To improve the accuracy of lane change trajectory prediction for autonomous vehicles in complex traffic scenarios, we propose an intent-driven vehicle lane change trajectory prediction model. This model incorporates the weights of lane change behavior characteristic parameters to quantify and adjust the influencing factors of vehicle lane change behavior, enabling the model to simulate and predict the lane change intentions and behaviors of vehicles more accurately. The model adopts a dual encoder-decoder structure, combining Long Short-Term Memory Networks (LSTM), Graph Neural Networks (GNN) and attention mechanisms to extract the temporal and spatial features of vehicle historical trajectories. This model first uses LSTM to capture the time series features of vehicle trajectories, such as speed and steering changes, and then constructs the interaction relationship representation between vehicles through GNN. In addition, an attention mechanism was introduced to enhance the model’s focus on key features, and these features were integrated through a Multi-Layer Perceptron (MLP) decoder to predict vehicle trajectories. The minimum average displacement error and the minimum final displacement error of vehicle lane change trajectory prediction have also reached the benchmark level, effectively improving the missed detection rate.