Discovering customer types using sales transactions and product availability data of 5 hotel datasets with genetic algorithm

• Published in: Journal of revenue and pricing management Vol. 19; no. 6; pp. 386 - 400
• Main Authors: HajMirzaei, Milad, Ziarati, Koorush, Nikseresht, Alireza
• Format: Journal Article
• Language: English
• Published: London Palgrave Macmillan UK 01.12.2020; Palgrave Macmillan
• Subjects: Business and Management; Genetic algorithms; Hotels & motels; Research Article; Revenue management; Revenue Management; Evolutionary algorithms; Genetic algorithm; Optimization
• ISSN: 1476-6930; 1477-657X
• DOI: 10.1057/s41272-020-00245-3

Demand forecasting is an integral part of every revenue management system. Demand raises from customers; therefore, knowing customers and their behavior is essential in this regard. Similar customers are grouped into a customer type. Discovering customer types from sales transactions and product availability data is a challenging topic. The basic idea of this paper is to use metaheuristic’s capability in exploring the search space instead of mathematical demand models in the research field of market discovery. In this work, a genetic algorithm is proposed to find efficient customer types. The main challenge of using a genetic algorithm in this field is to choose the proper fitness function. We use a two-phase fitness function for this problem to evaluate feasible and infeasible solutions. To evaluate the proposed method, a real publicly available dataset of five hotels is used. The results indicate that the genetic algorithm improves approximately 10% of the log-likelihood value of other proposed approaches with equal or lower number of customer types.