Clustering Analysis of Tourism Regions in Bali Province Using Similarity Matrix and Log-Log Regression Methods

Abstrak

The tourism accommodation sector in Bali Province has grown rapidly; however, its distribution across regencies/cities remains highly uneven, motivating the need for a rigorous clustering method to map accommodation-capacity similarities among regions. This study aims to group the nine regencies/cities of Bali Province based on the similarity of ten types of accommodation and tourism-support facilities (star hotels, non-star hotels, villas, apartments, homestays, camping grounds, restaurants, bars, spas, and water tourism facilities), and to examine their relationship with Gross Regional Domestic Product (GRDP). The method combines log-log regression to test the elasticity of star hotel capacity on GRDP, squared Euclidean distance and maximum distance (dmax) to compute a similarity index between every pair of regencies/cities, and the Kruskal-Wallis test to validate differences in GRDP across zones. The regression results show that the star hotel variable has a very strong and statistically significant effect on GRDP (R² = 84.89%; p = 0.00042). The similarity matrix reveals extreme inequality: Badung Regency shows a similarity value that is essentially zero relative to every other regency/city, whereas Jembrana and Tabanan demonstrate a very high similarity value (0.9547). Based on tertile scores, the regions are classified into three zones: Zone I, representing high capacity (Badung and Denpasar); Zone II, representing medium capacity (Tabanan, Gianyar, and Buleleng); and Zone III, representing low capacity with the highest homogeneity (Jembrana, Klungkung, Bangli, and Karangasem). The Kruskal-Wallis test confirms that the GRDP differences across zones are statistically significant (H = 7.00; df = 2; p = 0.0302). These findings indicate that tourism-equalization policies in Bali should be designed differently for each zone rather than applied uniformly across all regions.