Collaborative Filtering (CF) is one of the techniques in recommender system which exploits information of user preference in the form of ratings of items and produce recommendation based on the similarity of ratings pattern. In a collaborative filtering approach is divided into two classifications: memory-based and model-based, both have their respective advantages and disadvantages. In making an accurate MemoryBased CF depends on dataset used, the data which generally is sparse makes predictions becomes less optimal, to handling sparse rating data we propose preprocessing to fills empty rating scores with values based on matrix factorization. The process will be exploited reduce by estimation calculated in matrix data. The research involves Memory-Based CF, with and without preprocessing to analyze both prediction performances. Our result show indicate that the proposed approach with preprocessing increased prediction accuracy than existing approach without preprocessing.