Paper: arXiv 2411.08382

Authors: Abdul Rahman, Neelesh Upadhye

Abstract

In high frequency trading, accurate prediction of Order Flow Imbalance (OFI) is crucial for understanding market dynamics and maintaining liquidity. This paper introduces a hybrid predictive model that combines Vector Auto Regression (VAR) with a simple feedforward neural network (FNN) to forecast OFI and assess trading intensity. The VAR component captures linear dependencies, while residuals are fed into the FNN to model non-linear patterns, enabling a comprehensive approach to OFI prediction. Additionally, the model calculates the intensity on the Buy or Sell side, providing insights into which side holds greater trading pressure. These insights facilitate the development of trading strategies by identifying periods of high buy or sell intensity. Using both synthetic and real trading data from Binance, we demonstrate that the hybrid model offers significant improvements in predictive accuracy and enhances strategic decision-making based on OFI dynamics. Furthermore, we compare the hybrid models performance with standalone FNN and VAR models, showing that the hybrid approach achieves superior forecasting accuracy across both synthetic and real datasets, making it the most effective model for OFI prediction in high frequency trading.

Complexity vs Empirical Score

  • Math Complexity: 6.0/10
  • Empirical Rigor: 7.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced econometric modeling (VAR) and neural networks, indicating high mathematical density, and demonstrates rigor through real-world trading data from Binance, comprehensive model comparisons, and specific metrics for strategy evaluation.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Predict Order Flow Imbalance OFI in HFT"] --> B["Data Collection & Preprocessing"]
  B --> C{"Model Formulation"}
  
  C --> D["VAR Component<br>Captures Linear Dependencies"]
  C --> E["Residual Analysis<br>Extracts Non Linear Patterns"]
  
  D --> F["FNN Component<br>Models Non Linear Residuals"]
  E --> F
  
  F --> G["Hybrid VAR FNN Model<br>Combines Linear & Non Linear Predictions"]
  
  G --> H["Key Findings & Outcomes"]
  
  H --> I["Superior Accuracy<br>Hybrid outperforms standalone VAR and FNN"]
  H --> J["Trading Intensity Insights<br>Identifies Buy vs Sell Pressure"]
  H --> K["Strategic Utility<br>Facilitates HFT Strategy Development"]
  
  style A fill:#e1f5fe
  style H fill:#f3e5f5

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a hybrid predictive model that combines Vector Auto Regression (VAR) with a simple feedforward neural network (FNN) to forecast Order Flow Imbalance (OFI) and assess trading intensity.

Source sentence
this paper introduces a hybrid predictive model that combines Vector Auto Regression (VAR) with a simple feedforward neural network (FNN) to forecast OFI and assess trading intensity.

Assumes 6

  • The VAR component captures linear dependencies while the FNN models non‑linear residuals. model
    Source sentence
    The VAR component captures linear dependencies, while residuals are fed into the FNN to model non-linear patterns
  • Lag order p is selected using information criteria such as AIC or BIC. model
    Source sentence
    past observations are considered, and is often determined through criteria such as the Akaike Information Criterion (AIC) or Bayesian Information Criterion (BIC).
  • The dataset is used without any normalization or resampling. data
    Source sentence
    gathered for a single cryptocurrency without applying any normalization or resampling.
  • Synthetic data are generated to replicate characteristics of the real datasets. data
    Source sentence
    A synthetically generated dataset of 3,000 data points designed to replicate characteristics similar to the real datasets.
Show 2 more
  • Training uses early stopping to ensure model stability. model
    Source sentence
    Epochs: 50, with early stopping to ensure model stability.
  • Trading intensity can be measured as a separate output from the hybrid model. market
    Source sentence
    Output Layer for Hybrid Model: Two neurons for predicting buy and sell orders individually, from which OFI is calculated.

Data

  • Real Dataset 1 (BTCUSD) · Binance crypto-onchain public
    Source sentence
    Real Dataset 1 (BTCUSD): Contains approximately 3,000 data points representing BTCUSD transactions.
  • Real Dataset 2 (ETCUSDT) · Binance crypto-onchain public
    Source sentence
    Real Dataset 2 (ETCUSDT): Contains approximately 3,000 data points for ETCUSDT transactions.
  • Synthetic Dataset synthetic unknown
    Source sentence
    Synthetic Dataset: A synthetically generated dataset of 3,000 data points designed to replicate characteristics similar to the real datasets.

