Price Jump Prediction in Limit Order Book
Price Jump Prediction in Limit Order Book
This paper investigates price jump prediction in limit order books, focusing on the relationship between liquidity balance and market order dynamics. The authors define price jumps as market orders executed at prices worse than the prevailing best bid or ask prices, creating a framework for analyzing execution quality deterioration. They extract comprehensive features from limit order book data, including volumes, price gaps, and order event information, to build predictive models.
The methodology employs logistic regression to predict price jumps, enhanced with LASSO regularization for feature selection and importance assessment. To address intraday seasonality patterns, the analysis is conducted separately on morning and afternoon datasets. The empirical study covers forty major French stocks from the CAC40 index, revealing that trade sign, market order size, and liquidity balance at the best bid/ask levels consistently provide significant predictive power for anticipating price jumps, offering practical insights for market microstructure analysis and algorithmic trading strategies.
Highlights
- 1Empirical demonstration of the relationship between bid-ask liquidity balance and trade sign for predicting market order direction
- 2Definition of price jump based on market order execution relative to best bid/ask prices
- 3Application of LASSO logistic regression for feature selection in price jump prediction
- 4Separation of datasets into morning and afternoon to account for intraday seasonality
- 5Identification of consistently informative features across forty CAC40 stocks
Methods
- MLogistic regression for price jump prediction
- MLASSO logistic regression for variable selection and feature importance analysis
- MFeature extraction from limit order book data (volumes, price gaps, market/limit order information)
- MDataset segmentation into morning and afternoon periods to address intraday seasonality
Results
- RLiquidity balance on best bid/best ask is highly informative for predicting future market order direction
- RTrade sign and market order size are consistently significant predictors of incoming price jumps
- RLASSO regression identifies key features from limit order book data that contribute to price jump prediction
- RAnalysis on forty CAC40 stocks shows robust predictive patterns across different trading periods
Analyze Paper
Generate insights from "Price Jump Prediction in Limit Order Book".