SVM and fuzzy logic based forex prediction and trading

• 1、School of Computer Science,Beijing University of Posts and Telecommunications, Beijing 100876
• 2、School of Computer Science,Beijing University of Posts and Telecommunications,Beijing 100876

Abstract：In this paper the application of SVM (support vector machine) and fuzzy logic inference systems in the field of exchange rate forecast is studied, and the simulation of trading gains and losses is carried out according to the output of optimized model. We conduct staged training and testing by the use of sliding window. At each stage we first use genetic algorithms dynamically to select classification features, then we train and test radial basis, linear, polynomial, Sigmoid Support Vector Machines and Mamdani fuzzy logic inference systems of two different precision, thus in total of six models. We select one model with the best performance perspectively from the two machine learning techniques to conduct joint prediction, and simulate trade under the output of joint prediction. The experimental results show that our joint prediction mode gives a more satisfactory accuracy, while the corresponding trading strategy has a very substantial income within a particular market trends in the test period.

Keywords： Machine learning SVM Fuzzy logic Exchange rate forecast Forex transaction