Using Classification Trees to Predict Performance in Information Technology Projects

Abstract

Classification trees are introduced as a modeling technique to predict IT project performance. A comparative analysis of classification tree, regression and neural network techniques provides promising evidence that classification trees can provide more actionable output for project decisions. The relative accuracy from classification trees, vis-à-vis regression and neural network techniques is demonstrated using a data set of 440 projects including 8 performance factors and a binary dependent variable of performance. Results suggest classification tree techniques provide comparable, and possibly superior, predictive capabilities. We conclude that a performance assessment based on classification trees could provide an effective decision tool for managing IT project performance.