This repo contains work on my project, Michelin Survivor, that I created as part of my data science fellowship at Insight Data Science during the summer of 2019. The Jupyter notebook expands on details pertaining to my approach, and the two web-app directories contain code for my online Dashboard web applications, available at yasha.sheynin.com:8050.
Michelin-rated restaurants are some of the highest-grossing businesses in the fine-dining industry – a $10 billion market in the U.S. Inversely, losing a Michelin star can make a business less profitable, with up to 40% of these businesses closing within 5 years. In this notebook which I developed at Insight Data Science, I used Yelp reviews to develop a product that forecasts the risk of a restaurant losing a Michelin star and provides actionable insights to restauranteurs and investors on how to improve if their business is deemed at risk. Restaurants have vastly different timelines - from opening, getting Michelin-rated, to closing shop - so wrangling 200k+ Yelp reviews into usable features was a significant challenge to overcome, and vital in broader risk assessment and survival analysis problems.