Evaluating Clustering Algorithms and Autoencoders for Segmenting Customers in the Tourism Domain

Nowadays, recommender systems are used to give customers personalized recommendations on where to go on vacation, based on social media data. However, the collaborative filtering algorithms which are used by these recommender systems face scalability problems because of the exponential growth in the amount of social media data. Therefore, in this thesis we compare different clustering algorithms in order to segment customers in the tourism domain. We also use autoencoders as a dimensionality reduction technique for the considered data sets, with the aim of filtering out noise in the data and improving the performance of the clustering algorithms. The results show that the k-means algorithm, the k-medoids algorithm, and the hierarchical clustering algorithm with either complete or Ward linkage are the clustering algorithms which in general perform the best. Which of these algorithms to use depends on the data set. The results also show that autoencoders are able to improve the performance of the clustering algorithms, especially for large data sets.