Compaction quality is an important part of quality assessment of earth-rock dam. Current evaluation methods tend to use the graphical reports of compaction parameters based on real-time compaction monitoring system, or built compaction quality prediction model, such as the multiple linear regression model and the neural network model, to evaluate compaction quality of storehouse surface. However, the precision of these models is not enough. In this study the bacterial foraging-support vector regression algorithm is proposed and employed to establish a high prediction model between multiple attributes and compactness to evaluate compaction quality of storehouse surface. This compaction quality evaluation method gets rid of the one-sided shortcomings of the compaction graphical reports, and greatly improve the prediction accuracy compared with current models, which provides reasonable compaction quality evaluations for quality management works.