Comparing Factors and Models in a Multi-Model Stock Strategy

Summary

This assignment describes building a stock strategy from four factors, including small market capitalization and turnover. The author used a provided template and AI assistance to implement a linear regression strategy, then packaged three models as options in a multi-factor strategy and displayed the outputs in charts. The account focuses on the workflow rather than specifying all factor definitions or portfolio rules.

The author reports that results varied across the models and testing periods: random forest performed best in their trials, while linear regression performed worse than XGBoost and random forest. They also observed that their chosen factors mattered more to profitability than the model choice. These are personal backtest observations, not evidence of robust out-of-sample performance; the document gives no dates, metrics, transaction costs, or validation method. It suggests that factor selection and model comparisons need more systematic analysis, including runtime comparisons and replacing provisional factors with candidates from factor analysis.

Key ideas

• The assignment combines four stock selection factors with machine learning models.
• The author compares linear regression, XGBoost, and random forest under the same factor setup.
• In the author’s trials, model results differed, and random forest performed best.
• The author considers factor quality more influential than model choice, based on limited testing.
• The reported outcomes lack performance metrics and validation details, so they do not establish general effectiveness.

Tags

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