Learning Non-Stationary Time-Series with Dynamic Pattern Extractions

TL;DR

This paper introduces a novel approach combining seq2seq RNN models with dynamic feature extraction techniques to improve forecasting accuracy of non-stationary time-series data, such as Forex trends.

Contribution

It proposes a new model that integrates attention mechanisms and dynamic pattern features for better non-stationary time-series prediction.

Findings

High accuracy in 4-hour Forex trend prediction

Effective pattern recognition from historical data

Enhanced model performance with customized loss functions

Abstract

The era of information explosion had prompted the accumulation of a tremendous amount of time-series data, including stationary and non-stationary time-series data. State-of-the-art algorithms have achieved a decent performance in dealing with stationary temporal data. However, traditional algorithms that tackle stationary time-series do not apply to non-stationary series like Forex trading. This paper investigates applicable models that can improve the accuracy of forecasting future trends of non-stationary time-series sequences. In particular, we focus on identifying potential models and investigate the effects of recognizing patterns from historical data. We propose a combination of \rebuttal{the} seq2seq model based on RNN, along with an attention mechanism and an enriched set features extracted via dynamic time warping and zigzag peak valley indicators. Customized loss functions and evaluating metrics have been designed to focus more on the predicting sequence's peaks and valley points. Our results show that our model can predict 4-hour future trends with high accuracy in the Forex dataset, which is crucial in realistic scenarios to assist foreign exchange trading decision making. We further provide evaluations of the effects of various loss functions, evaluation metrics, model variants, and components on model performance.