both businesses and consumers. To meet these goals, this research investigated Yelp reviews and the City of Boston restaurant inspection data to identify relationships within and between the two datasets. We reviewed historical data from to 2017–2022 to balance the sample size and relevance from the vast scale of city data. Yelp data were selected first by the most popular restaurants in the city, and then by the restaurants with the most recorded code violations. Analytically, machine learning methods such as sentiment analysis, linear regression, and association rule mining are used to identify outliers, trends, and correlations within the data. Our findings serve as a tool to improve the efficiency and efficacy of Boston’s food code enforcement and aid restaurant survival. Consumers gain additional understanding of the inspection system, and restaurants gain information on where they can simultaneously improve their practices based on historical precedent and decrease the likelihood of escalating violations.