Object Classification with Classical Linear Discriminant Analysis and Robust Linear Discriminant Analysis
Justin Eduardo Simarmata · International Journal for Research in Applied Science and Engineering Technology · 2018
Discriminant analysis is one of multivariate analysis with dependency method. Discriminant analysis is a multivariate analysis that aims to classify observations based on several independent variables that are non-categorical and categorical dependent variables. Discriminant analysis requires the assumption of the normal multivariate distribution and the homogeneity of the variance-covariance matrix. Classical linear discriminant analysis and linear robust discriminant analysis can be applied to classify objects. The classification is based on 10 indicators of district/city poverty level in North Sumatra province, 10 indicators are as independent variable and low poverty classification and high poverty classification as dependent variable. The linear robust discriminant model classifies the object more precisely than the classic linear discriminant model. This can be seen from the total proportion of classification mistakes of 6%, less than the total proportion of classical linear discriminant classifier error classification of 21.2%. This is due to the large number of outlays in the district/city poverty data in North Sumatra.