Non-negative Matrix Decomposition-Based Power Package Recommendation Method

It is a common way for price comparison websites to recommend electricity sales packages based on users’ evaluation information. However, due to the wide variety of electricity sales packages and the limited number of users of electricity sales packages, the user's evaluation information on electricity sales packages will be missing, which will lead to a large deviation in the recommendation effect. In this paper, a non-negative matrix decomposition-based power package recommendation method is proposed. Firstly, the scoring data of original users on their used electricity sales packages are collected, and an original scoring matrix for the users can be obtained; Secondly, the non-negative matrix decomposition method is used to decompose the original scoring matrix into user feature matrix and package feature matrix; Thirdly, under the condition that the distance between the original matrix and the two decomposed matrices’ product is the minimum, the elements of the two matrices are iterated to obtain the users’ predictions on the packages; Then, the obtained prediction matrix is recommended in full order, which realizes the accurate filling of the original scoring matrix and helps the electricity sales company to achieve the accurate recommendation of the electricity sales package. Finally, numerical studies are conducted, and the results show that the non-negative matrix decomposition-based power package recommendation method has a high accuracy, and can provide technical support for electricity sales companies to recommend electricity sales packages.