Forecasting the trend of foreign exchange rates using time series analysis techniques

Abstract

Forecasting exchange rates from the foreign exchange market is a challenging research. The exchange rate forecast will be very beneficial to investors. Currently, time series techniques play a role in forecasting future data. The purpose of this research is to study and compare the efficiency of the models for predicting foreign exchange rates.In this paper, Using the upward trend of exchange rates.  From. February 2014 to January 2017 were used. four techniques including  Linear Regression (LR), Multi-Layer 9 Perceptron (MLP), Support Vector Machine  Regression (SVMR) and Sequential Minimal Optimization Regression (SMOR)  were employed. Sliding Windows was used to divide data into learning and testing sets. In this paper, 12 rounds of sliding windows were used to reduce the variance of experiment results. Moreover, Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) were used to measure performance of the models. The experiment result showed that SVMR technique is superior to LR, MLP and SMOR. The MAE and RMSE values of SVMR are the lowest which are were as low as 1.11 ± 2.10 and 1.13 ± 2.14, respectively.