In this paper, we propose a new clustering method to simultaneously analyze multi-dimensional spatial data of tourists’ floating population, represented in different age and gender groups. Based on graph theory, the whole geographical region is simplified into grid graphs, and we use minimum cost to partition them into subgraphs. To evaluate our method, we apply the proposed cluster analysis to examine visit patterns of tourists in Jeju Island, which is a popular tourist attraction in Korea, and compare it to results from the traditional heat map analysis. Through this comparison, we show that our new method is able to discover abnormal pattern of population density that are not found in heat maps. This research has two major implications. First, methodologically, we propose a clustering algorithm for geographic data which is faster to implement and enables the researcher to find patterns across different areas. Second, managerially, we can use the insights obtained from this research to provide the right products and services to different groups of tourists.