Date of Award


Document Type

Honors Thesis

Degree Name

Bachelor of Science


Computer Science

Advisor/Committee Chair

Charalampos Chelmis


With the rise in popularity of the Internet, data describing unique types of items has been collected into easy to access sources. Using this newly acquired data, is it possible to predict if an item will become a bestseller while another fade away with time? Popularity prediction is a problem that has attracted a great deal of research recently, and for good reason. The ability to predict an items future rise to popularity or fall to obscurity is a possibly priceless skill and sought out in many different industries such as sales, investments, and marketing. This report enumerates and analyzes a number of factors assumed to be an indicator of popularity. Additionally, we propose a number of popularity prediction methods, and evaluate using a cohost of evaluation metrics, and state of the art baselines. Our findings show promising potential for popularity prediction, based on a combination of structural and temporal properties indicative of popularity. The key proposed metrics include a measure of similarity between two items, and various definitions of popularity evolving with time. Experiments on a large scale real dataset from Yelp allow us to demonstrate the performance of our methods on predicting the popularity of businesses. We believe the methods described below can be extended to be used for diverse types of data.