International Journal of Innovative Information Systems

& Technology Research 14(3):519-532, July-Sept, 2026

 © SEAHI PUBLICATIONS, 2026    www.seahipublications.org   ISSN: 2467-8562

 

 

DOI:10.5281/zenodo.23174480

A Hybridized Machine-Learning Model For Predicting Software Piracy Using Social And Behavioural Characteristics

 

Timinipre Amabie*; Laeticia Nneka Onyejegbu; Barilee Baridam

 

Department of Computer Science,

Faculty of Computing,

University of Port Harcourt, Port Harcourt, Rivers State, Nigeria.

*Corresponding author: amabietimi@gmail.com

 

ABSTRACT

This study developed a hybridized machine-learning framework for classifying self-reported software piracy using social, demographic and behavioural characteristics. The dataset comprised 553 respondents and 24 questionnaire items from a SurveyPlanet survey on the pricing and consumption of music-production software. Preprocessing included missing-value treatment, geographical grouping, binary target construction and one-hot encoding. Six candidate feature groups yielded 35 binary features. All 63 non-empty combinations were evaluated across Decision Tree depths of 1–9 using five-fold stratified cross-validation, producing 567 configurations. Expertise Level, Age Group, Region, and Importance of Ease of Pirating formed the selected feature combination, and the optimized standalone Decision Tree had a maximum depth of five. The framework also incorporated AdaBoost with depth-one Decision Trees as weak learners. On the held-out test set of 111 observations, the Decision Tree achieved 73.87% accuracy, compared with 72.97% for AdaBoost. For the Pirate class, Decision Tree precision, recall and F1-score were 67.69%, 84.62% and 75.21%, respectively; the corresponding AdaBoost values were 67.19%, 82.69% and 74.14%. The trained classifier was integrated into a Streamlit application. The findings show that the selected characteristics supported classification of self-reported piracy within this dataset, although boosting did not improve accuracy over the optimized standalone tree. The resulting framework provides a practical approach to examining behavioural patterns associated with software piracy.

Keywords: software piracy; machine learning; Decision Tree; AdaBoost; behavioural prediction; classification.