Predictive Analytics for Swap Price Forecasting
Enhancing Price Prediction with Agile Data Science Challenges
Predictive Analytics for Swap Price Forecasting
Enhancing Price Prediction with Agile Data Science Challenges
The Challenge
As part of compliance with the Dodd-Frank Act, registered swap dealers are required to report credit and interest rate trades to public repositories within minutes of execution. This centralized trade data increases transparency—and opens the door to predictive analytics.
Credit Suisse wanted to capitalize on this by building models to forecast the prices of US$/Libor vanilla spot start swaps over short time intervals. With many factors influencing swap pricing, even small gains in predictive accuracy could deliver a strong competitive edge.
The Solution
Topcoder launched a 12-month predictive analytics program using crowdsourced data science challenges. Trading data from a 6-month period in 2016 was used to train and test algorithms across multiple 12-day sprints. Each challenge enabled rapid iteration and agile exploration of the problem space.
The program ultimately produced four winning models with low error rates—averaging just 4.3%—as well as recommendations for feature combinations that could further refine Credit Suisse’s in-house predictive tools.
12
Month Program
4
Winning Models Delivered
4.3 %
Avg. Prediction Error
The Impact
Credit Suisse gained a suite of high-performing models for swap price forecasting, built without overloading internal teams.
The crowdsourced approach allowed them to tap into global talent, explore diverse technical strategies, and make measurable improvements in predictive capabilities—enhancing their ability to transact more effectively in competitive financial markets.
Achieve high-quality outcomes with
Topcoder.
Achieve high-quality outcomes with Topcoder.