Skip to content
All library documents

LSTM Momentum Signals for Stock Portfolio Construction and Trading

Article arXiv papers · Author: Hsiang-Hui Liu et al.
Read the original

Summary

NoxTrader uses historical US stock price and volume data to build features such as return, weekly price, and monthly price momentum. It applies an LSTM model to predict returns and capture persistent price trends, updating the model dynamically during execution. Predictive scores guide portfolio management within a custom backtesting system, and the document says the predictions show a correlation range of 0.65 to 0.75.

The authors report that filtering the prediction data changed an initial investment return of -60% to 325%. They also compare prediction dispersion with observed market data. These are reported results, but the excerpt does not specify the test period, universe, costs, benchmark, risk exposure, or filtering procedure. The performance claims therefore cannot be assessed independently from the information provided.

Key ideas

  • The system engineers return, weekly price, and monthly price momentum features from US stock data.
  • An LSTM model predicts returns and is updated during trading to adapt to market trends.
  • Predictive scores are used to manage portfolios in a custom backtesting framework.
  • The document reports prediction correlations and a large improvement after filtering, but omits details needed to judge robustness.

Tags

EquitiesUS marketsMomentumMachine learningBacktestingPortfolio construction
Full text
# NoxTrader: LSTM-Based Stock Return Momentum Prediction for Quantitative Trading


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









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%.

Shown in full with attribution under the source's licence. Licence: abstract CC0

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.