Skip to content

Memetic Portfolio Optimization

February 18, 2010

While researching multi-objective optimization for large search spaces and clustering, Quantivity stumbled upon some interesting recent work by Aranha and Iba applying memetic and biologic algorithms. Latter work appears along the line of Brabazon and O’Neill Biologically Inspired Algorithms for Financial Modelling (which, responding to previous comments, may be potentially useful for those looking for published examples of GA applied to trading, as it includes numerous simplistic case studies in Part III).

First, applying memetic algorithms for portfolio optimization and trading rule evolution. Memetic seems potentially interesting as it combines local search with genetic evolution, and thus may have some interesting uses where classic GA has proven unsuccessful:

Genetic algorithm improves the solution in large strokes, while the local optimization fine tunes the solutions generated by the GA.

Interesting papers, coming from GECCO 09 and ACSE 09:

  • Aranha and Iba, Using Memetic Algorithms to Improve Portfolio Performance in Static and Dynamic Trading Scenarios
  • Aranha and Iba, Memetic Tree-based Genetic Algorithm and its application to Portfolio Optimization
  • Aranha and Iba, Optimization of the Trading Rule in Foreign Exchange using Genetic Algorithms

Second, applying biologic (ant) algos for complex shape clustering:

Shown that a form of clustering happens from this low level rules. Ant clustering has its positive and negative points when compared to more traditional techniques, such as k-means, but it can be very useful for the clustering of data that is shaped into complex shapes.

Having never used ant clustering in practice, a healthy dose of skepticism seems warranted (if only due to its name).

Full list of Aranha publications includes earlier work on more classic GAs.

5 Comments leave one →
  1. February 18, 2010 7:08 am

    Ant Colony Optimization (and other Swarm Algorithms, especially those using Stigmergy) can be extremely effective at clustering / detection. I’ve seen Particle Swarm Optimization (PSO) used for real-time face recognition and work extremely well. Mother nature evolved these techniques over billions of years — they are effective for a reason. Perhaps not always computationally efficient (and you can’t really prove a lower-bounds on their accuracy), but they get the job done.

    • quantivity permalink*
      February 18, 2010 9:47 am

      @Corey: thanks for your comment. Do you have specific literature references you recommend which illustrate effective applications?

  2. Chronos Phenomena permalink
    February 19, 2010 9:34 am

    It’s not the optimization algorithm… But how you define objective function… And avoid overfitting in my opinion.

  3. February 19, 2010 2:15 pm

    Have a look at Particle Swarm Optimisation (PSO), it can be used for many of the same tasks as GA and you’ll find that it is far more effective. GA rely on random mutations which sometimes produce a good solution but often not. PSO is a directed search that nearly always comes up with the goods. Great blog BTW.

    • quantivity permalink*
      February 19, 2010 9:52 pm

      @Andrew: thanks for suggestion and complement. Can you recommend text or literature references to dive into PSO, either theory or finance applications?

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: