As the most widely used recommendation algorithm, the traditional collaborative filtering can hardly solve the problems of data sparsity, cold start and different user interests. Aiming at these three problems, a personalized recommendation method based on weighted bipartite graphs was proposed to solve the problem of recommendation of personalized design scheme. A weighted bipartite graph was used to classify users and schemes based on user characteristics and scoring features to reduce the data sparsity and form a user-scheme rule base. A collaborative filtering algorithm based on weighted networks was used to calculate the improved similarity of user features between new user characteristics and user-scheme rules in the library, and recommended by the Top-N method to screen high similar solution sets to solve the problems of cold starting and different user interests. Finally, compared with the traditional collaborative filtering algorithm and weighted bipartite graph, the validity and practicability of the method were proved.