By virtue of their tunable physicochemical and electrochemical properties, ionic liquids (ILs) provide a promising solution for enhancing the performance and safety of batteries. Toward efficient design of IL-based electrolytes, a reliable electrical conductivity (κ) prediction model is highly desirable. In this work, the COSMO-RS derived QSPR model and its use as a basis for developing boosting machine learning (ML) methods for the κ prediction of ILs are systematically examined. Based on a large experimental κ database, the overall κ prediction performance and the description of temperature and IL structure dependencies by the COSMO-RS derived QSPR model are evaluated thoroughly. Following that, boosting ML based on two powerful ensemble algorithms, namely random forest (RF) and extreme gradient boosting (XGB), are employed to bridge the residual between experimental and QSPR predicted κ. The value of this proposed boosting strategy is evidenced by comparing with ML without boosting and the direct QSPR predictions. The results demonstrate the notably enhanced prediction performance of the boosting ML model and identify the boosting XGB as the best option for κ prediction.