Skip to content
All library documents

LSTM Forecasting of Next-Day Equity Trend Changes

Article arXiv papers · Author: Harris Buchanan et al.
Read the original

Summary

E-TRENDS proposes an LSTM forecasting framework for next-day changes in equity trends. It focuses on the 30 largest S&P 500 stocks and describes validation across market cycles from 2005 to 2025. Instead of forecasting the trend level directly, the method predicts its daily difference, which the authors argue can reduce the bias-variance tradeoff.

The paper reports comparisons with ordinary least squares, Ridge, Lasso, and LightGBM regression, and portfolio simulations that assess profit and loss against the alternative models. The abstract claims the simulations show economic gains, but it does not provide performance figures, portfolio construction details, transaction costs, or risk measures. The stated evaluation period and asset selection also limit how far the results can be generalized. The description offers a forecasting approach and benchmark plan, but further methodological and empirical details are needed to assess robustness or live-trading value.

Key ideas

  • The framework uses an LSTM to forecast next-day changes in equity trends.
  • The study focuses on 30 large S&P 500 equities and evaluates data spanning 2005 to 2025.
  • Forecasting differences rather than levels is proposed to reduce the bias-variance tradeoff.
  • The reported benchmarks include OLS, Ridge, Lasso, and LightGBM regression.
  • Portfolio simulations are reported to show gains, though the abstract gives no detailed performance or cost figures.

Tags

EquitiesUS marketsTrend followingMachine learningBacktesting
Full text
# E-TRENDS: Enhanced LSTM Trend Forecasting for Equities


# E-TRENDS: Enhanced LSTM Trend Forecasting for Equities









Trend-following strategies underpin many systematic trading approaches yet struggle under nonstationary and nonlinear market regimes. We propose an LSTM-based framework to forecast next-day trend differences ($Δ_t$) for the top 30 S\&P 500 equities, validated across market cycles (2005--2025). Key contributions include: (i) formal proof of bias-variance reduction via differencing, (ii) exhaustive empirical benchmarks against OLS, Ridge, and Lasso, (iii) portfolio simulations confirming economic gains in terms of overall PNL compared to other models like OLS, Ridge, Lasso or LightGBM Regressor

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.