Abstract:Objective：Due to the traditional maintenance methods relying on human experience and a single repair measure，the usage and turnover of endoscopes have decreased. This article proposes a data-driven solution. Through machine learning methods，analyze historical maintenance records，extract key features of fault text using TF-IDF，and construct a fault classification model based on the random forest algorithm. The experimental results show that the model has high accuracy in identifying minor and evolving faults，effectively reducing unplanned downtime of endoscopes and lowering maintenance costs. First，the TF-IDF method was used to extract key features from the fault text in historical maintenance records. Second，a fault classification model was constructed based on the random forest algorithm，and a visual decision tree was used to accurately recommend maintenance measures. Results：The model achieved an accuracy of 90% and 69.7% for identifying minor faults（level 3） and evolving faults（level 2），respectively，significantly outperforming traditional empirical methods. In addition，through visualization methods，the model can intuitively guide daily maintenance work and improve maintenance efficiency. In the future，research will further enhance the digital prevention and maintenance level of endoscopes by expanding data volume and integrating more advanced mining methods. To provide support for digital management of medical electronic endoscopes.