The high risk of random access collisions leads to huge challenge for the deployment massive Machine-Type Communications (mMTC), which cannot be sufficiently overcome by current solutions in LTE/LTE-A networks such as the extended access barring (EAB) scheme. The recently studied approaches of grouped random access have shown a great potential in simultaneously reducing the collision rate and the power consumption in mMTC applications, and exhibit a good compatibility with the concept of random access resource separation. In this work, we propose an optimized resource dedication strategy for grouped random access approaches, which inherits the advantage of resource separation to isolate device classes from each other, while providing an optional class preference with high flexibility and accuracy, which has been usually implemented with access class barring.