An Empirical Investigation of the Sensitivity of Inductive Decision Trees

Data mining (DM) techniques have gained a lot of attention lately as they may provide decision-makers with institutional knowledge required for making strategic decisions. Such techniques are able to handle and extract useful information from large data sets. Many commercial tools are now available for the novices to facilitate the data mining process. However, treating these tools as a black box can lead to misleading and unpredictable results. Even when comparing among tools that are based on the same data mining technique, results may vary according to the implementation process and may be affected by the user selection of parameters prior to the mining phase. In this paper, we present an empirical comparison among some of the tools that generate decision trees, and illustrate the difference in their results using the case of loan application processing.