Urban traffic congestion severely restricts lane change and overtaking opportunities, making the study of car follower behavior essential for improving traffic flow and safety. This research leverages large-scale vehicle trajectory data and machine learning techniques to develop high-precision vehicle follow-up models tailored to different driving styles. A data set of 100 vehicle trajectories is preprocessed using double exponential smoothing to mitigate measurement noise. While traditional models like IDM [1] and Gipps’ model [2] have laid foundational principles, their reliance on instantaneous parameters limits adaptability to dynamic scenarios. Recent advancements in hybrid machine learning architectures [3, 4] and deep temporal models [5, 6] offer promising solutions. This research leverages these innovations to develop a hybrid LSTM-RF framework, combining the robustness of ensemble learning [7] with the sequential modeling power of LSTM [8].