The development of Internet technology has changed the way of travel for tourists. More and more people get information from the Internet, which has brought great changes to the tourism industry. The same is true of hotels as a key industry in the tourism industry. How to stand out from the fiercely competitive market and meet the diverse needs of customers is an important issue at present. This paper takes the data of more than 140,000 Chinese reviews of five-star hotels in Chongqing on the "Trip.com Group" as the research object, uses machine learning algorithms to conduct topic mining and sentiment analysis, and analyzes customer needs and preferences according to the characteristics of online review texts. The results show that in addition to the location, meals, decoration, cleaning, housing facilities and other hardware facilities of the hotel, consumers pay more attention to the software facilities such as the service and attitude of hotel staff. The research reduces the risk of tourists’ purchase decision, provides an important reference for hotel managers to carry out management and marketing strategies, and expands the application of review big data and natural language processing in the hotel industry in the analysis of big data.