Opera Solutions Uses Advanced Analytics to Better Predict Click-Through Rates for Search Advertising

Posted by Todd Higginson on Mon, Sep 10, 2012

NEW YORK, N.Y., Sept. 10, 2012 — Opera Solutions, LLC, a leading global Big Data and advanced predictive analytics company, today announced its results in the KDD Cup 2012, a leading worldwide competition that attracts the best and brightest minds in machine learning and advanced analytics. Competing against 170 teams, Team Opera Solutions won second place in the Track 2 contest, edged out by 6.9 thousandths of a point by a team from academia.

The challenge for the Track 2 contest was to accurately and efficiently predict click-through rates (CTRs) for Internet ads. A vital component of revenue streams for millions of companies, CTRs determine the effectiveness of an online ad campaign.

The teams were given 10GB of data, consisting of 150 million records. Using its state-of-the-art hybrid recommender system that integrated collaborative filtering models and feature-engineered models, Opera Solutions won second place with a predictive accuracy score of 0.80824. In a photo-finish ending, the first-place winner, Catch up @National Taiwan University, scored 0.80873, only a 6.9-thousandths improvement. Opera Solutions salutes Catch Up for its achievements in this challenging competition.

"In 2011, Internet search advertisement revenue reached $14.8 billion, up almost 27% from 2010, and accounted for 47.7% of all Internet advertisement revenue. Being able to accurately predict CTR is critical in better pricing ads and increasing profits. Our performance in the KDD Cup 2012 demonstrates our capabilities for doing so," said Jacob Spoelstra, Global Research and Development Head at Opera Solutions.

"This competition once again showcases the quality of the advanced algorithms and predictive models that power our Signal Hub Suite and Signal Apps."

The scientific methodologies used by Team Opera Solutions apply to more than predicting CTR. Opera Solutions already uses similar models to predict customer behavior for a variety of global enterprises, including a food distribution company, an international airline, and a major credit card corporation.

In addition to its participation and thought leadership, Opera Solutions donated its $2,000 second-place award to sponsor student-travel awards for this KDD2012 conference.

Topics: Signal Hub Technologies, Machine Learning, Competitions, Press Releases