Decision tree induction based on autocortelation function

Abstract

Abslracf-The paper introduces definition of autocorrelation function and theory of perturbation. Time sequence model is different from ordinary linear regression model. The ordinary methods of detecting abnormal value do not have good effectiveness for time sequences such as data deletion method and single point derivation method. In order to detect the strong influence points in time sequences, we discuss the perturbation theory of autocorrelation function with several minute perturbations. A new kind of decision tree based on Autocorrelation Function (AF) is developed. The simulation results show the effectiveness and accuracy of the decision tree.