With the development of smart distribution network, the distribution automation system (DAS) is increasing rapidly. It is urgent to evaluate the state operation performance of DAS comprehensively. Based on the establishment of hierarchical evaluation index system for the performance, the indexes are clustered to train the index classifier by decision tree algorithm to analyze different types of the state operation automatically. Then, we use the entropy method to give weight to each index, and constructe the training sample set of comprehensive evaluation by using the weighted results and the original data of the indexes. Finally, we use the multiple regression algorithm to train the evaluation model for each state operation category. As shown by the experimental results, the evaluation model can objectively reflect the comprehensive state operation performance of DAS at the corresponding time section. Moreover, it obtains higher accuracy than the traditional neural network evaluation model and provides effective evaluation method and decision supporting for the operation management of DAS.