User-generated content (UGC) provides us with a rich channel to aggregate public information. The study examines whether UGC could predict the failure of P2P lending platforms, which also depends on crowds’ involvement and contribution. Specifically, we examine a web forum with user comments on the P2P platforms and extract consumer opinion valence and opinion concentration with natural language processing (NLP) techniques. Time-dependent Cox regression analysis shows that consumer opinion has a strong prediction power for platform survival. If consumer opinions on a platform are positive and concentrated, a platform is more likely to survive. Besides, the two factors show a substitution effect to each other. This study deepens our understanding of UGC, which has significant implications for research. It also forms an application of information technology on risk assessment, by expanding the risk indicators to UGC-related information sources.

用户生成内容（UGC）为我们聚合公共信息提供了丰富的渠道。该研究探讨了 UGC 是否可以预测 P2P 借贷平台的失败，这也取决于人群的参与和贡献。具体来说，我们检查 P2P 平台上的用户评论网络论坛，并使用自然语言处理 (NLP) 技术提取消费者意见效价和意见集中度。与时间相关的Cox回归分析表明，消费者意见对平台生存具有很强的预测力。如果一个平台上的消费者意见是积极的、集中的，那么这个平台就更有可能生存。此外，这两个因素还表现出相互替代效应。这项研究加深了我们对UGC的理解，对研究具有重要意义。它还将风险指标扩展到教资会相关信息源，形成信息技术在风险评估中的应用。