Abstract: Aiming at the problem of data sparsity in traditional collaborative filtering algorithm, this paper proposed a collaborative filtering algorithm combined with improved weighted Slope One.The algorithm firstly used the score prediction value calculated by the improved slope one algorithm to effectively fill in the initial user-score matrix, and then made prediction and recommendation through the memory-based collaborative filtering algorithm on the new score matrix.The matrix filled by the improved Slope One algorithm not only greatly reduced the sparsity of the scoring matrix, but also avoided the problem that the backfill data is too single.Simulation experiments on movieslens-100k datatset show that the improved algorithm effectively reduces the MAE and RMSE, and improves the recommendation accuracy while alleviating the problem of data sparsity.