Demand-Driven Acquisitions (DDA) have played an important role in academic libraries. Identifying driving forces of the DDA process and therefore being able to predict the purchasing triggers are beneficial to library collection management, budget planning, librarians’ reference, and the understanding of patrons’ behaviors. However, the literature to date has shown a lack of robust and accurate solutions to the issue. We propose a machine-learning approach called adaptive boosting (AdaBoost) to predict DDA purchasing patterns. As we can show, the results of our simulation studies show that the boosting algorithm possesses higher predictability than traditional logistic regression—an analytic approach adopted in the most current literature. This presentation provides a useful quantitative toolkit for library applications of machine learning.