INDICATORS
Component Hunt
Forecasting Singular Spectrum Analysis
by Sergiy Drogobetskii and Vladimir Smolynsky
Find components in the markets such as trends and random noise that can help you analyze and forecast the financial markets.Applying mathematical statistics to analyze and forecast the commodity and financial markets is nothing new. Traders have been applying classical models of trend plus noise or autoregression (moving average) and come up with satisfactory results. But those models tend to be simple in structure. The peculiarity of market behavior is that its characteristics (price, amount of transactions, indicators, and so forth) are made up of several components: a slow component, which is the trend; periodical, or oscillatory; and a random component described by the randomness of the price series. The periodical component is made up of periodicity with variable period and amplitude.
The classic methods of analysis such as Fourier analysis, regression analysis, or wavelet analysis tend to decompose the initial function into a series using a fixed system of basic functions such as sines and cosines, which produce strong periodicity property. In this article, I will discuss singular spectrum analysis (SSA), which is data-adaptive and a nonparametric method.
WHAT IS SINGULAR SPECTRUM ANALYSIS?
Singular spectrum analysis looks at the principal components of the time series method but doesn't need any preliminary stabilization of the series. Ssa allows you to analyze the structure of a time series, single out its separate components, and forecast both the series and its component development tendencies. Being able to visualize the calculation data and to apply it in a practical manner is what makes this method stand out among others.
THE PREMISE
The first idea that underlies the method is repeatability. This is done by transitioning from time series to something with structure. One example is a transition from a sequence of prices that are equidistant to a sequence of vectors consisting of time series segments of a certain length. So if the initial series has some structure, its segments inherit this structure.
...Continued in the July issue of Technical Analysis of STOCKS & COMMODITIES
Excerpted from an article originally published in the July 2008 issue of Technical Analysis of STOCKS & COMMODITIES magazine. All rights reserved. © Copyright 2008, Technical Analysis, Inc.
Return to July 2008 Contents