The rapid growth of technology has changed the dynamics in which consumers socialize and make their purchasing decisions. The volume of online reviews has grown rapidly over the past decade, leading the peer groups of consumer to carry a disproportionate weight in the purchasing decision process. The sheer volume of reviews can be a daunting task for an operator to attempt to incorporate the reviews in their analysis. Sentiment analysis allows for large volumes of consumer reviews to be processed in a relatively easy, and time sensitive manner. The information contained in these reviews, the sentiment score, is the same feeling hospitality consumers are gathering from other consumers prior to making their purchasing decision. To demonstrate the importance of these reviews, this study will seek to model the directional change of a company's stock price using the sentiment of the consumer's reviews as the primary predictor. Support Vector Machines will help to classify a year's worth of consumer reviews on nine distinct properties of a publicly traded Las Vegas gaming/hotel company. This is then modeled using ARIMA modelling techniques to forecast an out-of-time sample, and the accuracy will be assessed by showing that the results being due to random change are minimal. The model is able to accurately predict 28 out of 39 time periods in the out of time sample, which has less than a .0047 probability of being due to random chance.
展开▼
