NoxTrader: LSTM-Based Stock Return Momentum Prediction for Quantitative Trading

Hsiang-Hui Liu,Han-Jay Shu,Wei-Ning Chiu

2023-10-31

Abstract:We introduce NoxTrader, a sophisticated system designed for portfolio construction and trading execution with the primary objective of achieving profitable outcomes in the stock market, specifically aiming to generate moderate to long-term profits. The underlying learning process of NoxTrader is rooted in the assimilation of valuable insights derived from historical trading data, particularly focusing on time-series analysis due to the nature of the dataset employed. In our approach, we utilize price and volume data of US stock market for feature engineering to generate effective features, including Return Momentum, Week Price Momentum, and Month Price Momentum. We choose the Long Short-Term Memory (LSTM)model to capture continuous price trends and implement dynamic model updates during the trading execution process, enabling the model to continuously adapt to the current market trends. Notably, we have developed a comprehensive trading backtesting system - NoxTrader, which allows us to manage portfolios based on predictive scores and utilize custom evaluation metrics to conduct a thorough assessment of our trading performance. Our rigorous feature engineering and careful selection of prediction targets enable us to generate prediction data with an impressive correlation range between 0.65 and 0.75. Finally, we monitor the dispersion of our prediction data and perform a comparative analysis against actual market data. Through the use of filtering techniques, we improved the initial -60% investment return to 325%.

Portfolio Management,Artificial Intelligence

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to utilize deep - learning techniques, especially Long Short - Term Memory (LSTM) networks, in quantitative trading to predict the momentum returns in the stock market, thereby achieving medium - and long - term profit goals. Specifically, the paper aims to:

1. **Improve the accuracy of market prediction**: By using the LSTM model to capture the time - series features in historical trading data, especially price and volume data, to generate effective features such as Return Momentum, Week Price Momentum and Month Price Momentum.

2. **Develop a comprehensive back - testing system**: Build the NoxTrader system for managing portfolios and evaluating trading strategies based on prediction scores. This includes monitoring the dispersion between predicted data and actual market data and optimizing the return on investment through filtering techniques.

3. **Explore new methods of label generation**: Introduce the concept of "Return Momentum" as a prediction label instead of the traditional raw return to improve the predictive ability of the model.

4. **Verify the application effect of the model in the actual market**: Not only focus on the prediction accuracy of the model, but also emphasize its practical application value in the real financial market, and demonstrate the effectiveness and profitability of the model through empirical research.

Through these efforts, the paper hopes to provide an efficient and practical solution for quantitative trading, especially in medium - and long - term investment strategies.