Predictive Analytics Using Text Classification for Restaurant Inspections

Abstract

According to the Center for Disease Control (CDC), there are almost 48 million people affected by foodborne diseases in the U.S. every year, including 3,000 deaths. The most effective way of avoiding food poisoning would be its prevention. However, complete prevention is not possible, therefore Public Health departments perform routine restaurant inspections, combined with the practice of inspecting specific restaurants once a disease outbreak is identified. Following other health applications (e.g., prediction of a flu outbreak using Twitter), we use social media and a predictive analytics approach to identify the need for targeted visits by city inspectors.

Analysis of the Impact of COVID-19 on Education Based on Geotagged Twitter

“…Sadilek et al [15] and Wang et al [21] perform studies to detect and predict foodborne illness using respectively Twitter and Yelp. They analyze the tweets using language features and language models.…”

“…In this paper our focus is on Twitter. In fact, there is a wealth of millions of tweets with different times, locations, and users, in what constitutes a stream of snapshots throughout the pandemic, which we analyze in this paper, following the lead of so many others who have analyzed tweets in a variety of domains [3,4,7,9,10,12,13,15,16,21,23].…”

Ontology-based Instance Matching for Geospatial Urban Data Integration

“…We are also currently working on performing predictive analytics using text classification for restaurant inspections based on Yelp reviews. This approach extracts features from existing Yelp reviews and uses various machine learning algorithms such as SVM, RNN, Naive Bayes, and Random Forest to identify restaurants that require immediate inspections [27].…”