Abstract: A method for prediction of multidimensional heterogeneous time series using logical decision functions is suggested. The method implements simultaneous prediction of several goal variables. It uses deciding function construction algorithm that performs directed search of some variable space partitioning in class of logical deciding functions. To estimate a deciding function quality the realization of informativity criterion for conditional distribution in goal variables' space is offered. As an indicator of extreme states, an occurrence a transition with small probability is suggested.

Keywords: multidimensional heterogeneous time series analysis, data mining, pattern recognition, classification, statistical robustness, deciding functions.

ACM Classification Keywords: G.3 Probability and Statistics: Time series analysis; H.2.8 Database Applications: Data mining; I.5.1 Pattern Recognition: Statistical Models

Link:

EXTREME SITUATIONS PREDICTION BY MULTIDIMENSIONAL HETEROGENEOUS TIME SERIES USING LOGICAL DECISION FUNCTIONS1

Svetlana Nedel’ko

http://www.foibg.com/ijita/vol13/ijita13-3-p14.pdf