Authors:
P. Kroha
and
K. Kröber
Affiliation:
University of Technology Chemnitz, Germany
Keyword(s):
Classification, Market time series, Fractal analysis, Fuzzy technology, Stocks, Market behavior.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence
;
Decision Support Systems
;
Enterprise Software Technologies
;
Knowledge Acquisition
;
Knowledge Engineering and Ontology Development
;
Knowledge-Based Systems
;
Software Engineering
;
Symbolic Systems
Abstract:
In this paper, we propose an improvement of a method for market time series’ classification based on fuzzy and fractal technology. Usually, the older values of time series will be cut off at a specific time point. We investigated the influence of the fractal features on the classification result. We compared a normal time series representation, a representation having a smaller box dimension (achieved by exponential smoothing), and a representation having a greater box dimension (achieved by addind scaled noise). We used different types of noises and scales to improve the classification result. Our application concerns time series of stock prices. The market performance of those approaches is analyzed, discussed, and compared with the system without the scaled noise component.