How Chaos Theory improves data mining in research by means of ALEV

In this paper a method for data reduction is introduced. Aspects of Lyapunov, entropy and variance (ALEV) provide an approach for mining large stocks of time series data. Methods of artificial intelligence (AI) offer two different ways for modeling observation data: the recall times of expert systems (XPS) depend on the size of a knowledge base. Connectionist approaches like the multi-layer perceptron (MLP) have to be trained with a representative data set for mapping system behavior. While the duration of this learning process also depends on the amount of representative data the recall times are very short. On basis of the Mackey-Glass function a technique for visual data mining (VDM) is proposed. Performance tests on basis of real world traffic speed patterns from different observation time periods show that ALEV thins out large pattern stocks. Viability of data mining methods is increased and generalization quality remains the same.