Paper: arXiv 2311.10719

Authors: Jiahao Chen, Xiaofei Li

Abstract

In recent years, high-frequency trading has emerged as a crucial strategy in stock trading. This study aims to develop an advanced high-frequency trading algorithm and compare the performance of three different mathematical models: the combination of the cross-entropy loss function and the quasi-Newton algorithm, the FCNN model, and the vector machine. The proposed algorithm employs neural network predictions to generate trading signals and execute buy and sell operations based on specific conditions. By harnessing the power of neural networks, the algorithm enhances the accuracy and reliability of the trading strategy. To assess the effectiveness of the algorithm, the study evaluates the performance of the three mathematical models. The combination of the cross-entropy loss function and the quasi-Newton algorithm is a widely utilized logistic regression approach. The FCNN model, on the other hand, is a deep learning algorithm that can extract and classify features from stock data. Meanwhile, the vector machine is a supervised learning algorithm recognized for achieving improved classification results by mapping data into high-dimensional spaces. By comparing the performance of these three models, the study aims to determine the most effective approach for high-frequency trading. This research makes a valuable contribution by introducing a novel methodology for high-frequency trading, thereby providing investors with a more accurate and reliable stock trading strategy.

Complexity vs Empirical Score

  • Math Complexity: 4.0/10
  • Empirical Rigor: 6.0/10
  • Quadrant: Street Traders — empirical and applied, lower math complexity

Why this score: The paper applies standard machine learning models (logistic regression, FCNN, SVM) with basic mathematical notation but lacks advanced derivations or novel theory. It demonstrates empirical rigor by using real stock data, detailing preprocessing steps, and reporting performance, aligning with practical trading research.

Research Flowchart

  flowchart TD
  A["Research Goal:<br>Analyze ML Models for HFT"] --> B{"Data Processing:<br>Stock Market Data"}
  B --> C1["Model 1:<br>Cross-Entropy + Quasi-Newton"]
  B --> C2["Model 2:<br>FCNN (Deep Learning)"]
  B --> C3["Model 3:<br>Vector Machine"]
  
  C1 --> D{"Performance Comparison"}
  C2 --> D
  C3 --> D
  
  D --> E["Findings:<br>Optimal Strategy Identified"]

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 an advanced high‑frequency trading algorithm and compares three mathematical models for stock prediction.

Source sentence
This study aims to develop an advanced high-frequency trading algorithm and compare the performance of three different mathematical models

Assumes 3

  • A fixed trading fee of $10 per transaction is assumed for profit calculations. execution
    Source sentence
    Assuming a fixed trading fee of $10 per transaction, we can calculate the net profit or loss for each model’s predictions.
  • The empirical analysis is based on a small dataset of ten stocks. data
    Source sentence
    It is important to note that these results are based on a small dataset of ten stocks
  • A linear equation can approximate stock price movements using the quasi‑Newton optimizer’s BFGS method. model
    Source sentence
    an approximately accurate linear equation can be obtained to predict the stock price movements at that time

Data

  • Ten stocks (Fengdong Co., Ltd., Transfer of intellectual education, An Naier, ...) market-prices unknown
    ten different companies
    Source sentence
    The table2,3,4 presented in the table shows the results obtained from applying three different models, logic regression, SVM, and FCNN, to predict the rise and fall signals of ten stocks.

Costs

Partial Costs appear only as a robustness check, a flat fee without spread or impact, or outside the headline numbers.

A flat $10 fee per transaction is used in net‑profit calculations but not incorporated into headline performance numbers.

Source sentence
Assuming a fixed trading fee of $10 per transaction, we can calculate the net profit or loss for each model’s predictions.

Finds 4

  • The FCNN model outperforms the other models in terms of risk and return, delivering higher ROI and lower loss risk. performance
    Source sentence
    Analysis shows that the FCNN model outperforms other models in terms of risk and return.
  • Logic regression has the highest average train accuracy (87.5%) while FCNN has the highest average test accuracy (62.1%). predictive-accuracy
    Source sentence
    The test accuracy results show a different trend, with the FCNN model having the highest average test accuracy of 62.1%
  • Average total revenue is low; the highest observed is 1.9% for Zhejiang Guangxia predicted by the SVM model. performance
    Source sentence
    The highest value being 1.9% for Zhejiang Guangxia, which was predicted by the SVM model.
  • Logic regression has the lowest average training time (7.9 s), SVM 7.4 s, and FCNN the longest (6.9 s). performance
    Source sentence
    From the table, we can see that the average total time taken by the logic regression model is 7.9 seconds, which is the lowest among all the models. The SVM model takes an average of 7.4 seconds, while the FCNN model takes the longest time, with an average of 6.9 seconds.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 47 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.