The existing distribution network node vulnerability assessment methods face problems such as the difficult selection of indicators and one-sided weights of indicators, and they thus fail to be used in the assessment of vulnerable nodes in active distribution networks. To address these issues, a vulnerable node assessment method of active distribution networks based on the improved K-shell mixed degree decomposition (MDD) is proposed. Firstly, an improved K-shell MDD based on the information entropy theory is proposed to divide the distribution network node vulnerability hierarchy; subsequently, a stochastic output model of distributed generation based on kernel density estimation and Copula theory is established by combining the geographic location and stochastic characteristics of the distributed generation; finally, a vulnerable node assessment method of distribution networks based on the risk theory is proposed with the node operation risk as a weighting correction factor. The proposed method can effectively assess the vulnerable nodes of active distribution networks and has stronger computational efficiency when facing large-scale or ultra-large-scale vulnerable node assessment of distribution networks. The feasibility and superiority of the proposed method are verified by analyzing the case of the IEEE 123 system.