Abstract: The lane change decision directly affects the autonomous lane change of intelligent vehicles in complex traffic environments，yet current process of decision-making is afflicted with low prediction accuracy and poor safety.To address these problems，this paper proposes a lane change decision model based on driving scenarios and decision rules.First，the new decision feature variables，desired velocity after lane change and distance difference from the vehicles before and after lane change，are introduced，considering the influence of the traffic conditions of post-lane change.The lane change decision rules are made based on the correlation between the feature variables and the lane change decision，considering the human decision logic.Then，the lane change scenarios dataset simulating the real-time driving environment is built and validated，which augments the NGSIM dataset.The support vector machine model based on the Bayesian optimization kernel function is proposed for the multi-parameter and nonlinear problem of lane change decision.Finally，the model is tested and validated on the lane change scenarios dataset.Our comparison results show the newly introduced decision feature variables exert positive effects on lane change behavior and the lane change scenarios dataset simulates the real-time driving conditions，which can be further applied to the research of lane change decision-making and trajectory planning.The support vector machine achieves a prediction accuracy of 95.40%，higher than other machine learning classifiers，improving the safety of lane change behaviors.