Stock Analysis Notes

Change of direction (ironic ;-):

Measured the dimensions of some price series - using the counting boxes method (executed in Excel VBA):

FTSE 1.8241

BP 1.425

GKSR 1.1058

They are progressively less fractal.

Now for the FTSE and BP it seems that a velocity approach is not the most appropriate approach since dy/dx is not defined for fractals. While moving averages are appealing in watching trends: I wonder what theoretical foundation exists for their use now.

The shift in approach now (altho I will return to other approaches) is in search of attractors that drive stock charts.

===

This paper is exactly what I wanted to do all laid out tho I've not got the IFS to work yet:

http://it.swufe.edu.cn/UploadFile/other/xsjl/sixwuhan/Paper/IF132.pdf

===

From discussion board

http://www.ai-stockmarketforum.com/showthread.php?t=8

As so many fields investigate variables which change over time, there is widespread interest in the analysis of time-series. Probably three of the most well-developed time-series specialties (there is sginficant overlap, despite varying terminology) are: business forecasting, spectral analysis and digital signal processing. For further research, try these keywords:

-Business Forecasting: "moving average", exponential smoothing, dynamic regression, deseasonalization, ARIMA, autocorrelation, correlogram

-Spectral Analysis: "Fourier transform", FFT ("fast Fourier transform), DCT "discrete cosine transform", wavelet, coherence

-Digital Signal Processing: FIR ("finite impulse response"), IIR ("infinite impulse response"), Wiener

Of course, many other, more generic, techniques have been applied to time-series. I suggest an on-line search for terms like "fuzzy logic" or "neural network" in combination with time-series terms such as "forecast", "prediction" or "denoise".