34th Annual Student Research Conference
An Exploration of Digital Advertising Optimization Using Thompson Sampling
Austin D. Davis
Dr. Ruthie Halma, Faculty Mentor
As consumerism continues to be a driving force of economic growth across the world, companies are increasingly looking for methods to optimize their advertisements and reach a broader market of potential consumers. Click-through rate (CTR) is an essential measure of how a digital advertisement performs over the Internet, and is measured by recording the number of clicks on an advertisement per a given number of total impressions. Thompson Sampling, a Bayesian optimization method, can dynamically analyze a finite number of inputs to determine which input yields the most optimal results. This research investigates the implementation of a Thompson Sampling algorithm analyzing a simulated advertising environment. To simulate this artificial environment, Python notebooks are used to implement a Thompson Sampling algorithm that iterates over a large advertising dataset. This enables us to determine which advertisements yield the highest CTR.
Keywords: Advertising, Thompson Sampling, Bayesian Optimization, Machine Learning
Topic(s):Computer Science
Presentation Type: Asynchronous Virtual Oral Presentation
Session: 7-1
Location: https://flipgrid.com/3853f3aa
Time: 0:00