AdaBoost Classification for Short-Term Market Timing

Summary

This report frames next-day direction in China’s Wind All A-share index as a binary classification task. It trains a decision-tree-based AdaBoost model on 51 daily features drawn from several markets and asset classes, including repo rates, credit spreads, commodity returns, the gold-to-oil ratio, and US equity data. The report presents the model as a way to select useful signals, represent nonlinear effects, and adapt to relationships among features.

A rolling backtest reports long-short and long-only results over a historical sample, first without transaction costs. It also describes a combined model that adds an options-based timing signal and reports results under an assumed one-sided trading cost. These are historical backtest findings, not evidence of future performance; the supplied text does not include detailed model settings, validation procedures, or enough information to assess robustness independently.

Key ideas

• The report treats next-day index direction as a binary prediction problem.
• AdaBoost decision trees use 51 daily features from multiple markets and asset classes.
• The method is presented as selecting relevant features and modeling nonlinear relationships.
• Reported rolling backtests include both a standalone model and a version combined with an options-based timing signal.
• The reported performance is historical and depends on stated transaction cost assumptions.

Tags

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