Anyone Use Kalman Filtering?
There's a small firm which uses Kalman filtering to track the markets. The predictive accuracy is reported to fade about 3 months out. This is the foecasting ability I read about in a Federal Reserve white paper 10 years ago. (Darn, I lost it).
So the last couple of weeks I have begun trying the Kalmann filter technical analysis approach. My initial studies involved trying to track a combination of 3 sinusoids. Hurst theory clearly proves the market is a collection of cycles, and I've verified using classical signal procwssing techiques, so a small collection of sinusoids is reasonable. The results for the initial studies were extremely promising. As long the period was more than 8 samples, the target was tracked accurately.
When I used data from the Dow Jones Industrial Average, the predictive ability goes down the drain. This is most likely due to a noisy signal.
Any thoughts?