The emergence of ideation methods that generate large volumes of early-phase ideas has led to a need for reliable and efficient metrics for measuring the creativity of these ideas. However, existing methods of human judgment-based creativity assessments, as well as numeric model-based creativity assessment approaches suffer from low reliability and prohibitive computational burdens on human raters due to the high level of human input needed to calculate creativity scores. In addition, there is a need for an efficient method of computing the creativity of large sets of design ideas typically generated during the design process. This paper focuses on developing and empirically testing a machine learning approach for computing design creativity of large sets of design ideas to increase the efficiency and reliability of creativity evaluation methods in design research. The results of this study show that machine learning techniques can predict creativity of ideas with relatively high accuracy and sensitivity. These findings show that machine learning has the potential to be used for rating the creativity of ideas generated based on their descriptions.