Application of AI Algorithms Integrating High-Frequency Trading Data in Financial Market Liquidity Risk Prediction

Abstract

With the rapid development of financial technology and the increasing fragmentation of the financial market, high-frequency trading (HFT) has become a core driving force affecting market liquidity, and liquidity risk has gradually become one of the key risks endangering the stability of the financial system. Traditional liquidity risk prediction methods rely on low-frequency financial data, which are difficult to capture the dynamic and nonlinear characteristics of high-frequency trading markets, leading to insufficient prediction accuracy and timeliness. In this study, we integrate high-frequency trading data (including order book data, transaction tick data, and order flow data) and construct a liquidity risk prediction framework based on artificial intelligence (AI) algorithms, including Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Transformer. First, we define and calculate key liquidity indicators based on high-frequency data to characterize market liquidity from multiple dimensions such as tightness, depth, and resiliency. Second, we preprocess the high-frequency data through normalization, outlier detection, and sequence construction to solve the problems of high noise and high dimensionality of HFT data. Then, we establish AI prediction models to predict the liquidity risk level (measured by Liquidity at Risk, LAR) and verify the effectiveness of the models using real high-frequency trading data of the U.S. stock market and cryptocurrency market. The experimental results show that the AI algorithms integrating high-frequency trading data significantly outperform traditional models (such as Logistic Regression, LR, and Support Vector Machine, SVM) in terms of prediction accuracy, recall rate, and F1-score. Among them, the Transformer model, which can capture long-term dependencies in time series, achieves the best prediction effect, with an average accuracy of 89.73%. This study provides a new technical path for financial institutions to conduct real-time liquidity risk management and has important theoretical and practical significance for maintaining the stability of the financial market.