Publicação
Using machine learning and deep learning techniques to predict hotel booking cancellations
| Resumo: | Booking cancellations are one of the most concerning topics in the hotel industry. From the revenue lost from canceled bookings, the uncertainty created in occupancy rates or even the inaccuracy of the hotel’s resource management, the negative impacts of booking cancellations are numerous, making it a sensitive subject for those working in the field. To combat these obstacles, hotels have been forced to adopt strategies such as creating cancellation policies or managing the sale of their rooms with overbooking. Although these strategies are effective in controlling the negative impacts caused by booking cancellations, they can also have a negative impact on a hotel’s reputation and customer satisfaction, especially when it comes to overbooking. This project aims to present a solution to help identify possible booking cancellations by creating predictive models using machine learning and deep learning methodologies. Based on the CRISP-DM paradigm and a real and current dataset of a Portuguese hotel chain, this project addresses various techniques used in the creation of binary classification models, from data processing to modeling and model evaluation. Given the constant superiority of machine learning algorithms in models based on tabular data, making deep learning solutions still not very viable in this field, this project also seeks to test some deep learning architectures that have been developed in recent years. The machine learning model created using the XGBoost algorithm showed promising results, with an AUC of 0.77 and a precision and recall of 0.93 and 0.54, respectively. Although the results are not yet excellent, they show that the use of this solution is viable and has room for further improvements. |
|---|---|
| Autores principais: | Sousa, Diogo Abrantes de |
| Assunto: | Machine learning Deep learning Predictive analytics Booking cancellations |
| Ano: | 2024 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | Booking cancellations are one of the most concerning topics in the hotel industry. From the revenue lost from canceled bookings, the uncertainty created in occupancy rates or even the inaccuracy of the hotel’s resource management, the negative impacts of booking cancellations are numerous, making it a sensitive subject for those working in the field. To combat these obstacles, hotels have been forced to adopt strategies such as creating cancellation policies or managing the sale of their rooms with overbooking. Although these strategies are effective in controlling the negative impacts caused by booking cancellations, they can also have a negative impact on a hotel’s reputation and customer satisfaction, especially when it comes to overbooking. This project aims to present a solution to help identify possible booking cancellations by creating predictive models using machine learning and deep learning methodologies. Based on the CRISP-DM paradigm and a real and current dataset of a Portuguese hotel chain, this project addresses various techniques used in the creation of binary classification models, from data processing to modeling and model evaluation. Given the constant superiority of machine learning algorithms in models based on tabular data, making deep learning solutions still not very viable in this field, this project also seeks to test some deep learning architectures that have been developed in recent years. The machine learning model created using the XGBoost algorithm showed promising results, with an AUC of 0.77 and a precision and recall of 0.93 and 0.54, respectively. Although the results are not yet excellent, they show that the use of this solution is viable and has room for further improvements. |
|---|