Multi-Cluster DBSCAN for Analysing Tourism Data

This paper presents a novel multi-clustering approach using the optimized DBSCAN algorithm to analyze tourism data with diverse and complex attributes.Through feature engineering, we transformed raw data into more informative representations.By fine-tuning DBSCAN parameters and using five distance metrics, the optimal clustering configuration for 347 destinations was identified.The experimental results show that DBSCAN significantly outperformed K-Means and FCM, achieving perfect Silhouette Scores for categorical features, type, and popularity.In addition, DBSCAN demonstrated superior cluster separation and density, as reflected by lower DBI and higher CHI values.For geographic features, DBSCAN achieved the highest Silhouette Score (0.54940), surpassing K-Means (0.48374) and FCM (0.45724), despite challenges in clustering spatial data.DBSCAN also recorded the shortest computation time, highlighting its efficiency.This research underscores the importance of feature engineering and parameter tuning in gaining deeper insights and improving the clustering process for tourism data analysis.