A recommended method of electricity sale package based on similarity matching and dual attention mechanism

When recommending electricity sale packages, due to the relatively sparse information data of new customers and the lack of sufficient historical records and preference information, the system may be difficult to accurately capture their real needs and preferences, which may cause a large recommendation bias. This bias will not only affect the satisfaction and trust of new users, but also reduce the effectiveness and reliability of the entire recommendation system. Therefore, this paper proposes a similar matching knowledge graph attention network for electricity sale packages recommendation. First, based on the similarity matching algorithm, the knowledge graph is constructed to provide support for new users to match similar historical users. Then, based on the dual attention mechanism, the personalized representation of the user to the electricity sale packages is constructed to solve the incomplete problem of the attribute weight information of the user's electricity sale packages. Then, based on the equivalent evaluation information of similar users, the satisfaction of new users' packages is evaluated, and the packages is recommended in full order. Finally, the accuracy and effectiveness of the proposed electricity sale packages recommendation method are verified by an example analysis of users in a certain area of China.