Clustering of hotel customers by K-means

Abstract

In this study, hotel customers are clustered by using k-means partitioning technique. Eight attributes are used in clustering the customers which are gender, age, frequency of stay, length of stay, total expenditure, number of companions, season, and location / addres. Data collected for randomly selected 100 customers of a 5-star hotel in Antalya region are used in the analysis. Everett's approach is preferred in determining the preliminary number of clusters. Clustering operation is terminated with respect to the minimization of total mean squared error. A simple special software program is developed to carry out the analysis. Fuzzy k-means and expectation maximization methodologies are also used in the study in compariosn with the k-means algorithm. Cluster validation is achieved through the Dunn and Davies-Bouldin Indices. Further, the results are verified according to a target value calculated by using the given indices. Operation stages in the software developed and the graphical presentations of the data are also supplied in the paper. The research as a whole bears the distinction of being the first in the literature.