Fractional Fourier Transform in Time Series Prediction

Several signal processing tools are integrated into machine learning models for performance and computational cost improvements. Fourier transform (FT) its variants, which powerful spectral analysis, employed in the prediction of univariate time series by converting them to sequences domain be processed further recurrent neural networks (RNNs). This approach increases reduces training compared conventional methods. In this letter, we introduce fractional (FrFT) RNNs. As a parametric transformation, FrFT allows us seek select better-performing transformation domains providing access continuum between frequency. flexibility yields significant improvements power underlying without sacrificing efficiency. We evaluated our FrFT-based on synthetic real-world datasets. Our results show that gives rise over ordinary FT. 1 1 Source codes available at https://github.com/koc-lab/FrFTimeSeries