With the advent of the era of big data, the problem of information overload has become particularly serious. The recommendation system can provide personalized recommendation services for users by analyzing users' basic information and users' behavior information. How to push information accurately and efficiently has become an urgent issue in the era of big data. Based on the Alternating Least Squares (ALS) collaborative filtering recommendation algorithm, this paper reduces the loss of the invisible factor item attribute information by merging the similarity of the item on the loss function. At the same time, the cold start strategy is introduced into the model to complete the recommendation. The algorithm is implemented on the Spark distributed platform and single node by using the Movie Lens dataset published by the GroupLens Lab. The experiment results show that the proposed recommendation algorithm can preferably alleviate the data sparsity problem compared with the traditional recommendation algorithm. Moreover, the algorithm improves the accuracy of recommendation and the efficiency of calculation.