Link prediction mainly seeks hidden connections on a social network. In a recommendation system, how to recommend a highly relevant object for users is a primary task. Here, we extracted several features from edges. Scoring functions are defined as cosine similarity, Jaccard coefficients, etc.