People are living in the background of the rapid development of the Internet, and the network data generated has also shown an explosive growth. For this reason, the need for access to personalized data has become more apparent. Therefore, recommendation systems have become one of the hot topics of discussion in recent years. The traditional recommendation algorithms, such as collaborative filtering recommendation algorithms, ignore the influence of time factor on the rating, and also have the problem of low recommendation accuracy on the sparse data. Based on the above two problems, this paper proposes an Alternating Least Squares (ALS) recommendation algorithm based on Kullback-Leibler divergence (KL) and time weighting, which uses KL divergence to calculate item similarity while incorporating a time factor, referred to as KL-TW-ALS algorithm. Three experiments were also conducted on the Spark platform to compare with the item based collaborative filtering algorithm and the ALS-based collaborative filtering algorithm, showing that the optimized ALS algorithm has improved accuracy and performance.