This paper investigates the revenue maximization problem in mobile edge computing (MEC) systems, where an access point provides computation offloading services to multiple users and charges service fees. While deep reinforcement learning approaches achieve optimal performance, their computational complexity limits practical deployment. This paper proposes a simplified policy gradient algorithm that maintains near-optimal revenue performance with significantly reduced complexity. By employing radial basis function (RBF) features and experience replay, the proposed approach achieves within 5% of optimal revenue while using only 9 parameters compared to approximately 20,000 parameters in deep learning methods. Simulation results demonstrate that the proposed algorithm converges faster than both conventional Q-learning and deep policy gradient methods, achieving a 95% reduction in training time while maintaining competitive performance......................... Keywords:....... Mobile edge computing, dynamic pricing, policy gradient, reinforcement learning, revenue maximization, basis functions.