更科学准确地识别以及预测撤销论文的特征，更好地纠正科研失范、减少学术欺诈行为、保护科学的公平性。

基于PubPeer平台上受质疑论文相关数据信息，经数据处理与筛选，建立包含1792篇完整论文数据的撤销论文数据集，分析该部分撤销论文自然属性和评论属性特征，并通过构建PubCancel模型进行撤销论文特征识别及准确性检验。

构建的PubCancel模型可准确、有效地识别撤销论文，撤销论文状态验证准确率高达98.24%。该方法不仅对于评估论文质量具有重要的实际意义，也可为研究人员提供一种快速评估论文可信度的方法。

研究PubPeer平台质疑论文中撤销论文的情况对学术预警研究有重要意义，应用撤销论文特征识别模型能够帮助期刊及编辑及时发现异常论文，规范论文发表并及时撤回问题论文，加强科研诚信建设。

This study aims to scientifically and accurately identify and predict the characteristics of retracted papers, better correct scientific misconduct, reduce academic fraud, and protect the integrity of science.

We collected data and information related to questioned papers from the PubPeer platform. After data processing and screening, a dataset of 1792 retracted papers with complete data was established. The natural and comment attributes of these retracted papers were analyzed, and a PubCancel model was developed to identify the features and assess the accuracy of the retracted papers.

The PubCancel model developed in this study accurately and effectively identifies the retracted papers, with an accuracy rate of 98.24%. This method has significant practical implications for evaluating paper quality and provides researchers with a fast way to assess paper credibility.

Studying the situation of retracted papers in questioned papers on the PubPeer platform is of great significance for academic warning research. The application of the feature recognition model of retracted papers can assist journals and editors in detecting abnormal papers promptly, standardizing paper publication processes, and promptly retracting problematic papers, thereby enhancing the integrity of scientific research.