Vacation Recommendation System Using New York Times Articles

A good relaxing vacation is required for every individual. But when it comes to vacation planning, one usually spends maximum amount of time finalizing the location. In the process of finalizing the location, one follows a recommender model. In this research, it is aimed to build a recommendation model that can predict the desired destination based on user inputs to certain topics which can describe places in unique way. This paper does the recommendation using Cosine similarity after generating feature matrix from web scrapped New York Times articles using NMF and TF-IDF and the results obtained by this approach is compared with other feature extraction techniques. Out of the four mentioned techniques, NMF with TF-IDF showed more efficient feature vector distribution. Using this feature vectors and user input, cosine similarity was computed to give recommendations to user. The recommendation model was then tested using the manual feature sets, which shows that the recommender model can predict 12 out of 13 places correctly.