I have data on reservation requests for hotels (your booking information:searching date, check-in, check-out, # of rooms and etc. on hotel booking websites) and am trying to do some analysis on one variable called advance day (i.e. check-in date subtract booking date). My goal is to predict, say using 7 days' of data, the most frequent requests on the next day so that I can prepare room prices in advance.

I tried analyzing hourly modes because that's the stat that describes "most frequent" and my data are not regularly spaced. After creating a time series in R, I tried fitting several models, like ARIMA, HoltWinters smoothing and ETS. Below is the example result of forecasting using HoltWinters smoothing. I am not satisfied with my results because none of the forecasts match with my future (8th day's) modes perfectly. I guess it's because these modes are independent? Everybody search independently? Am I missing anything? Is analyzing modes a good idea?

I have no idea what other ways I could try. Any suggestions besides analysis on modes? Any help would be greatly appreciated. Thanks!


the red line is my fitted values.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.