Tree-based methods in supervised learning with Estonian Health Insurance Fund data

Abstract

The main aim of this master’s thesis work is to provide an overview of some tree-based models and to test the suitability of these models in finding the incorrectly submitted invoices received by the Estonian Health Insurance Fund. C4.5, CART and bagged CART are the three algorithms that are used to train the models and to apply binary classification with these models in order to reduce the number of invoices that must be checked manually.