A Personalized Time-Bound Activity Recommendation System
Activity-listing services like TripAdvisor, Foursquare and Facebook events make use of community opinions/reviews to help users identify ‚??points of interest‚?? from a large search space and typically make use of collaborative filtering algorithm. These algorithms analyze reviews from a group of people to find correlations between users and/or items in order to suggest top items to a querying user. The main issues with collaborative filtering are scalability and sparsity. Data sparsity creates a problem in scenarios where, a new location or activity that has not been reviewed by enough users is less likely to turn up as a recommendation. Also, users with unique interests do not benefit from such a model since the recommendations are irrespective of how relevant they may be to the user. There is also no system known to the authors, which recommends activities taking into consideration the time available with the user. Hence, a content based approach to generate a time-bound, relevant, personalized feed of ‚??activities‚?? in the nearby area is proposed in this paper. If the user is a die-hard trekker, into indie bands, or dislikes pub hopping, the system will always take that into consideration. Similarly, if someone only wants to spend a couple of hours and is interested in outdoor activities, he/she would be recommended ice-skating instead of hiking.
Personalization, Recommendation System, Information Retrieval, Feature Extraction, Machine Learning, Natural Language Processing, Automatic Travel Guide, Clustering, Sentiment Analysis, Time Estimation