Determine Intraday Trading Currency's Trend Framework Evidence From Machine Learning Techniques

Abstract

The purpose of this study is to determine the intraday hourly trading trends of currencies using predictive modeling techniques. The study encompasses two distinct intraday time intervals of 30 minutes and 1 hour, analyzing currencies from 8 different countries. It incorporates the use of wavelets MODWT to identify trends and noise in intraday currency analysis. Three predictive models, namely Support Vector Regression, Recurrent Neural Network, and Long Short-Term Memory, are applied to relative time series data to predict intraday trading currency trends. The study reveals significant noise presence in three currencies based on MODWT analysis. Additionally, it demonstrates that deep learning techniques, such as LSTM, outperform traditional machine learning approaches in accurately predicting intraday currency trends. This study contributes substantially to the theoretical understanding of international finance and provides practical insights for real-time problem-solving in currency markets. Further, this research adds to the discourse on leveraging sophisticated analytical methods within the domain of business intelligence to enhance decision-making processes in organizations operating within dynamic and complex financial environments.