By aiming at huge user groups in the power market and the difficult selection of power selling packages in the trading process, a recommendation method of power selling packages based on the optimal feature subset of power trading users was proposed. First of all, the Optimal Feature Subset of Power Trading Users (OFSPTU) was defined, and the optimal subset discovery algorithm based on coverage rate with weighing increase was designed, the optimal user feature subset could be screened out reasonably from mass trading users. Next, the similar computing method of power selling packages based on attribute correlation was proposed, in the form of clustering and confirming package attribute weighing, the package similarity was calculated, obtaining the nearest neighbor of package projects. At last, the accurate recommendation of power selling packages based on the OFSPTU and package attribute was realized. The experiment shows that the effectiveness of OFSPTU and accuracy of the recommended algorithm.