Booming noise influences perceived vehicle quality and passenger comfort. Therefore, it becomes important for automotive manufacturers to have a reliable method for the detection of booming noise. Machine learning models can effectively address this need. However, in the absence of sufficient data samples, machine learning models have been known to perform poorly. In this paper, we study the efficacy of simulation-driven machine learning and transfer learning techniques to overcome this data scarcity problem. In this study, the source and target datasets are synthesized starting from a limited set of original recordings. The machine learning database is obtained through modification of baseline vehicle runup sounds, by defining booming events on a pre-existing vehicle order profile. A modified definition of booming noise and change in vehicle orders and background noise are the primary differences between the source and target datasets. We use transfer learning techniques to leverage the knowledge gained from the source dataset to make better predictions in the target dataset.