A comprehensive lane-change prediction method integrating driving style recognition, vehicle characteristics, interaction features, traffic flow characteristics, and truck ratio was proposed, in order to address the limitations of existing lane-change prediction models in modeling the multi-vehicle driving styles of lane-changing and surrounding vehicles and their interaction features, as well as their insufficient consideration of the influencing factors of lane-changing behavior. Based on the HighD dataset, basic and interaction features of lane-changing and surrounding vehicles were extracted, and statistical indicators under different time windows were calculated. Principal component analysis and K-means clustering were employed to quantify multi-vehicle driving styles during lane-changing. Subsequently, Long Short-Term Memory (LSTM) networks and double-layer Convolutional Neural Networks (2-CNN) were used to compare predictive performance across time windows. Results showed that incorporating multi-vehicle driving styles and interaction features improved accuracy by 5.64%, with the LSTM achieving a maximum F1-score of 99.26% under a 2.0-second window. Furthermore, the combined contribution of vehicle length, truck ratio, traffic density, and the interaction-aware driving style features of lane-changing vehicles reached 11.08%. Overall, the proposed method enhanced the accuracy of lane-change behavior prediction and the interpretability of influencing factors, providing insights for autonomous driving and traffic management.