Design Of Recommendation System For Electricity Sales Package Based On Collaborative Filtering Algorithm

With the deepening reform of China’s power market,the key content and development path of the national power reform have been clarified.Electric power reform requires opening up the supply of the selling side,encouraging the introduction of competition mechanism in the selling link,thus restoring the commodity attributes of electric power products,thus giving users the right to choose freely.Therefore,on the one hand,it is necessary to encourage the cultivation of market-oriented power selling companies and make rational planning for them;on the other hand,how to recommend personalized power selling packages to power users is related to the competitive interests of power selling companies in the electricity market.This paper does the following three aspects of research.(1)According to the problem of data sparseness and cold start in collaborative filtering algorithm,this paper proposes a collaborative filtering algorithm based on weighted similarity explicit-implicit feedback.Firstly,the algorithm excavates the historical data of power users,analyses their implicit behavior preference information,and combines with the active evaluation of power users for the power selling packages.The results show that the improved algorithm can achieve good performance.It can effectively solve the problems of data sparseness and cold start in collaborative filtering algorithm.(2)In view of the difficulty of recommendation caused by the variety of power selling packages in the recommendation system,this paper considers clustering the attributes of power selling packages,mining the historical information of power users,and effectively analyzing the behavior preferences of power users.By introducing time weights and correlation factors,and using the optimized collaborative filtering algorithm,we can achieve the accurate recommendation of the package and avoid the reduction of the effect of the recommendation caused by the complexity of the package attributes.(3)In view of the large amount of tasks that need to be dealt with in the process of recommendation,and the problem that users’ needs can not be responded to and processed in time,this paper mainly designs and analyses the recommendation algorithm,data collection and storage,and system user interaction,so as to achieve the purpose of accurate and timely recommendation for power users.