Hotel room sales prediction for a travel agency

Official URL: https://risc01.sabanciuniv.edu/record=b2579098_(Table of contents)

Predicting sales can be extremely beneficial to the tourism industry because it allows planners and managers to foresee future performance. This allows travel agencies to make more informed decisions about facilities, improve their contracts with more favorable terms, and offer better deals to customers in order to maximize their revenue and minimize their loss. Sales prediction enables travel agencies to adjust prices based on facility supply and customer demand, focus on sales to different demographics or change their marketing strategy to attract more customers of a specific segment. In this thesis, we compare various statistical and machine learning models on several datasets containing basic information on hotels, hotel features, and points of interests (PoI) near hotels in order to present a robust and accurate solution to hotel room sales prediction problem based on real-life data from one of the largest travel agencies in the Turkish tourism market. The results show that machine learning regression models have a great potential for hotel sales prediction. Random Forest Regression is outstanding with the highest goodness of fit and Support Vector Regression is good at accuracy values in the majority of the cases. Besides, there is a significant difference between the predictive performances by using All Segments and Two Adults Segment datasets. Additionally, the results with PoI datasets are also as good as the results without PoI datasets.