Recommendation Algorithm based on Link Prediction and Domain Knowledge in Retail Transactions

Abstract

In this paper, we propose a new recommendation algorithm, which extends the idea of linkage measure to recommendation in bipartite network, and incorporate domain knowledge with topological property in recommendation process. Through calculating domain similarities between products, we weigh the products recommended to potential customer with larger weights, whose categories are more close to the categories meeting with users’ preference, so as to improve the recommendation quality. Our preliminary experimental results based on a retail transaction dataset indicate that domain-based link prediction measures achieved better performance than general linkage measures algorithms.