Fourier Genetic Series: An Evolutionary Time Series Modeling Technique

Abstract

There are several digital trading platforms provided by stock exchanges and financial brokers, through which one can program buy and sell orders without human intervention. Thus, the present work conducts a study of the state of the art of forecasting techniques in time series found in the most recent research, such as Artificial Neural Networks and Deep Learning, as well as the classical statistical regression techniques, where they highlight Linear Regression, Exponential Smoothing, and the ARIMA (Autoregressive Integrated Moving Average) technique. By analyzing the related works, the present work proposes an automatic method that mixes strengths of existing techniques, shaping time series with different characteristics, abstracting the attributes of seasonality, trend variance and correlation in a concise representation through of a finite sum of strongly inspired mathematical expressions in the Fourier Series.