Forecasting business software piracy rates A machine-learning approach

Abstract

The aim of this research is to examine how effective and accurate machine-learning algorithms can be in predicting business software piracy rates. Computer software is mainly protected by copyright law. Cross-national data on software piracy rates are extracted from the Business Software Alliance (BSA). For that purpose, we employ a sample of 96 countries over the period 2000–2014. We implement the linear regression, and support vector regression models for a wide set of features. The choice of features (or explanatory variables) is based on previous empirical literature on business software piracy. In particular, we conduct a comparison performance of both machine-learning algorithms. Our results show that the support vector regression model has the lowest error rate in comparison to the linear regression model. Future research implications are also discussed.