Improved Parametric and Nonparametric Classification Techniques

This text considers different parametric and nonparametric classification techniques to classify objects, and make a comparative study among these techniques. In most of the situations, classification techniques give few misclassifications under large samples as well as under the normal populations. If the data set comes from the non-normal populations, then we apply Box-Cox transformation to transform this data set into near normal. Hence, we investigate the effect of Box-Cox transformation and see that Box-Cox transformed data generates better discrimination and classification techniques. Also if the sample size is small, then we use the Bootstrap approach for classifying objects, and investigate that the Bootstrap classification technique used in this analysis performs better than the usual techniques of small samples. There is no unique classification technique that is suitable for all the situations, also examines that nonparametric classification techniques perform better than the parametric classification techniques, whereas the Neural Network classification technique gives optimum solutions among the nonparametric classification techniques.