As society develops, the comfort and safety characteristics of automobiles are increasingly focused on. Much industrial and academic research has been therefore involved in the study of active and semi-active suspension control systems to improve ride comfort and road holding of automobiles. The new control methods for these systems are still being studied and improved to increase these two performances. Recently, Machine Learning has created rapid development in many fields of science and technology. In which, Reinforcement Learning has been applied in vehicle dynamics and control systems to increase both ride comfort and road holding. Along with that trend, this paper applies Deep Reinforcement Learning for controlling a quarter-car active suspension model based on the Deep Deterministic Policy Gradient (DDPG) algorithm. The changing of the reward function is key to ride comfort and road safety enhancement. The reward function is designed to reduce the sprung mass displacement. The simulation results show that the proposed DDPG active suspension system increases ride comfort compared with the passive suspension system.