Big Data Forex Analysis using GPU Computing

The largest financial market in the world is the Forex (Foreign Currency Exchange) market, by virtue of the highest volume of trading that takes place on a daily basis. Being able to read underlying market patterns and making smart choices amidst the turbulent and organic Forex marketplace is the first step to decision making. Traders and investors harvest profitable returns from Forex market by buying and selling when the exchange rates are respectively low and high. Indicator analyses can be used to locate the ideal times to convert back and forth between currencies. These Indicator analyses themselves involve parameters, which are usually chosen manually from experience, which usually are not the optimal choices. The parameters can be optimized from historic data using software, but this is computationally intensive and time consuming. In this paper, we propose a method to speed up the optimization of indicator parameters, using CUDA parallel processing API of NVIDIA GPUs (Graphical Processing Units) as opposed to the classic CPU based sequential approach. While it seemed logical to incorporate several high-end processors (CPUs) in order to harness more computing power, we aim at demonstrating that a GPU based implementation, based on suitably written kernels and threads, has the potential to be scaled for industrial use.