Friday, May 23, 2008

Machine Learning + Regime Switching = Profitability?

My article on a trading strategy based on regime switching and machine learning techniques is now available on Automated Trader magazine (subscription required). The software I used to research this model is Alphacet Discovery, an industrial-strength backtesting, optimization, and execution platform.


Damian said...

Anything you can share about the platform itself? It looks very expensive.

Ernie Chan said...

Alphacet Discovery has an integrated historical high-frequency equity and futures database for backtesting, and a suite of machine learning algorithms (e.g. neural net and genetic algorithm) for parameter optimization. Many of these operations can be done with the mouse only, without writing any programming codes.

Furthermore, once a trading algorithm is backtested and optimized, it can be immediately switched on for paper/real trading with no further work.

It is primarily targeted to institutional investors.

Ziqing said...
This comment has been removed by the author.
Ziqing said...

I am wondering how you solve the optimization problem in real time regime-switch models. Usually the optimization for estimating one models is at seconds level for each input. Unless you use simple regime-switch models like threshold / homo-markovian models, but I really do not believe these static regime-switch models can achieve any stable profit.

Also, neutral network models face significant over-fitting problem, which leads to high frequent rebalancing. I only agree that neutral network works in field such as FX where the transaction cost is small.