Hotel Classification Using Meta-Analytics: A Case Study with Cohesive Clustering

Abstract

We present a new clustering algorithm for handling complexities encountered in analysing data sets of hotel ratings and analyse its performance in a clustering case study. In the setting we address, business constraints and coordinates (among other individual attributes of objects) are unknown and only distances between objects are available to the clustering algorithm, a situation that arises in a wide range of clustering applications. Our algorithm constitutes an application of meta-analytics, in which we tailor a metaheuristic procedure to address a challenging problem at the intersection of predictive and prescriptive analytics. Our work builds on and extends the ideas of our clustering algorithm introduced in previous work which employs the Tabu Search metaheuristic to assure clusters exhibit a property we call cohesiveness. The special characteristics of the present hotel classification problem are handled by integrating our previous method with a new form of hierarchical clustering. Our computational analysis discloses that our algorithm obtains clusters that exhibit greater cohesiveness than those produced by the classical K-means method.