Parallel Monte Carlo Method With Mapreduce For Option Pricing
Shijia Wang,Liang Zhou,Yuanyuan Li,Junfeng Wu
DOI: https://doi.org/10.1109/trustcom.2016.0325
2016-01-01
Abstract:Option and other financial derivatives are becoming more and more important in financial market. As one of the most important financial activities, reasonable option pricing not only makes the trade market steady and orderly, but also provides investors valuable information to make decisions. As the growth of the financial data and the inherent complexity of option pricing methods, option pricing is facing more and more challenges, such as the random problem of solution and time consuming. Monte Carlo is one of the most common used methods in option pricing. However, in order to obtain a better solution, Monte Carlo method requires huge number of simulations. It may be up to tens of millions of simulations and generates large amounts of data. That is, Monte Carlo simulation is inherently computing-intensive. Meanwhile, the implementation of traditional parallel Monte Carlo method is complicated. Massive data and huge computational cost limit the further application of Monte Carlo method.In order to deal with the above problems, this paper researches the efficient B-S option pricing problem with Monte Carlo and proposes a parallel Monte Carlo method for option pricing. This method extends the Monte Carlo simulation to the MapReduce framework, which is a simple powerful parallel programming technique. It divides the Monte Carlo simulation into three phases and they are implemented with one MapReduce job. With the help of a large-scale cluster computing power and the excellent scalability of MapReduce, the proposed method scales well and solves the option pricing efficiently. The experimental results also demonstrates the good characteristic of speedup and sizeup.