For high levels autonomous driving functions, the Decision Layer often takes on more responsibility due to the requirement of facing more diverse and even rare conditions. It is very difficult to accurately find a safe and efficient lane change timing when autonomous vehicles encounter complex traffic flow and need to change lanes. The traditional method based on rules and experiences has the limitation that it is difficult to be taken into account all possible conditions. Therefore, this paper designs a lane-changing decision algorithm based on data-driven and machine learning, and uses the DQN (Deep Q Network) algorithm in Reinforcement Learning to determine the appropriate lane-changing timing and target lane. Firstly, the scene characteristics of the highway are analyzed, the input and output of the decision-making model are designated and the data from the Perception Layer are processed. The DQN network architecture is built, and the reward function is designated considering multi-objective requirements. Then, a virtual scene is built in PreScan, and random factors are added to simulate various scenarios that vehicles might encounter when making lane change decisions. Based on this scenario, the Reinforcement Learning model is trained. Finally, a bench test comparing between the trained Reinforcement Learning model and rule-based model and a real vehicle test on a highway are carried out to verify the effectiveness of the algorithm. By analyzing the results of both bench test and real vehicle test, it can be concluded that the lane-changing decision model based on Reinforcement Learning not only improves driving efficiency but also ensures safety under subjective and objective assessment, and has better adaptability to different working conditions, which is closer to human decision.