The increasing scale and topological complexity of distribution systems have led to a sharp rise in maintenance demands. Since operations inherently carry operational risks, this paper constructs an risk assessment index system for maintenance and proposes a novel data-driven risk assessment framework based on the XGBoost algorithm, simultaneously considers the impact of distribution network resilient reconfiguration, and distributed energy resources (DER). The model captures nonlinear interactions between physical indicators (e.g., load loss ratio, voltage violations) and structural metrics (e.g., node isolation, topological centrality), providing accurate and interpretable risk classification. The paper conducts tests based on the improved IEEE 123 node examples and, in conjunction with the assessment results, analyzes the impact of resilient resources on the maintenance risk of distribution system. The examples validate the effectiveness and feasibility of the maintenance risk assessment method.