Publication Date

Spring 2019

Degree Type

Master's Project

Degree Name

Master of Science (MS)


Computer Science

First Advisor

Mike Wu

Second Advisor

Robert Chun

Third Advisor

Hailin Wu


Click-through rrate prediction, online advertising


Internet has become the most prominent and accessible way to spread the news about an event or to pitch, advertise and sell a product, globally. The success of any advertisement campaign lies in reaching the right class of target audience and eventually convert them as potential customers in the future. Search engines like the Google, Yahoo, Bing are a few of the most used ones by the businesses to market their product. Apart from this, certain websites like the that has more traffic also offer services for B2B customers to set their advertisement campaign. The look of the advertisement, the maximum bill per day, the age and gender of the audience, the bid price for the position and the size of the advertisement are some of the key factors that are available for the businesses to tune. The businesses are predominantly charged based the number of clicks that they received for their advertisement while some websites also bill them with a fixed charge per billing cycle. This creates a necessity for the advertising platforms to analyze and study these influential factors to achieve the maximum possible gain through the advertisements. Additionally, it is equally important for the businesses to customize these factors rightly to achieve the maximum clicks. This research presents a click through rate prediction system that analyzes several of the factors mentioned above to predict if an advertisement will receive a click or not with improvements over the existing systems in terms of the sampling the data, the features used, and the methodologies handled to improve the accuracy. We used the ensemble model with weighted scheme and achieved an accuracy of 0.91 on a unit scale and predicted the probability for an advertisement to receive a click form the user.