Abstract: 【Purpose/significance】This research aims to explore the public's sentiments and key concerns regarding ChatGPT and large language models, which can assist AI companies in advancing AIGC technology development and provide recommendations for governments in formulating corresponding regulatory strategies.【Method/process】After web-scraping and preprocessing data from Weibo, a BERT model was constructed for sentiment classification. The Latent Dirichlet Allocation (LDA) model was used for topic clustering, and the ARIMA time series model was employed to reveal the public's focal points and sentiment tendencies toward Chat‑ GPT, as well as to predict the emotional trajectory of user comments【. Result/conclusion】The public's attitudes toward large language models, represented by ChatGPT, vary among individuals, with a relatively balanced distribution of sentiments. Users' key concerns are characterized by increasing complexity, and there are positive expectations for public recognition and emotional attitudes toward large language models like ChatGPT in the future.【Innovation/limitation】By integrating fine-grained sentiment classification with LDA topic clustering and combining coarse-grained classification results with the ARIMA time series model, this study overcomes the limitations of relying solely on static sentiment analysis. It provides a multi-perspective and comprehensive analysis of the public's at‑ titudes, concerns, and future expectations regarding ChatGPT, offering in-depth insights into the developmental trends and potential risks of large language models like ChatGPT.