Skip to content
All library documents

Genetic Algorithms for Similar-Stock Grouping and Portfolio Optimization

Article BigQuant
Read the original

Summary

This research summary describes a method for grouping stocks with similar price behavior so that an investor can substitute one holding for another within a group. It builds on a grouping genetic algorithm, representing groups, stocks, and portfolio components in candidate solutions. To compare long price histories, it applies Symbolic Aggregate approXimation (SAX) or its extended form, ESAX, and introduces a sequence-distance factor to measure similarity. The fitness functions also incorporate dividend-based stability and measures of unit and price balance.

Experiments on real data are reported, with SAX producing returns of roughly 16% to 18% and outperforming ESAX on that measure, while ESAX yields greater portfolio similarity. The summary therefore presents a trade-off between return and similarity. It does not provide the underlying dataset, test period, benchmark, transaction costs, or implementation details in the supplied text, limiting how confidently the reported results can be generalized or reproduced.

Key ideas

  • The method groups stocks with similar price sequences to support within-group substitutions.
  • A grouping genetic algorithm searches for portfolios that satisfy grouping and portfolio objectives.
  • SAX and ESAX compress price sequences, while a sequence-distance factor evaluates similarity.
  • Fitness measures also account for dividend-based stability and balance in units and prices.
  • The reported experiments favor SAX on return and ESAX on portfolio similarity.

Tags

EquitiesMachine learningPortfolio constructionStatistics

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