Handling of Missing Values in Static and Dynamic Data Sets

This thesis contributes by first, conducting a comparative study of traditional and modern classifications by highlighting the differences in their performance. Second, an algorithm to enhance the prediction of values to be used for data imputation with nonlinear models is presented. Third, a novel algorithm model selection to enhance prediction performance in the presence of missing data is presented. It includes an overview of nonlinear model selection with complete data, and provides summary descriptions of Box-Tidwell and fractional polynomial methods for model selection. In particular, it focuses on the fractional polynomial method for nonlinear modelling in cases of missing data. An analysis ex- ample is presented to illustrate the performance of this method.