Abstract

Given data in a time series we will create a phase space using methods based upon the work of Takens and Whitney. Our phase space will be approximated using a single record observed s (n ) of the New York Stock Exchange. This procedure of creating a phase space will create a complete vector space by defining s (n ) to be the first coordinate, s (n + T ) the second and s (n + ( DE - 1)T ) the last coordinate, where T is a suitable delay and DE is the embedding dimension. The observed phase space will be shown to be chaotic in its behavior and a reconstructed attractor in the phase space will provide us with predictions of future the stock market prices. All algorithms for computation are written in Borland C++ version 5.