A new method to diagnose open-line fault of the branch line of the medium voltage distribution network without ground is presented based on electric power distribution information system data and association rules algorithm. First through the analysis of the correlated data of electric power distribution information system, a method of mining association rules based on data feature selection is established. Secondly the continuous features is converted to Boolean feature by chi-square splitting algorithm, at the same time the MSApriori algorithm is used to solve the problem of rare items in the fault information, and then on the basis of this, the kulc rule is applied to eliminate the redundant rules to form the simple rule family. At last the practical numerical analysis based on the historical data of electric power distribution information system in an area in East China shows that the proposed method can reduce the invalid mining greatly and improve efficiency and accuracy significantly, which applies to on-line diagnosis of open-line fault in medium-voltage distribution network.