作为智能情报分析中的重要应用场景,算法推荐提供的个性化和精准化信息服务为现代快速决策增加了价值，但算法推荐风险问题也尤为突出，探寻算法推荐风险影响因素对科学地提出算法风险治理策略至关重要。文章采用LDA模型对科研论文进行主题聚类，聚类结果与《互联网信息服务算法推荐管理规定》进行相似度计算，以识别算法推荐风险影响因素，从风险产生和风险治理两个维度构建系统动力学模型，然后利用Vensim PLE软件和文本计算数据进行仿真与灵敏度分析。研究识别出算法素养、大数据技术、算法偏见、网络安全审查等影响因素，通过文本计算获得的数据进行仿真能够较好的拟合算法推荐风险治理现实情况，并基于灵敏度分析提出如下建议：加强算法素养教育，提高个人隐私保护意识；建立算法全流程监管机制，提升算法的可解释性；建立“制度+技术”机制，提高平台风险防范能力。

As an important application scenario in intelligent intelligence analysis, personalized and precise information services provided by algorithm recommendation add value to modern fast decision-making, but the problem of algorithm recommendation risk is also particularly prominent. Exploring the influencing factors of algorithm recommendation risk is crucial to scientifically proposing algorithm risk governance strategies, so this study has important practical significance. This paper using the LDA model for research paper topic clustering, clustering results should be compared with the "Regulations on the Recommendation Management administration of Internet Information Service Algorithms" for similarity calculation, to identify the risk influencing factors of algorithm recommendation, the system dynamics model is constructed from two dimensions of risk generation and risk governance, and then use Vensim PLE calculation software and text data to analyze simulation and sensitivity. Identify algorithm literacy, big data technology, and algorithm of prejudice, network security, and other factors, for the calculation of the data obtained through the calculation of the text to a better fitting method recommended management reality, and put forward the following suggestions based on sensitivity analysis: strengthen algorithm literacy education, to improve personal privacy protection awareness; Establish the whole process supervision mechanism of the algorithm to improve the interpretability of the algorithm; We will establish an "institutional and technological governance" mechanism to improve the platform's ability to prevent risks.