The paper introduces LC-LLM, a novel model for lane change prediction in autonomous driving systems. LC-LLM leverages the strong reasoning and self-explanation capabilities of Large Language Models (LLMs) to predict lane change intentions and trajectories. The model reformulates the lane change prediction task as a language modeling problem, processing heterogeneous driving scenario information as natural language prompts for LLMs. Supervised fine-tuning is employed to tailor the LLMs specifically for lane change prediction, and Chain-of-Thought (CoT) reasoning is used to enhance prediction transparency and reliability. During inference, explanatory requirements are included in the prompts to enable the model to provide CoT reasoning and explanations for its predictions. Extensive experiments on the highD dataset demonstrate that LC-LLM outperforms baseline models in lane change intention prediction (17.7% improvement) and trajectory prediction (64.4% improvement in lateral and 66.1% improvement in longitudinal). The model also shows superior interpretability, with a CoT reasoning accuracy score of 97.2. Future work includes extending the model to urban driving scenarios and improving inference speed.The paper introduces LC-LLM, a novel model for lane change prediction in autonomous driving systems. LC-LLM leverages the strong reasoning and self-explanation capabilities of Large Language Models (LLMs) to predict lane change intentions and trajectories. The model reformulates the lane change prediction task as a language modeling problem, processing heterogeneous driving scenario information as natural language prompts for LLMs. Supervised fine-tuning is employed to tailor the LLMs specifically for lane change prediction, and Chain-of-Thought (CoT) reasoning is used to enhance prediction transparency and reliability. During inference, explanatory requirements are included in the prompts to enable the model to provide CoT reasoning and explanations for its predictions. Extensive experiments on the highD dataset demonstrate that LC-LLM outperforms baseline models in lane change intention prediction (17.7% improvement) and trajectory prediction (64.4% improvement in lateral and 66.1% improvement in longitudinal). The model also shows superior interpretability, with a CoT reasoning accuracy score of 97.2. Future work includes extending the model to urban driving scenarios and improving inference speed.