It’s difficult to screen the main loss-reduce factors effectively and analyze the urgency of the reconstruction with a large number of mid-voltage feeders. In this case, a program of large-scale mid-voltage feeders’loss-reduce reconstruction is provided based on clustering analysis and improved ordinal relationship method. Firstly, considering the factors which affect the feeder loss, an index system and its evaluation function are established to evaluate the priority of loss-reduce construction. Then, fuzzy Cmeans clustering is performed on mid-voltage feeders according to the evaluation indexes of each feeder, and the weights of each clustering index is determined by improved ordinal relationship method. Thus, the comprehensive evaluation value of the loss-reduce reconstruction of all kinds of feeders is calculated, as well as the priorities of the schemes. The case of 861 10kV feeders in a city power supply bureau is analyzed and compared with other 2 modification schemes, which verifies the effectiveness of the proposed scheme.