In the process of digital transformation, electric power enterprises can effectively reduce their operating costs and provide customers with more convenient and efficient services by building a multi- channel service system and making full use of the Internet + physical channels. Under the above background, a power marketing channel diversion strategy based on improved collaborative filtering algorithm is proposed. Firstly, the customer attribute data matrix is constructed, and the matrix decomposition algorithm is used to recover the missing data in the original customer attribute matrix, and the K-means algorithm is used to cluster the customer attributes. Then, using the customer mixed type attribute dissimilarity measure, through the user based collaborative filtering recommendation algorithm, the target customer’s K- nearest neighbor matrix is found, and the diversion strategy of travel alienation is formulated.Finally, taking 100 000 payment work order data as an example, the influence of customer attribute matrix filling, different measurement methods and the number of nearest neighbors on the drainage accuracy are analyzed, and the effectiveness and feasibility of the proposed algorithm are found.