Abstract: In 5G and beyond wireless networks, enhanced mobile broadband (eMBB) and ultra-reliable low latency communications (URLLC) are two core services with different quality of service requirements. Achieving the heterogeneous coexistence of these two services using limited radio resources is a challenging problem. This paper proposes an intelligent resource allocation framework based on the perforation technique introduced by 3GPP, modeling the eMBB/URLLC resource allocation problem as an optimization problem that aims to maximize the eMBB user data rate while satisfying the URLLC reliability constraints. Additionally, considering the uncertainty of wireless channels and the impact of URLLC random perforation, a proportional fairness (PF) algorithm is introduced to balance the trade-off between total throughput and user fairness. To address this issue, this paper proposes a proportional fairness-based deep reinforcement learning algorithm, the proportional fairness-twin delayed deep deterministic policy gradient (PF-TD3A), to intelligently allocate resources for the two services. Experimental results show that the proposed algorithm can further increase the eMBB user data rate while meeting the eMBB reliability requirements, achieving an average improvement of approximately 7.4%.