Existing personalized recommendation systems are facing many problems such as cold start, data sparseness and high complexity. Users' interests exist more widely and are more personalized compared with purchasing history in traditional recommendation systems. Thus, applying the interest graph in the recommendation process can make up certain shortages. This paper builds the mechanism of a user-interest-goods recommendation which is a tripartite network recommendation, and finally on the basis of the interest graph, it proposes the IGGRA (Interest Graph-based Goods Recommendation Algorithm) to recommend goods to customers. The empirical study demonstrates that the IGGRA is better than the collaborative filtering in accuracy.