Advancement Through Research
Identifying tourist hotspots and its explanatory variables using limited information sources are quite routine, although these are key clues for tourist planning and city management perspectives. Unfortunately, most of current practices are not cost-effective and quite outdated due to lacking of utilizing diverse sources. So far, only limited studies have been tried to discover and define tourist hotspots and its contributing factors with considering a few of data sources such as tourist counting in attraction sites. With leveraging state-of-the-art technology, this study successfully applied diverse machine learning techniques such as principal component analysis (PCA), Random forest, K-Means clustering, and t-distributed stochastic embedding (t-SNE) algorithms in conjunction with two optimization techniques including Hill climbing and genetic algorithm (GA). Six major explanatory variable groups were collected and incorporated into 100m grid-level GIS unit to this study. Which are demographic, time of day, ground truth inbound visitor counting, geographical information, accessibility from transportation network, and 18 different local business information by types. As case study, Jeonju-City, South Korea was chosen. Finally, clustered results and its contributing factors tell how travelers visited and moved so differently with unique interest by time and space. The results can be expected to contribute more vivid and insightful clues to city management and tourist planning.
Machine Learning, t-SNE Clustering, Tourist Hotspots, Explanatory Analysis, BigData
