Improved Affinity Propagation Clustering for Business Districts Mining

Abstract

Business districts serve as basic structures for understanding the organization of real-world economic network. Discovering these business districts in cities establish new types of valuable applications that can benefit end users: Business investors can better identify the proximity of existing business districts and hence, can contribute a better future planning for investing. In this paper, we propose improved affinity propagation clustering for business districts mining. Given check-in data, whose geography information represents business venues' location, we introduce a affinity propagation clustering algorithm(AP), a basic solution, to cluster venues. This strategy requires that real-valued messages are exchanged among business venues until a set of centers and corresponding business districts gradually emerges. However, the computational complexity of AP is affected by the scale of input. And it's not adaptive for random distribution of venues when mining business districts. To conduct business districts mining efficiently, we introduce a pruning method, termed as PAP. And then present merging based mine approach, termed as MAP. We conduct experiments from Yelp data, and experimental results show that our proposed method outperforms the basic solutions and resolves the problem well.