Detection of Black Regions in the Forex Market By Analyzing High-Frequency Intraday Data

摘要

The Foreign Exchange (Forex) market is one of the most exciting, and fast-evolving financial markets in the world. The evolution of the prices of the Forex market depends on a variety of factors including unexpected news, rumors, scheduled news announcements, and micro-blogging posts. Hence accurate prediction of prices, as well as high-quality backtesting, is more challenging as it is greatly controlled by these anomalous time periods (also called black regions) made by user generated content in social media and news websites. Most of the existing literature related to anomaly detection in financial time series are supervised learning approaches. They rely on a training dataset and have lesser generalizing power. This research proposes an unsupervised pipeline based on ARIMA, GARCH and LOF techniques to detect black regions in the Forex Market. The proposed method demonstrates high accuracies compared to other benchmark methods.