摘要
Abstract
Traffic prediction,as a key component of intelligent transportation systems,plays a vital role in traffic planning,management,and control.With the rapid development of artificial intelligence(AI),large language mod-els(LLMs)have demonstrated significant potential in traffic prediction because of their advanced language compre-hension and generation capabilities.Firstly,the concept,development history,core principles and key technical foundations of LLMs were introduced in this paper.Secondly,current traffic prediction methods based on LLMs were summarized.According to various keys,the methods were classified into two categories of data-driven and model-driven approaches.Data-driven methods include tokenization,prompt engineering,and embedding;While model-driven methods involve fine-tuning,zero-shot/few-shot learning,and integration.The applications of LLMs in traffic prediction were proceeded and the technical methods employed by various studies were analyzed.Finally,the challenges faced by LLMs technologies in the field of traffic prediction were discussed,as well as future devel-opment.关键词
大语言模型/交通预测/智能交通系统/人工智能Key words
large language models/traffic prediction/intelligent transportation systems/artificial intelligence分类
交通工程