Recognised sources: Binance

Costs

Not applicable The paper reports no trading or execution result.

Finds 3

  • The hybrid VAR‑FNN model achieved lower MSE (0.00127) and higher R² (0.9946) than the FNN‑only model (MSE 0.00133, R² 0.9843) on training data. performance
    Source sentence
    Hybrid VAR-FNN 0.00127 0.00541 0.9946
  • On validation, the hybrid model outperformed VAR‑only and FNN‑only models with MSE 0.002, MAE 0.019, R² 0.997 and intensity accuracy 98.18% for BTCUSD. predictive-accuracy
    Source sentence
    Hybrid VAR-FNN 0.002 0.019 0.997 98.18% 96.32%
  • The hybrid model consistently achieved higher trading‑signal accuracy (≈96‑98%) compared to VAR‑only and FNN‑only models across real and synthetic datasets. performance
    Source sentence
    The hybrid model achieved high accuracy in predicting trading signals for both real and synthetic datasets, showing significant improvements over the VAR-only and FNN-only models.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 10 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (40)

  1. A vector-autoregression analysis of credit and liquidity factor dynamics in US LIBOR and Euribor swap markets (2012) — link
  2. Asymmetric heavy-tailed vector auto-regressive processes with application to financial data (2020) — link
  3. Asymmetric post earnings announcement drift and order flow imbalance: The impact on stock market returns (2024) — link
  4. Bond futures and order imbalance (2013) — link
  5. Capturing the Order Imbalance with Hidden Markov Model: A Case of SET50 and KOSPI50 (2020) — link
  6. Capturing the Order Imbalance with Hidden Markov Model: A Case of SET50 and KOSPI50 (2020) — link
  7. Cross-impact of order flow imbalance in equity markets (2023) — link
  8. Deep order flow imbalance: Extracting alpha at multiple horizons from the limit order book (2023) — link
and 32 more
  1. Dynamic relations between order imbalance, volatility and return of top gainers (2012) — link
  2. Effect of order flow imbalance on market impact across market states (2016) — link
  3. Enhancing trading strategies with order book signals (2018) — link
  4. Estimation of slowly decreasing Hawkes kernels: application to high-frequency order book dynamics (2016) — link
  5. Flow Toxicity and Liquidity in a High-frequency World (2012) — link
  6. Forecasting High Frequency Order Flow Imbalance — → 2408.03594
  7. Hawkes model for price and trades high-frequency dynamics (2014) — link
  8. High Frequency Trading (2012) — link
  9. High Frequency Trading (2014) — link
  10. Limit order books (2013) — link
  11. Modeling high-frequency order flow imbalance by functional limit theorems for two-sided risk processes (2015) — link
  12. Multi-Level Order-Flow Imbalance in a Limit Order Book (2019) — link
  13. Neural networks in business: a survey of applications (1992–1998) (1999) — link
  14. NEURAL NETWORKS IN FINANCE AND ECONOMICS FORECASTING (2007) — link
  15. Optimal Execution with Dynamic Order Flow Imbalance (2015) — link
  16. Order imbalance and liquidity supply: Evidence from the bubble burst of Nasdaq stocks (2005) — link
  17. Order imbalance and the dynamics of index and futures prices (2007) — link
  18. Order Imbalance, Liquidity, and Market Efficiency: Evidence from the Chinese Stock Market (2011) — link
  19. Price Impact of Order Flow Imbalance: Multi-level, Cross-sectional and Forecasting (2021) — link
  20. Price Impact of Order Flow Imbalance: Multi-level, Cross-sectional and Forecasting (2021) — link
  21. Reducing Misclassification Costs (1994) — link
  22. Shallow Neural Hawkes: Non-parametric kernel estimation for Hawkes processes (2022) — link
  23. The Effects of Order Flow Imbalance on Stock Prices in Tehran Stock Exchange (2022) — link
  24. The impact of neural networks in finance (1997) — link
  25. The interaction between order imbalance and stock price (1997) — link
  26. The Price Impact of Order Book Events (2014) — link
  27. The Price Impact of Order Book Events (2014) — link
  28. Trade size, order imbalance, and the volatility–volume relation (2000) — link
  29. Trade size, order imbalance, and the volatility–volume relation (2000) — link
  30. Vector autoregression and causality: a theoretical overview and simulation study (1994) — link
  31. Vector Autoregressions (2001) — link
  32. Vector Autoregressive Models for Multivariate Time Series — link

1 of 40 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.