Evaluating Artificial Neural Networks and Traditional Approaches for Risk Analysis in Software Project Management - A Case Study with PERIL Dataset
Carlos Timoteo, Mêuser Valença, Sergio Murilo Maciel Fernandes · 2014
Many software project management end in failure. Risk analysis is an essential process to support project success. There is a growing need for systematic methods to supplement expert judgment in order to increase the accuracy in the prediction of risk likelihood and impact. In this paper, we evaluated support vector machine (SVM), multilayer perceptron (MLP), a linear regression model and monte carlo simulation to perform risk analysis based on PERIL data. We have conducted a statistical experiment to determine which is a more accurate method in risk impact estimation. Our experimental results showed that artificial neural network methods proposed in this study outperformed both linear regression and monte carlo simulation.