Prediction of Financial Time-Series Signals Using a Trous Wavelet Transform

摘要

This paper proposes a financial time-series prediction method consisting of a Trous wavelet transform and polynomial regression. The main purpose of employing a Trous wavelet transform is to decompose financial time-series signals into different resolutions where only relevant signal components are used for prediction. Also, a Trous wavelet transform is used to avoid the edge problem where only the past and present components of the time-series signal are taken into account. The decomposed time-series signals are then fed into the polynomial regression part to predict time-series signals. Using real-world financial data, performance evaluation is conducted based on total benefit and profit/loss where it is shown that a Trous wavelet transform contributes to a significant performance improvement.