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.