Research on multi-source POI data fusion based on ontology and clustering algorithms

Traditional point-of-interest (POI) data are collected by professional surveying and mapping organizations and are distributed in electronic maps. With the booming Internet and the development of crowdsourcing, the POI data defined in various formats are issued by some Internet companies and non-profit organizations. Due to the multiple sources and diverse formats of POI data, some problems occur in the data fusion process, such as conceptual definition differences, inconsistent classification, inefficient fusion algorithms, inaccurate fusion results, etc. To overcome the challenges of multi-source POI data fusion, this paper proposes a standardized POI data model and an ontology-based POI category system. Furthermore, a fusion framework and a fusion algorithm based on a two-stage clustering approach are proposed. The proposed method is compared with existing algorithms using datasets of different sizes, including POI surveying and mapping data from Kunming, China, Weibo check-in POI data, and real estate POI data. The experimental results demonstrate that the fusion effects of the proposed algorithm are superior to those of existing algorithms in terms of different evaluation indexes and operational efficiency.