A Deep Learning Approach to Anomaly Detection in High-Frequency Trading Data

2025-03-31

AI Summary

This paper addresses the challenge of detecting microstructural anomalies in high-frequency foreign exchange (EUR/USD) trading data. We propose a novel two-stage sliding-window Transformer model that jointly leverages self-attention and microstructure-aware weighted attention to simultaneously capture local abrupt changes and global temporal dependencies. The model explicitly encodes multi-scale microstructural features—including order book depth, bid-ask spread, and trading volume—and is end-to-end optimized for highly noisy, non-stationary financial time series. Key innovations include the first-ever two-stage sliding-window architecture and a microstructure-guided weighted attention mechanism. Evaluated on real-world high-frequency data, our model achieves 0.93 accuracy, 0.91 F1-score, and 0.95 AUC-ROC—substantially outperforming state-of-the-art tree-based and deep temporal models. Ablation studies and attention visualization confirm the efficacy of each component and the model’s robust representation capability for complex market dynamics.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• Web Mining and Content Analysis: Bridging structured and unstructured data
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

This paper proposes an algorithm based on a staged sliding window Transformer architecture to detect abnormal behaviors in the microstructure of the foreign exchange market, focusing on high-frequency EUR/USD trading data. The method captures multi-scale temporal features through a staged sliding window, extracts global and local dependencies by combining the self-attention mechanism and weighted attention mechanism of the Transformer, and uses a classifier to identify abnormal events. Experimental results on a real high-frequency dataset containing order book depth, spread, and trading volume show that the proposed method significantly outperforms traditional machine learning (such as decision trees and random forests) and deep learning methods (such as MLP, CNN, RNN, LSTM) in terms of accuracy (0.93), F1-Score (0.91), and AUC-ROC (0.95). Ablation experiments verify the contribution of each component, and the visualization of order book depth and anomaly detection further reveals the effectiveness of the model under complex market dynamics. Despite the false positive problem, the model still provides important support for market supervision. In the future, noise processing can be optimized and extended to other markets to improve generalization and real-time performance.

Problem

• Detects anomalies in high-frequency EUR/USD trading data
• Improves accuracy over traditional and deep learning methods
• Addresses false positives for better market supervision

Innovation

• Staged sliding window Transformer architecture
• Self-attention and weighted attention mechanisms
• Classifier for abnormal event identification