关键词: 差分整合移动平均回归模型（ARIMA模型）, 信息构建, 研究主题识别, 趋势预测, 可视化

Abstract: ［Purpose/Significance］Identifying the main research topics of journal articles in a specific subject and predicting their development trends could be helpful for understanding the research hotspots and trends in this area. It is also significant for a profound analysis on the development trend in this field.［Design/Methodology］Firstly, the LDA model was used for topic recognition, and the annual probability distribution of each topic was obtained through a userdefined function. Besides, time series data of topic evolution was gained. Then, the ARIMA model was established to predict and analyze the time series of main topics in the field of information architecture.［Findings/Conclusion］Currently, there is a good development momentum in the topics including evaluation indicators of information architecture, information organization, network information and knowledge construction in China. ［Originality/Value］On the one hand, this paper identifies the research topics in the field of information architecture domestically in the past 20 years, and predicts the evolutionary trend of the main topics. On the other hand, this study verifies the feasibility and effectiveness of the topic prediction method proposed in this paper, and provides reference for the research on information architecture and topic prediction.

Keywords: ARIMA model, Information architecture, Identification of research topics, Trend prediction, Visualization