Class Project: Inferring Customer Characteristics and Predicting Satisfaction from Reviews

Python Code

Project Paper

I used the Yelp data to provide a new measure of customer satisfaction level by reviews and rating stars. Using the restaurants in Las Vegas as an example, this project uses 8 machine learning models with scikit-learn to predict the satisfaction of all customers. The model only uses features that are available for all customers, so that we can predict even when they don’t leave the reviews. The key insight is that the customer heterogeneity variables (gender, date of visits, with group or not, etc), extracted from data, are not predictive for the satisfaction level.