Title: Evolutionary relaxed support vector regression for exchange rates trading
Authors: Shaolong Sun - Xi'an Jiaotong University (China) [presenting]
Shouyang Wang - Academy of Mathematics and System Science, Chinese Academy of Sciences, (China)
Yunjie Wei - Academy of Mathematics and Systems Science, Chinese Academy of Sciences (China)
Abstract: A new evolutionary learning approach, namely CS-RSVR, is proposed for exchange rate forecasting and trading. The proposed CS-RSVR approach can dynamically optimize the values of all SVR's parameters through the CS evolutionary algorithm, and use acquired parameters to construct optimized RSVR in order for proceeded forecasting foreign exchange rates. Many researchers have discussed exchange rate forecasting with the majority focusing on forecasting performance, however; accuracy is only one part of exchange rate forecasting. More important is how integrated approaches such as this can guide professional practice. We extend our forecasting to test trading performance of exchange rates between the USD and four other major currencies, EUR, GBP, CNY and JPY. The experimental results demonstrate the CS-based optimized SVRs models significantly improve efficient in trading terms compared with other optimized SVRs models. Generally speaking, our proposed CS-RSVRb model can be considered as a promising solution for exchange rates forecasting and trading.