Thursday, April 5, 2012

About Particle Swarm Optimization, Part II

I did not elaborate as much as I should have in the previous post, and some people had good questions, which I hope the following will clarify:

1. Optimization is performed over the "Months to test" period - so different periods provide different results and yield different time to completion due to more work to be done.
2. The initial positions do take effect. The algorithm uses the following initial particles:
a. The current portfolio
b. 5 First equities at 100% allocation each.
c. The rest of the particles are random allocations

This means that
a. Results will differ between times the optimization is run.
b. A result can only provide a better sharpe then your current allocation

Cheers!

No comments:

Post a Comment