Classification rules for multi-level multivariate data

ANURADHA ROY; RICARDO LEIVA

Kumamoto City International Center, Kumamoto, Japón

Conferencia; “Session on Intelligent Technique for Time Series Data Mining” del 2nd International Conference on Innovative Computing, Information and Control; 2007

In this article we study new classifications rules for multiple m-variate observations over u-sites and over v-time points under the assumption of multivariate normality. We assume that the m-variate vector of observations has "jointly equicorrelated" covariance structure. The new classification rules are effective in discriminating individuals when of observations is very small and thus unable to estimate the unknown variance-covariance matrix. The classifications rules are demonstrated by using a real world data set. Our results show that the performances of our new classification rules are superior to the traditional classification rule.