Using customer segmentation to build a hybrid recommendation model

Abstract

The growing trend in leisure tourism has been closely followed by the number of hospitality services. Nowadays, customers are more sophisticated and demand a personalized and simplified experience, which is commonly achieved through the use of technological means for anticipating customer behavior. Thus, the ability to predict a customer’s willingness to buy is also a growing trend in hospitality businesses to reach more customers and consolidate existing ones. The acquisition of a transfer service through website reservation generates data that can be used to perform customer segmentation and enable recommendations for other products or services to a customer, like recreation experiences. This work uses data from a Portuguese private transfer company to understand how its private transfer business customers can be segmented and how to predict their behavior to enhance services cross-selling. Information extracted from the data acquired with the private transfer reservations is used to train a model to predict customer willingness to buy, and based on it, offer leisure services to customers. For that, a hybrid classifier was trained to offer recommendations to a customer when he/she is booking a transfer. The model employs a two-phase process: first, a binary classifier asserts if the customer who’s buying the transfer would eventually buy a service experience. In that case, a multi-class model decides what should be the most likely experience to be recommended.