Abstract: Temporal data analysis helps to make accurate predictions and decisions by capturing the changing trends of temporal data and revealing the underlying rules and patterns. In recent years, inspired by the strong generalization ability of large models, many works have extended the large models originally designed for natural language processing to the field of temporal data analysis, which empowers the temporal data analysis models with zero-shot and multimodal inferences. However, there are fewer systematic classifications and discussions on temporal data analysis algorithms based on large models in the existing works. The authors classify the related works of temporal data analysis models based on large language models and temporal foundation models respectively according to the directions of targeted optimization in the large model pipeline, describe the various types of methods for applying the large models to the temporal data analysis algorithms, evaluate the strengths and weaknesses and analyze the applicability scenarios of each type of method, to provide methodological references for applying the large model to the temporal data analysis model. Then it is concluded that the temporal data analysis models based on large models perform strongly under the zero-shot inference task. Because the temporal data analysis models based on large language models can utilize the pre-training knowledge, they exhibit better generalization ability with lower computational cost compared to the temporal foundation model. However, the analysis performance is not as good as that of the temporal foundation model. Finally, the remaining challenges and potential future research directions are highlighted.