A Time Series Forecasting Model Based on Deep Learning Integrated Algorithm with Stacked Autoencoders and SVR for FX Prediction

摘要

This paper proposes a Deep Learning integrated algorithm with Stacked Autoencoders (SAE) and Support Vector Regression (SVR), it is also for the first time that applies the SAE-SVR integrated algorithm to Foreign Exchange (FX) rate forecasting. We adopt 28 currency pairs pertaining to G7 currencies and RenMinBi, and collect the real daily FX data for simulation. To implement the empirical study, we develop the program of SAE-SVR integrated algorithm independently, and benchmark the results with ANN and SVR models, which are considered as the best performance in Artificial Intelligence. Ultimately, the simulation results indicate that the SAE-SVR integrated algorithm performs much better over other benchmarks.